A Price-Based Demand Response Scheme for Discrete ... - MDPI

energies Article

A Price-Based Demand Response Scheme for Discrete Manufacturing in Smart Grids Zhe Luo, Seung-Ho Hong * and Jong-Beom Kim Department of Electronic Systems Engineering, Hanyang University, 1271 Sa 3-dong, Sangnok-gu, Ansan-Si, Gyeonggi-do 426-791, Korea; [email protected] (Z.L.); [email protected] (J.-B.K.) * Correspondence: [email protected]; Tel.: +82-31-400-5213 Academic Editor: Paras Mandal Received: 30 May 2016; Accepted: 9 August 2016; Published: 17 August 2016

Abstract: Demand response (DR) is a key technique in smart grid (SG) technologies for reducing energy costs and maintaining the stability of electrical grids. Since manufacturing is one of the major consumers of electrical energy, implementing DR in factory energy management systems (FEMSs) provides an effective way to manage energy in manufacturing processes. Although previous studies have investigated DR applications in process manufacturing, they were not conducted for discrete manufacturing. In this study, the state-task network (STN) model is implemented to represent a discrete manufacturing system. On this basis, a DR scheme with a specific DR algorithm is applied to a typical discrete manufacturing—automobile manufacturing—and operational scenarios are established for the stamping process of the automobile production line. The DR scheme determines the optimal operating points for the stamping process using mixed integer linear programming (MILP). The results show that parts of the electricity demand can be shifted from peak to off-peak periods, reducing a significant overall energy costs without degrading production processes. Keywords: demand response (DR); factory energy management system (FEMS); state-task network (STN) model; mixed integer linear programming (MILP); discrete manufacturing; automobile manufacturing

1. Introduction In recent decades, manufacturing has played a major role in society and economies, e.g., it employs 11% of the workforce and constitutes 12% of the economy in United States [1]. Although different manufacturing processes have different energy consumption requirements, industrial electricity consumption is characterized by two distinguishing characteristics. First, industrial plants are one of the major consumers of electrical energy. A survey by the International Energy Agency indicated that the industrial sector accounted for 42.3% of the world’s electricity consumption in 2013 [2]. Second, industrial facilities often connect to the power grid at high voltage levels, e.g., by directly connecting to the transmission network [3]. In conclusion, reducing energy consumption costs and maintaining power grid stability are critical issues in factory energy management systems (FEMSs), and therefore introducing intelligent management of electricity demand, also known as demand side management (DSM), has been regarded as an effective way to deal with the aforementioned issues [4]. Based on energy consumption data measured and obtained from factory facilities, FEMSs undertake strategies to control the operations of equipment to minimize energy consumption without affecting the industrial production [5]. As a modern electric power grid infrastructure system, smart grids (SGs) integrate advanced information and communication technologies (ICTs) within the grids, and deliver electricity between suppliers and consumers through two-way digital technologies [6]. Meanwhile, through employing appropriate DSM programs, energy consumers can actively participate in the electricity market via demand response (DR) which is considered as one of effective mechanisms within DSM [7], so as to Energies 2016, 9, 650; doi:10.3390/en9080650

www.mdpi.com/journal/energies

Energies 2016, 9, 650

2 of 18

schedule electricity consumption in response to electricity prices from electrical utility and the current state of system reliability [8]. DR is both a significant component of the emerging SG paradigm and an important element of energy market design for stabilizing the potential supply power through which occurrences of energy market meltdown can be effectively prevented [9]. By scheduling energy consumption of industrial facilities, adjusting the supplement for demand, and postponing the need to invest in greater capacity, a DR strategy can make a great influence on balancing the electricity demand between supplier and consumer sides [10]. DR includes all potential modifications to the electricity consumption patterns of end-use customers that schedule the timing of energy usage, change the instantaneous demand during crucial periods, and change energy consumption patterns to respond to electricity prices [11]. Integration of DR into SGs can provide consumers with optimal energy allocation, minimize energy consumption costs, and curtail peak-hour usage, all of which help to avoid power quality degradation or blackout. Furthermore, it takes advantage of in-facility distributed energy resources (DERs), such as energy storage systems (ESSs), to play a further role in costs reduction and electrical reliability improvement. By integrating with DR in SG, FEMSs improve flexibility in factory management by realizing more functions, such as communicating with power utilities through two-way digital communication technology. Moreover, by scheduling industrial equipment to work in an optimal state, FEMSs integrated with DR can shift parts of the electricity demand from peak to off-peak periods, through which the energy consumption costs of consumers can be effectively reduced. In this study, a DR scheme was applied to a typical discrete manufacturing—automobile manufacturing—in which the stamping process was chosen as a scenario to be optimized. In order to demonstrate the impact of a DR scheme on an FEMS, we made the comparison for different test cases in which either the DR algorithm or the ESS was applied. The rest of the paper is organized as follows: the second section describes the related studies on FEMS. Section 3 introduces a brief background to this study, including a state-task network (STN) model and a DR scheme. The fourth section describes a modeling study of automobile manufacturing, in particular a DR model of the stamping process. The fifth section summarizes the DR scheme and the specific algorithm proposed in previous studies. The sixth section establishes an operational scenario and compares three test cases. The final section concludes this study and proposes future work in this area. 2. Related Studies In the industrial sector, FEMSs play an important role in factory management by managing industrial equipment to work at high energy efficiency and adopting strategies to ensure that the energy demands of factory facilities are met. Giacone and Mancò [12] described a methodology for building a framework that defines and measures energy efficiency more precisely, while Wang et al. [13] proposed a service-oriented architecture for FEMSs, on which smart factories could feasibly be established. Ikeyama et al. [14] proposed an approach for optimizing energy use based on the example of food plants, in which typical FEMS configurations were introduced. However, none of these studies provided a solution to integrate SG factors or DR schemes into the management of industrial manufacturing. Focusing on the discrete manufacturing area—an important subdomain of the industrial sector—Georgiadis et al. [15] implemented optimization scheduling for a concrete case study, in which inventory planning was optimized to ensure that customer demands for production were met. Chong et al. [16] proposed a simulation-based scheduling for dynamic discrete manufacturing, which dynamically adapted to changing manufacturing conditions to prevent disturbances in the production process. However, neither of them took energy consumption in the manufacturing process into account. In terms of energy efficiency, Cannata et al. [17] proposed a scheduling model to investigate the operation of a production system, however this method only allowed for a simple case as it was formulated with strict assumptions and only took into account the

Energies 2016, 9, 650

3 of 18

energy efficiency of a single piece of equipment without considering internal connections in the production line. Similarly, Sohail et al. [18] concentrated on reduction of energy consumption for the ICT infrastructure which is only one component of a discrete manufacturing facility, and hence did not provide a comprehensive solution for factory management. In summary, most studies on discrete manufacturing were concentrated on scheduling sequences of processes to guarantee a production process or on improving the properties of an individual machine to increase energy efficiency, but none have made full use of SG factors in the design of a DR scheme. In past decades, several related studies have reported on the benefits and feasibility of DR. Samad and Kiliccote [3] introduced some SG technologies for use in the industrial sector and provided an overview of automated DR. Mohagheghi and Raji [19] put forward a DR scheme for energy management in industry. In order to perform a thorough study of DR for sustainable energy systems, Shariatzadeh et al. [7] presented a review of DR resources which could be carried out under the SG framework, as well as made a classification of various existing DR schemes. These researches, however, only introduced the concept of a DR scheme without proposing a specific DR algorithm. Introducing real-time pricing as primary input, Karwan and Keblis [20] presented optimal operation planning for an air separation plant, but this solution requires a large computational burden for evaluating the feasible operating space by solving a nonlinear problem. Mitra et al. [21] proposed a DSM scheme for scheduling production planning of a continuous process with time-dependent electricity prices as the input. Similarly, this solution was also computationally intractable, because it obtained a surrogate model to evaluate the feasible region for operational constraints for the plant. To minimize total energy consumption costs, Shrouf et al. [22] proposed a mathematical model to optimize the production scheduling of a single machine. However, this model was only applied to an individual machine and did not consider the connection between machines, which is not a practical scenario for industrial facilities. Choosing the paper-making process as an example of continuous manufacturing, Pulkkinen [23] proposed a scheduling scheme and corresponding formulation with dynamic electricity prices as the input. However, although a mathematical expression for the scheme was provided, no consideration was made of logic control or information transmission between facilities. Castro et al. [24] proposed a continuous-time scheduling formulation for continuous manufacturing with time-of-use (TOU) electricity prices as the input. However, this method only considered a rather simple case of TOU electricity prices, and required a significant computational burden to solve the problem in a continuous-time formulation. In later research, Castro et al. [25] focused on optimizing the model to reduce the computational burden. However, the optimization solution only adequately takes effect under the TOU prices, as it aggregates periods with the same price level. In summary, these previous studies either required significant computational resources, restricting their practical applications, or did not provide an appropriate DR algorithm that included real-time prices. In this study, in combination with FEMS, we implemented the STN representation to discrete manufacturing, and formulated a detailed description of the integral structure of a factory management system and the internal structure of each task, providing a solution for practical production. On this basis, and with day-ahead hourly electricity prices as the input, a DR scheme using the computationally affordable mixed integer linear programming (MILP) optimization algorithm was applied to a typical discrete manufacturing system. To evaluate the performance of the DR scheme, a DR model was established with the automobile manufacturing as an example, and the operational scenario of the stamping process within automobile manufacturing was investigated. 3. Background Based on process manufacturing, Kondili et al. [26] proposed a general STN model which consisted of state nodes and task nodes. Georgiadis et al. [15] extended this model to discrete manufacturing, which differs from process manufacturing. In order to implement energy management on the consumer side, Ding et al. [27] expanded this general STN model to propose a DR scheme for industrial facilities in SG, while in further research, Ding et al. [28] proposed a specific DR algorithm and evaluated its

Energies 2016, 9, 650

4 of 18

performance with a typical continuous industrial process as a scenario. As shown in Figure 1a, the DR scheme has 2016, state9, nodes including feed, intermediate product and final product, and task nodes which Energies 650 4 of 18 can be subdivided into non-schedulable tasks (NSTs), for which the electricity demand cannot be and final and task nodes which can (for be subdivided non-schedulable tasks (NSTs), for scheduled andproduct, must be satisfied immediately technical into or economic reasons); and schedulable which the electricity demand cannot be scheduled and must be satisfied immediately (for technical tasks (STs), for which the electricity demand can be scheduled among pre-specified multiple operating or economic reasons); and schedulable tasks (STs), for which the electricity demand can be scheduled points. Figure 1b shows the simple internal structure of each task with the corresponding industrial among pre-specified multiple operating points. Figure 1b shows the simple internal structure of each equipment. In this DR scheme, the industrial equipment is classified into three types: non-shiftable task with the corresponding industrial equipment. In this DR scheme, the industrial equipment is equipment (NSE), should have a sufficient supply of electricity whenever it is required; shiftable classified into which three types: non-shiftable equipment (NSE), which should have a sufficient supply of equipment (SE), which can be switched on or off to balance the electricity supply; and controllable electricity whenever it is required; shiftable equipment (SE), which can be switched on or off to equipment which has multiple with different electricity demands. Based on the balance(CE), the electricity supply; and operating controllablestates equipment (CE), which has multiple operating states with different electricity demands. Based on the of each of type of equipment, NSTs will be properties of each type of equipment, NSTs will beproperties only composed NSEs, and STs can consist of one onlyNES, composed NSEs, and STs can consist of one or more NES, SE, or CE pieces. or more SE, orofCE pieces. Non-schedulable task

Feed (state node)

Schedulable task

Final Product (state node)

Intermediate Product (state node) (a)

NSE

NSE

NSE

SE

NSE

CE

(b)

Figure 1. (a) Simple representation of a demand response (DR) model and (b) internal structure of each

Figure 1. (a) Simple representation of a demand response (DR) model and (b) internal structure of each task with corresponding industrial equipment. task with corresponding industrial equipment.

The specific algorithm for the DR scheme is formulated using MILP, which consists of one objective function definedfor to minimize the energy of manufacturing processes, andconsists a series of of one The specific algorithm the DR scheme is cost formulated using MILP, which linear constraints that ensure achievement of the objective without degrading the production line. objective function defined to minimize the energy cost of manufacturing processes, and a series of specified the process modeling (including operating constraints material line. linearConstraints constraintsarethat ensurefor achievement of the objective without degrading the and production balance) and ESS modeling. Once the MILP problem is solved, the outputs of the DR algorithm Constraints are specified for the process modeling (including operating constraints and material include the optimal operating point for all STs in each time interval, the charging/discharging rate of balance) and ESS modeling. Once the MILP problem is solved, the outputs of the DR algorithm include ESS in each time interval, the amount of electricity purchased from the grid, and the total energy costs the optimal operatingprocess. point for all STs in each time interval, the charging/discharging rate of ESS in of manufacturing

each time interval, the amount of electricity purchased from the grid, and the total energy costs of 4. Modeling of Automobile Manufacturing manufacturing process. Based on the description of an automobile production line given in [29], an STN model for DR in automobile manufacturing was developed as shown in Figure 2. With the utility meter and utility gateway boundaries, the DR model is composed of a utilityline supply sideinand anan industrial demand Based onasthe description of an automobile production given [29], STN model for DR side. The manufacturing utility meter measures total electricity supplied to the 2. industrial in automobile was developed as shown in Figure With thedemand utility side, meterand andthe utility utility gateway exchanges information between a wide area network (WAN) and an industrial area gateway as boundaries, the DR model is composed of a utility supply side and an industrial demand network (IAN). side. The utility meter measures total electricity supplied to the industrial demand side, and the The utility supply side includes power stations to provide electrical energy for industrial utility gateway exchanges information between a wide area network (WAN) and an industrial area manufacturing through the electrical grid, and a utility data center to communicate with the network (IAN).demand side via the WAN. The industrial demand side is also composed of two industrial The utility supply side includes power process stationssystem to provide electrical energy the for whole industrial subsystems, the automobile manufacturing (AMPS) which includes manufacturing the process electrical grid, and abyutility datamodel, centerand to communicate with the industrial automobilethrough production represented the STN the FEMS which includes an energy management systemThe (EMS), ESS, the demand power source thecomposed power supply network, and demand side via the WAN. industrial sideswitch, is also of two subsystems, the IAN. the automobile manufacturing process system (AMPS) which includes the whole automobile

4. Modeling of Automobile Manufacturing

Energies 2016, 9, 650

5 of 18

production process represented by the STN model, and the FEMS which includes an energy Energies 2016, 9, 650 5 of 18 management system (EMS), ESS, the power source switch, the power supply network, and the IAN. Industrial Demand Side Automobile Manufacturing Process System (AMPS)

legends

Body parts

Feed Intermediate product

Pressure processing

Machining A

Final product

Spray paint

Steel plate

Non-schedulable task Schedulable task

M

The car body

Plate cutting

Machining B

Steel

Spare parts

Utility meter

Gear box assembly

GW

Gear box

Utility gateway

Casting

Auxiliary parts A

Semi finished parts A

Engine assembly

Semi finished parts B

Automobile

Auxiliary parts B Chassis assembly Chassis

Drive axle assembly Die-casting

Vehicle mounting

Engine

Drive axle Auxiliary parts C

Chassis parts

Tire

Utility Supply Side Factory Energy Management System (FEMS) Power Stations Grid Power flow

Power Supply

M

WAN Information flow

GW

Grid IAN

Switch

Data Center

ESS

EMS

Figure 2. 2. State-task State-task network Figure network (STN) (STN) model model for for DR DR in in automobile automobile manufacturing. manufacturing.

The AMPS consists of 17 state nodes that include six kinds of feeds represented as broken circles, The AMPS consists of 17 state nodes that include six kinds of feeds represented as broken circles, 10 kinds of intermediate products represented as solid circles, and one final product represented as 10 kinds of intermediate products represented as solid circles, and one final product represented as an inside-broken and outside-solid circle. Additionally, 12 task nodes include three NSTs represented an inside-broken and outside-solid circle. Additionally, 12 task nodes include three NSTs represented as double-bordered rectangles and nine STs represented as single-bordered rectangles. as double-bordered rectangles and nine STs represented as single-bordered rectangles. In the FEMS, the EMS manages and monitors the electricity demand of the whole automobile In the FEMS, the EMS manages and monitors the electricity demand of the whole automobile manufacturing process. It communicates with the utility supplier via the WAN to obtain day-ahead manufacturing process. It communicates with the utility supplier via the WAN to obtain day-ahead hourly electricity prices and transmit energy consumption information. Additionally, it schedules hourly electricity prices and transmit energy consumption information. Additionally, it schedules STs STs to work on the optimal operating points to manage the electricity demand of the whole industrial to work on the optimal operating points to manage the electricity demand of the whole industrial manufacturing, through which parts of the electricity demand can be shifted from peak to off-peak manufacturing, through which parts of the electricity demand can be shifted from peak to off-peak demand periods. By controlling the power source switch, the EMS schedules the ESS to charge demand periods. By controlling the power source switch, the EMS schedules the ESS to charge electricity from the electrical grid during off-peak periods, or discharge to supply energy for the electricity from the electrical grid during off-peak periods, or discharge to supply energy for the automobile manufacturing process during peak periods. automobile manufacturing process during peak periods. In the whole automobile manufacturing, the stamping process emphasized by a red rectangle in In the whole automobile manufacturing, the stamping process emphasized by a red rectangle in Figure 2, was selected as an example for precise analysis, since it not only determines the quantities Figure 2, was selected as an example for precise analysis, since it not only determines the quantities of of final production but also relates critically to the steady production of subsequent processes final production but also relates critically to the steady production of subsequent processes through through providing sufficient materials for them. Based on the significant feature of discrete providing sufficient materials for them. Based on the significant feature of discrete manufacturing manufacturing whereby each process deployed in the discrete manufacturing can be started or whereby each process deployed in the discrete manufacturing can be started or stopped individually stopped individually with a variety of production rates [30], separate processes shown in Figure 2 with a variety of production rates [30], separate processes shown in Figure 2 can be combined in can be combined in a parallel or serial way to expand the scale of the production. Hereon, this study a parallel or serial way to expand the scale of the production. Hereon, this study focuses on the focuses on the stamping process to represent DR management in automobile manufacturing; stamping process to represent DR management in automobile manufacturing; however, this approach however, this approach and results can be expanded to the whole process of automobile and results can be expanded to the whole process of automobile manufacturing. Applying a DR energy manufacturing. Applying a DR energy management scheme to the stamping process reduces the management scheme to the stamping process reduces the energy costs of industrial manufacturing energy costs of industrial manufacturing without degrading production processes. without degrading production processes. 4.1. The Response Representation Representation for for the the Stamping Stamping Process Process 4.1. The Demand Demand Response Figure 33 shows shows the the DR DR representation representation for for the the stamping stamping process, process, which which incorporates incorporates production production Figure processes from steel (raw material) to spare parts (intermediate products of the whole production processes from steel (raw material) to spare parts (intermediate products of the whole production line). The stamping process can be regarded as two subsystems: the automatic cutting system (ACS) corresponding to plate cutting task in Figure 2, and the parts production system (PPS) corresponding to machining A task or machining B task in Figure 2.

Energies 2016, 9, 650

6 of 18

line). The stamping process can be regarded as two subsystems: the automatic cutting system (ACS) corresponding to plate cutting task in Figure 2, and the parts production system (PPS) corresponding to machining A task or machining B task in Figure 2. Energies 2016, 9, 650 6 of 18

Steel

Energies 2016, 9, 650

6 of 18 ACS NSE A

SE SteelA

CE A

ACS NSE A

Steel plate with SE A appropriate size

CE A

Steel plate with appropriate size

PPS#1

PPS#2

PPS#3

PPS#4

PPS#5

NSE P 11

NSE P 21

CE P 31

NSE P 12 PPS#1

NSE P 22 PPS#2

PPS#3 CE P 32

NSE NSEP 23

CE P 33 CE P 31

P 43 CECE P 41

CE P 51CE P 53

CE P 32

CE P 42

CE P 52

CE P 33

CE P 43

CE P 53

NSENSE P 13

P 11

P 21

NSE P 12

NSE P 22

NSE P 13

NSE P 23

CE P 41

CE P 51

PPS#4 CE P 42 PPS#5

CE P 52

Spare parts

Spare parts

ACS: Automatic cutting system PPS: Parts production system ACS: Automatic cutting system PPS: Parts production system Figure 3. 3. STN STN model model for for DR DR in in the the stamping stamping process. Non-schedulable Non-schedulable tasks tasks (NSTs) (NSTs) include include PPS#1 PPS#1 Figure process. and 2. Schedulable tasks (STs) include ACS and PPS#3, 4 and 5. Each task is composed of three pieces Figure 3. STN model for DR in the stamping process. Non-schedulable tasks (NSTs) include PPS#1 and 2. Schedulable tasks (STs) include ACS and PPS#3, 4 and 5. Each task is composed of three pieces and 2. Schedulable tasks (STs) include ACS and PPS#3, 4 and 5. Each task is composed of three pieces of industrial industrial equipment with corresponding types. of equipment with corresponding types.

of industrial equipment with corresponding types.

The ACS is an ST and consists of three pieces of industrial equipment, which are classified as The The ACSACS is an andand consists of industrial industrialequipment, equipment, which are classified is ST an ST consistsofofthree threepieces pieces of which are classified as as NSE, SE and CE, respectively. PPS#1 and 2 (both of which correspond to the machining A task) are NSE,NSE, SE and CE,CE, respectively. PPS#1 whichcorrespond correspond machining A task) SE and respectively. PPS#1and and22(both (both of which to to thethe machining A task) are are NSTs and consist of three pieces of industrial equipment regarded as NSEs. PPS#3, 4, and 5 (each of and consist of three pieces industrialequipment equipment regarded NSEs. PPS#3, 4, and 5 (each of NSTsNSTs and consist of three pieces ofofindustrial regardedasas NSEs. PPS#3, 4, and 5 (each of which corresponds to thethemachining BB task) are STs STs and and consistof of three pieces of industrial corresponds machining task) areand three pieces of industrial whichwhich corresponds to thetomachining B task) are STs consistconsist of three pieces of industrial equipment equipment regarded as CEs. equipment regarded as CEs. regarded as CEs. 4.2. The The Internal Structure of Each Each Task ininthe the Stamping Process 4.2.Internal The Internal Structure of Each Task theStamping Stamping Process Process 4.2. Structure of Task in Figure 4 shows the internalstructure structure of the the ACS task, which consists of a local task monitoring Figure shows theinternal internal ACS task, which consists a local monitoring Figure 44shows the structure ofof the ACS task, which consists of aof local tasktask monitoring and and control system (TMCS), a local task energy management agent (TEMA), and threethree pieces of and control system (TMCS), a local task energy management agent (TEMA), and pieces of control system (TMCS), a local task energy management agent (TEMA), and three pieces of industrial industrial equipment—the differenttypes typesof of cutting cutting machine. industrial equipment—the different machine. equipment—the different types of cutting machine.

EMS

EMS

IAN

TMCS

TEMA

TMCS

TEMA

IAN Power Power

Cutting machine (NSE)

Cutting machine (NSE) Cutting machine (SE) Steel

Steel

Cutting machine Cutting machine (SE) (CE)

Cutting machine (CE)

Steel plate with appropriate size

Steel plate with appropriate size

Figure 4. The internal structure of the automatic cutting system (ACS). Figure 4. The internal structure of the automatic cutting system (ACS).

Figure 4. The structure of the automatic cutting system (ACS). TMCS monitors andinternal controls the statestate of each of equipment in that task. The The TMCS monitors and controls theoperational operational of piece each piece of equipment in that The TEMA monitors electricity consumption and controls the electric load of that task, while task. The TEMA monitors electricity consumption and controls the electric load of that task, while The TMCS information monitors and operational state of each piece of equipment in that task. exchanging withcontrols EMSthe through IAN. exchanging information with thethe EMS through IAN. The monitors informationelectricity exchangedconsumption from TEMA to and EMS includes of eachload task (NST or ST), thewhile The TEMA controlsthe thetype electric of that task, operating points of thewith task the and EMS the current storages exchanging information through IAN.of related states. While the operating point itself contains the information of thefrom consumption rate of states (feedthe or/and The information exchanged TEMA to EMS includes typeintermediate of each taskproduct), (NST or the ST), the operating points of the task and the current storages of related states. While the operating point itself

Energies 2016, 9, 650

7 of 18

The information exchanged from TEMA to EMS includes the type of each task (NST or ST), the operating points of the task and the current storages of related states. While the operating point itself contains the information of the consumption rate of states (feed or/and intermediate product), Energies 2016, 9, 650 7 of 18 the production rate of states (intermediate product or/and final product), and the corresponding electricity the task. productiondemand rate ofofstates (intermediate product or/and final product), and the corresponding In thedemand ACS task, these cutting machines cut the steel into steel plate with several appropriate electricity of the task. sizes,Inwhich willtask, be used materials for other tasks. The ACS assumed have appropriate three kinds the ACS theseascutting machines cut the steel into steelis plate with to several of cutting machines to NSE, SE and CE,The respectively. The first may be sizes, which will be corresponding used as materials for other tasks. ACS is assumed to type have(which three kinds of acutting general-purpose machine tool) regarded as an NSE, only one operating it has machines corresponding to NSE, SE and CE,has respectively. The first state type since (which maya long be a start-up time andmachine cannot tool) be stopped arbitrarily once The second type (which be general-purpose regarded as an NSE, haslaunched. only one operating state since it hasmay a long astart-up high-speed numerical control machine tool with a fixed strokes type per minute) regarded time precision and cannot be stopped arbitrarily once launched. The second (which may be a as an SE, has two work states (on or off) since ittool canwith be started stopped The third high-speed precision numerical control machine a fixedor strokes perimmediately. minute) regarded as an type (which may be a high-speed numerical control machineimmediately. tool with a variable strokes SE, has two work states (on or off)precision since it can be started or stopped The third type (which may regarded be a high-speed control machine tool different with a variable strokesrates per per minute) as a CE,precision which hasnumerical multiple operating states with consumption minute) regarded as a CE, which has multiple states electricity with different consumption of of materials, generation rates of products, andoperating corresponding demands, since itrates can be materials,from generation rates state of products, and sincestates it can be changed one working to another in acorresponding short time. Byelectricity managingdemands, the operating of SE changed from one working state to another in apoints short for time. managing operating states of SE and CE, the TEMA defines multiple operating theBy ACS task andthe reports the current state andeach CE,operating the TEMApoint defines multiple points theoperating ACS taskpoint and reports state of to the EMS operating through the IAN.for The with a the lowcurrent production of each operating point toenergy the EMS thewhile IAN.the Theoperating operatingpoint pointwith withaahigh low production production rate rate consumes electrical at athrough low rate, energy at at aahigh low rate. rate, Based while on thethe operating with a electricity high production rate consumes electrical energy dynamicpoint day-ahead prices from consumes electrical energy at a high rate. Based on the dynamic day-ahead electricity prices from the the utility supply side, the EMS makes optimal decisions on operating points for each task during utilitytime supply side, and the EMS optimal decisions ontooperating points for eachtotask during each each interval, then makes transmits this information the TEMA. According the information time interval, and then transmits this information to the TEMA. According to the information from from the TEMA, the TMCS controls each piece of equipment to work on the corresponding operating the TEMA, the TMCS piece of equipment to work oneach the corresponding operating state. The TEMA then controls transmitseach consumption data collected from piece of equipment to thestate. EMS The TEMA transmits consumption collected from each piece of equipment theaEMS through thethen IAN. As shown in Figuredata 5, the internal structure of each PPS task to has localthrough TMCS, IAN. As shown in Figure the internal structure ofmatched each PPSwith task corresponding has a local TMCS, a local athe local TEMA and three pieces5,of industrial equipment three stepsTEMA in the and three PPS task. pieces of industrial equipment matched with corresponding three steps in the PPS task. EMS EMS

IAN

TMCS

TEMA

TMCS

TEMA

IAN Power Power

Steel plate with appropriate size Steel plate with appropriate size

Rough turning Finish turning Milling punch punch punch Rough turning Finish turning (NSE) (NSE) (NSE) Milling punch punch punch (NSE) (NSE) (NSE)

Spare parts Spare parts

(a)

(a) (a) EMS EMS

IAN

TMCS

TEMA

TMCS

TEMA

IAN Power Power

Steel plate with appropriate Steel plate size with appropriate size

Rough turning Finish turning Milling punch punch punch Rough turning Finish turning (CE) (CE) (CE) Milling punch punch punch (CE) (CE) (CE)

Spare parts Spare parts

(b) (b)

(b) Figure and (b) (b) the the internal internal structure structure of of PPS#3, PPS#3, 44 and and 5. 5. Figure 5. 5. (a) (a) The The internal internal structure structure of of PPS#1 PPS#1 and and 2; 2; and

The PPS task basically involves three steps (rough turning, finish turning, and milling processes) to produce spare parts from steel plate. Each step matches with one kind of punch machine. During the rough turning, the steel plate is produced into semi-finished parts with low accuracy. These semifinished parts can then be precision machined through the finish turning step. After milling, the processed semi-finished parts are used to produce different spare parts with a variety of surfaces,

Energies 2016, 9, 650

8 of 18

The PPS task basically involves three steps (rough turning, finish turning, and milling processes) to produce spare parts from steel plate. Each step matches with one kind of punch machine. During the rough turning, the steel plate is produced into semi-finished parts with low accuracy. These semi-finished parts can then be precision machined through the finish turning step. After milling, the processed semi-finished parts are used to produce different spare parts with a variety of surfaces, Energiesis 2016, 650 product of the stamping process. 8 of 18 which the9,final Figure 5a shows the internal structure of PPS#1 and 2. They are NSTs, and each of them is Figure 5a shows the internal structure of PPS#1 and 2. They are NSTs, and each of them is composed of three pieces of equipment, which are all NSEs matched with the three steps of the PPS composed of three pieces of equipment, which are all NSEs matched with the three steps of the PPS task. Figure 5b shows the internal structure of PPS#3, 4 and 5. They are STs, and each of them is task. Figure 5b shows the internal structure of PPS#3, 4 and 5. They are STs, and each of them is composed composed of three pieces of equipment, which are all CEs corresponding to the three steps of the of three pieces of equipment, which are all CEs corresponding to the three steps of the PPS task. PPS task.

5. Demand 5. DemandResponse ResponseScheme Schemefor forFactory FactoryEnergy EnergyManagement ManagementSystems Systems This section section consists DR scheme proposed in [27]. TheThe firstfirst one This consists of oftwo twosubparts subpartstotointroduce introducethe the DR scheme proposed in [27]. discusses the pure DR scheme (excluding ESS) for an FEMS in SG. The second includes an ESS, which one discusses the pure DR scheme (excluding ESS) for an FEMS in SG. The second includes an ESS, can becan integrated into the TheThe DRDR scheme is is operated day-ahead which be integrated intoDR the scheme. DR scheme. scheme operatedononthe thebasis basisof of the the day-ahead hourly electricity electricity prices, prices, and and the the time time horizon horizon of of 24 24 hh is is discretized discretized into into 24 24 time time intervals intervals with with each each hourly interval having the same time duration of 1 h, as shown in Figure 6. interval having the same time duration of 1 h, as shown in Figure 6. time interval

0

1

2

3

21

22

23

24

24 h Figure 6. 6. Discrete-time Discrete-time representation representation of of the the DR DR scheme. scheme. Figure

There are some operational scenario requirements for practical application of the DR scheme. There are some operational scenario requirements for practical application of the DR scheme. First, tasks must take operations (including start operation, end operation, and changes between two First, tasks must take operations (including start operation, end operation, and changes between two different operating points) at the boundaries of time intervals shown in Figure 6. Second, the electricity different operating points) at the boundaries of time intervals shown in Figure 6. Second, the electricity demands and related states (i.e., the feed, intermediate product, and final product) of the NSTs and demands and related states (i.e., the feed, intermediate product, and final product) of the NSTs and STs STs during each time interval are assumed to be known a priori at the start of the time horizon. during each time interval are assumed to be known a priori at the start of the time horizon. Based on the inputs (including the day-head hourly electricity prices, the STN model of Based on the inputs (including the day-head hourly electricity prices, the STN model of industrial industrial manufacturing, and the operating information of each task), the DR algorithm selects the manufacturing, and the operating information of each task), the DR algorithm selects the optimal optimal operating point for each ST within each pre-specified time interval to efficiently manage operating point for each ST within each pre-specified time interval to efficiently manage energy energy consumption in the manufacturing process. In other words, DR can schedule STs to work at consumption in the manufacturing process. In other words, DR can schedule STs to work at different operating points through which parts of the electricity demand can be shifted from peak to different operating points through which parts of the electricity demand can be shifted from peak to off-peak periods. off-peak periods. 5.1. Demand Demand Response 5.1. Response Algorithm Algorithm This section section describes describes the the optimization optimization problem problem for for the the MILP MILP in in DR DR algorithm, algorithm, which which includes includes This the objective function and a series of constraints on processing modeling. the objective function and a series of constraints on processing modeling. 5.1.1. Objective Objective Function function 5.1.1. As shown shown in in Equation Equation (1), (1), the the objective objective function function is is defined defined as as the the total total energy energy cost cost of of industrial industrial As manufacturing, which which should should be be minimized: minimized: manufacturing, Cost = ∑ pt · Dt 𝐶𝑜𝑠𝑡 = ∑ 𝑝 ∙ 𝐷𝑡 t 𝑡

(1) (1)

Dt = ∑ dk,t ∀k, t ≥ 0 𝐷𝑡 = ∑ 𝑑 k 𝑘,𝑡 ∀𝑘, 𝑡 ≥ 0

(2) (2)

𝑡

𝑘

where pt is the unit electricity price purchased from the utility during time interval (t, t + 1), and Dt is the total electricity demand for the industrial manufacturing process from the electrical grid, which is equal to the summation of the electricity demand for each task within the same time interval, dk,t.

Energies 2016, 9, 650

9 of 18

where pt is the unit electricity price purchased from the utility during time interval (t, t + 1), and Dt is the total electricity demand for the industrial manufacturing process from the electrical grid, which is equal to the summation of the electricity demand for each task within the same time interval, dk,t . For an NST, dk,t is fixed during whole time horizon, while for a ST, dk,t is constrained by operating point, will be described later. Energies which 2016, 9, 650 9 of 18 5.1.2. Constraints Constraints on on the the Process Process Modeling Modeling 5.1.2. Constraints on on the the processing processing modeling modeling are are subdivided subdivided into into operating operating constraints constraints for for STs STs and and Constraints balance constraints constraints on on materials. materials. balance Operating Constraints for Schedulable Tasks 5.1.2.1. Operating constraints for Schedulable Tasks Assuming that task k is a ST as shown in Figure 7, it consumes state s to produce state s by Assuming that task k is a ST as shown in Figure 7, it consumes state s11 to produce state s22 by choosing one of n operating points during each time interval. Each operating point is composed choosing one of n operating points during each time interval. Each operating point is composed of of two sets of parameters. The first is a pair of parameters including the consumption rate of raw two sets of parameters. The first is a pair of parameters including the consumption rate of raw material crk,n,s1 , generation rate of product prk,n,s2 and corresponding electricity demand of that task material crk,n,s1 , generation rate of product prk,n,s2 and corresponding electricity demand of that task dk,n . The second is a binary variable Mk,n,t to indicate if task k operates at operating point n during time dk,n . The second is a binary variable Mk,n,t to indicate if task k operates at operating point n during time interval (t, t+1) or does not (if the task operates, M = 1, otherwise Mk,n,t = 0). Among all operating interval (t, t+1) or does not (if the task operates, Mk,n,t k,n,t = 1, otherwise Mk,n,t = 0). Among all operating points (from operating point 1 to operating point n), Equation (3) restricts that a ST k must operate at points (from operating point 1 to operating point n), Equation (3) restricts that a ST k must operate at only one operating point within the same time interval. only one operating point within the same time interval.

∑MMk,n,t = 11 k , n, t

n

∀tt≥0,  0,k∀k ST ∈ ST

(3) (3)

n

Task k (ST) Operating point 1 (crk ,1, s1 , prk ,1, s2 , d k ,1 ) | M k ,1,t Operating point 2 (crk ,2, s1 , prk ,2, s2 , d k ,2 ) | M k ,2,t

Cs1 , k ,t

Ps2 , k ,t

State s1

State s2 Operating point n (crk , n, s1 , prk , n, s2 , d k , n ) | M k , n,t

Figure Figure 7. Example Example of of an an schedulable schedulable task task (ST). (ST).

Figure 7 shows the relation between state s1 and state s2 through one ST to indicate the operation Figure 7 shows the relation between state s1 and state s2 through one ST to indicate the operation mechanism of the ST. In terms of any STs which can consume/produce states, Equations (4)–(6) mechanism of the ST. In terms of any STs which can consume/produce states, Equations (4)–(6) indicate indicate general relations between states and tasks. general relations between states and tasks. Cs , k , t   M k , n, t  crk , n, s s, t  0, k  Tc, s  ST

∀s, t ≥ 0, ∀k ∈ Tc,s ∩ ST

(4) (4)

, k, t k , n, t k , n, s p, s Ps,k,t s= M · prk,n,s ∀s, t ≥ 0, ∀k ∈ Tp,s ∩ ST ∑ n k,n,t

(5) (5)

Cs,k,t =

∑ Mk,n,t · crk,n,s n

n

P



M

 pr

s, t  0, k  T

 ST

n

dk,t Mk,n,t d· dk,n d = ∑M k, t

n n

k , n, t

k, n

∀tt ≥0,0, k ST ∈ ST  k∀

(6) (6)

where Cs,k,t is the quantity of state s consumed by task k during time interval (t, t + 1). Ps,k,t is the where Cs,k,t is the quantity of state s consumed by task k during time interval (t, t + 1). Ps,k,t is the quantity of state s produced by task k within time interval (t, t + 1). dk,t is the electricity demand of task quantity of state s produced by task k within time interval (t, t + 1). dk,t is the electricity demand of k during time interval (t, t + 1). In Equations (4)–(6), binary variables Mk,n,t are restricted by Equation (3) task k during time interval (t, t + 1). In Equations (4)–(6), binary variables Mk,n,t are restricted by to ensure that only one operating point of ST can be operational within the same time interval. Equation (3) to ensure that only one operating point of ST can be operational within the same time interval.

5.1.2.2. Balance Constraints on Materials In Equation 7, SAs,t is the storage of each state s at time boundary t (t > 0), which is equal to the storage at the previous time boundary t − 1 minus the total amount of state s consumed within time interval (t − 1, t) plus the total amount of state s produced within time interval (t − 1, t):

Energies 2016, 9, 650

10 of 18

Balance Constraints on Materials In Equation (7), SAs,t is the storage of each state s at time boundary t (t > 0), which is equal to the storage at the previous time boundary t − 1 minus the total amount of state s consumed within time interval (t − 1, t) plus the total amount of state s produced within time interval (t − 1, t): Energies 2016, 9, 650

SAs,t = SAs,t−1 −

∑

Cs,k,t−1 +

k ∈ Tc,s

SAs , t  SAs , t 1 

∑

Ps,k,t−1

∀s, ∀t > 0

10 of 18

(7)

k ∈ Tp,s

C

s , k , t 1



P

s , k , t 1

s, t  0

(7)

kT T Ssmin ≤ SAs,t k≤ Ssmax c, s

p, s

(8)

max (8)task k where Cs,k,t−1 and Ps,k,t−1 are the total amounts state s consumed and produced by all Ssmin  SAs ,of t  Ss (including NSTs and STs) within time interval (t − 1, t), respectively. Meanwhile, the storage is Where of state s consumed and produced by all Cs , k , t 1 and Ps , k , t 1 are the total amounts typically constrained within fixed limits (Ss min , Ss max ) as shown in Equation (8), which aretask the klower (including NSTs and STs) within time interval (t − 1, t), respectively. Meanwhile, the storage is storage bound and the upper storage bound of each state s, respectively.

typically constrained within fixed limits (Ssmin, Ssmax) as shown in Equation (8), which are the lower storage bound andScheme the upper storage with bound each state s, respectively. 5.2. Demand Response Integrated theofEnergy Storage System

In to the processing modeling section 5.2.addition Demand Response Scheme Integrated with theabove, Energy this Storage System integrates an ESS into the previous DR scheme. In this new scheme, the DR also schedules the operation each time In addition to the processing modeling above, this section integratesofanthe ESSESS intoduring the previous interval. For example, during off-peak periods, the DR schedules the ESS to charge energy from the DR scheme. In this new scheme, the DR also schedules the operation of the ESS during each time power grid, whereas during peak periods, it schedules the ESSthe toESS discharge toenergy supply energy interval. For example, during off-peak periods, the DR schedules to charge from the for industrial processes. this basis, the objective function should modified to for include powermanufacturing grid, whereas during peakOn periods, it schedules the ESS to discharge to be supply energy industrial manufacturing processes. On this basis, the objective function should be modified to include the interaction of the ESS, as well the constraints on ESS modeling need to be taken into account. the interaction of the ESS, as well the constraints on ESS modeling need to be taken into account.

5.2.1. Modified Objective Function 5.2.1. Modified Objective Function

With the interaction of ESS, the objective function is modified as Equation (9). Dtess is the total the interaction of ESS, the objective function is modified as Equation (9). Dtess is the total electricityWith demand of industrial manufacturing from the electrical grid within the interaction of the demand theconsumption electrical grid of within the interaction the ESS inelectricity time interval (t, tof+ industrial 1), and it manufacturing is equal to the from energy all tasks Dt plus theofcharging ESS in time interval (t, t + 1), and it is equal to the energy consumption of all tasks Dt plus the charging energy of the ESS Ec,t , minus the discharging energy of the ESS Ed,t : energy of the ESS Ec,t, minus the discharging energy of the ESS Ed,t: Cost =  Cost ∑ptpt D· tessDtess t

t

ess DDtess = E  DD Ec ,E t+ c,tE− t t  t  d , t d,t

(9)

(9)

(10)

(10)

wherewhere pt and Dt are as in Equation (1). pt and Dt are as in Equation (1). 5.2.2. Constraints on Energy Storage System Modeling 5.2.2. Constraints on Energy Storage System modeling

Figure 8 shows thethe STN model consistsofofstate state nodes including electricity Figure 8 shows STN modelofofthe theESS, ESS, which which consists nodes including electricity from from the power grid, electricity ofofthe fortasks; tasks;and and task nodes including the power grid, electricity theESS, ESS,and and electricity electricity for task nodes including the the charging andand discharging tasks, of which whichare are regarded as STs. In order the charging charging discharging tasks,both both of regarded as STs. In order to mapto themap charging and discharging efficiency γdγrepresent charging and and discharging and discharging efficiencyofofthe theESS ESSininactual actualplants, plants,γcγand charging discharging c and d represent coefficients, respectively. coefficients, respectively.

c

d

Charging

Discharging

Electricity from power grid

Electricity of ESS

Electricity for tasks

Figure8.8.STN STN model model of Figure ofthe theESS. ESS.

Equation (11) shows the balance relationship of the stored energy in the ESS. Eess,t is the energy

Equation balance relationship of the stored energy in theinstant ESS. Etess,t thethe energy storage at (11) time shows instant the t, which is equal to the stored energy at previous time − 1 is plus storage at time instant t, whichwithin is equal the stored at previous instant t − 1 plus energy input from charging timetointerval (t − 1,energy t) multiplied by the time charging efficiency, i.e., the Ec , tinput minus the energy output discharging time interval (t −charging 1, t) divided by the i.e., energy from charging within timefrom interval (t − 1,within t) multiplied by the efficiency, 1  c and discharging efficiency, i.e.,

Ed , t 1   d

.

Energies 2016, 9, 650

11 of 18

Ec,t−1 ·γc and minus the energy output from discharging within time interval (t − 1, t) divided by the discharging efficiency, i.e., Ed,t−1 ÷ γd . Eess,t = Eess,t−1 + Ec,t−1 · γc − Ed,t−1 ÷ γd 0 ≤ Eess,t ≤ Cess

∀t > 0

(11)

∀t > 0

(12)

0 ≤ Ec,t ≤ θc · Mc,t

∀t ≥ 0

(13)

0 ≤ Ed,t ≤ θd · Md,t

∀t ≥ 0

(14)

Mc,t + Md,t ≤ 1

∀t ≥ 0

(15)

where Eess,t is typically constrained within fixed limits (0, Cess ), while Ec,t and Ed,t are typically constrained with fixed limits (0, θc ) and (0, θd ), respectively. The binary variables Mc,t and Md,t indicate whether the ESS is charging or discharging respectively, within time interval (t, t + 1). Mc,t = 1 and Md,t = 1 indicate that the ESS is charging or discharging respectively, during time interval (t, t + 1), otherwise Mc,t = 0 and Md,t = 0. Equation (15) restricts that the ESS cannot be both charging and discharging during the same time interval. 6. Results This section presents the operational scenario and corresponding simulation results of the DR scheme for the stamping process in Figure 3. 6.1. Operational Scenario The ACS has five operating points (Table 1), each of which has related parameters including the cutting frequency of each cutting machine, the parts production amount of the task, and the corresponding electricity demand of the task. Table 1. Operating points of the automatic cutting system (ACS) during each time interval. Operating Point

NSEA (pcs/h)

SEA (pcs/h)

CEA (pcs/h)

Parts Production (pcs/h)

Electricity Demand (kWh)

1 2 3 4 5

1200 1200 1200 1200 1200

0 0 0 0 1800

0 900 1500 2100 2700

1200 2100 2700 3300 5700

5 9 11 13 22

PPS#1 and 2 have only one operating point (Table 2), with related parameters including the production frequency of each punch machine, the parts production amount of the task, and the corresponding electricity demand of the task. Among those parameters, the production frequency indicated the feed consumption rate and the product generation rate of the corresponding punch machine. For example, the production frequency of NSEpi1 was 600 pcs/h, meaning that NSEpi1 consumed 600 pieces of steel plate and produced 600 pieces of spare parts. Table 2. Operating points of parts production system (PPS)#1 and 2 during each time interval, i = 1, 2. Operating Point

NSEpi1 pcs/h

NSEpi2 pcs/h

NSEpi3 pcs/h

Parts Production pcs/h

Electricity Demand kWh

1

600

600

600

600

39

PPS#3, 4 and 5 have five operating points (Table 3), each of which has the related parameters including the production frequency of each punch machine, the parts production amount of the task, and the corresponding electricity demand of the task.

Energies 2016, 9, 650

12 of 18

Table 3. Operating points of PPS#3, 4 and 5 during each time interval, i = 3, 4, 5. Operating Point

CEpi1 (pcs/h)

CEpi2 (pcs/h)

CEpi3 (pcs/h)

Parts Production (pcs/h)

Electricity Demand (kWh)

1 2 3 4 5

0 300 500 700 900

0 300 500 700 900

0 300 500 700 900

0 300 500 700 900

0 30 36 42 48

Energies 2016, 9, 650

12 of 18

Table 3. Operating points of PPS#3, 4 and 5 during each time interval, i = 3, 4, 5.

Table 4 lists the related parameters of steel plate and spare parts. The initial storage and storage Operating CEpi1 CEpi2and 5000, CEpi3 respectively, Parts Production Demand and storage capacity of steel plate were set to 1200 while theElectricity initial storage Point (pcs/h) (pcs/h) (pcs/h) (pcs/h) (kWh) capacity of spare1parts were 0respectively 0 set to 200 0 and 3000. In0industrial processes, 0 maintaining some 2 production 300 is preferable. 300 300 300the parameter SP_Init 30 margin for reliable Therefore, we set to 1200 and the 3 500 500 500 500 36 parameter PP_Init to 200 (rather than700 0). In the 700 whole production process, the consumption of spare 4 700 700 42 parts in every hour was 3000 and the storage capacity was assumed to 5 900 pieces, 900 900 900of the feed (the steel) 48 be infinite. Table 4 lists the related parameters of steel plate and spare parts. The initial storage and storage Parameters for products. capacity of steel plate were setTable to 12004.and 5000, respectively, while the initial storage and storage capacity of spare parts were respectively set to 200 and 3000. In industrial processes, maintaining some margin forParameter reliable production is preferable. Therefore, we set the parameter SP_Init to 1200 Description Value and the parameter PP_Init to 200 (rather than 0). In the whole production process, the consumption SP_Init Initial storage of steel plate 1200 pieces of spare parts in SP_Cap every hour was Storage 3000 pieces, andofthe storage of the feed (the steel) was capacity steel plate capacity5000 pieces assumed to be infinite. PP_Init Initial storage of parts production 200 pieces

PP_Cap

Storage capacity of parts production

3000 pieces

Table 4. Parameters for products. Parameter

Description

Value

Table 5 lists the parameters related toInitial ESS. The initial storage was 0 1200 kWh and the storage capacity SP_Init storage of steel plate pieces was assumed to be 300SP_Cap kWh. The maximum charging/discharging rate was set to 75 kWh, and the Storage capacity of steel plate 5000 pieces PP_Init Initial storage of parts production 200 pieces charging/discharging efficiency coefficient was set to 0.9. The rate of electricity charged/discharged PP_Cap Storage capacity of parts production 3000 pieces during a given time interval was assumed to be continuously controlled between zero and the Table 5 lists the parameters related to ESS. The initial storage was 0 kWh and the storage capacity maximum charge/discharge rate. Electricity prices were generated using price data obtained from was assumed to be 300 kWh. The maximum charging/discharging rate was set to 75 kWh, and the Ameren Illinois Power Company (Belleville, MI, USA) [31] as day-ahead hourly electricity prices charging/discharging efficiency coefficient was set to 0.9. The rate of electricity charged/discharged (Figure 9).during a given time interval was assumed to be continuously controlled between zero and the maximum charge/discharge rate. Electricity prices were generated using price data obtained from Ameren Table 5. Parameters for the Energy Storage System (ESS). prices (Figure 9). Illinois Power Company (Belleville, MI, USA) [31] as day-ahead hourly electricity Table 5. Parameters forDescription the Energy Storage System (ESS). Parameter Value ParameterIess Iess Cess Cess θc θc θd θd γc γc γd γd

ESSDescription initial storage ESSstorage initial storage ESS capacity ESS storage charging capacity rate ESS maximum ESS maximumdischarging charging raterate ESS maximum ESSESS maximum discharging rate charging efficiency ESS charging efficiency ESS discharging efficiency ESS discharging efficiency

0 kWhValue 0 kWh 300 kWh 300 kWh 75 kWh 75 kWh 75 kWh 0.9 75 kWh 0.9 0.9 0.9

Figure 9. Day-ahead hourly electricity prices from Ameren Illinois for 18 June 2014. The prices can be

Figure 9. Day-ahead hourly electricity prices from Ameren Illinois for 18 June 2014. The prices can be classified into three types: low-price (green), high-price (yellow), and peak-price (red). classified into three types: low-price (green), high-price (yellow), and peak-price (red).

Energies 2016, 9, 650 Energies 2016, 2016, 9, 9, 650 650 Energies

13 of 18 13 of of 18 18 13

6.2. Optimization Optimization Results Results 6.2.

Table666shows showsthe the determined operating points for (including STs (including (including ACSPPS#3, and PPS#3, PPS#3, and 5) 5) Table determined operating points for STs ACS and 4 and 5) Table shows the determined operating points for STs ACS and 44 during and during each time interval. interval. For example, example, during time 0, interval 0, was the ACS ACS was scheduled scheduled to work at at each time interval. For example, during time interval the ACS scheduled to work atto operating during each time For during time interval 0, the was work operating point 4, so the TMCS of this task controlled the NSE A to cut steel at a speed of 1200 pieces point 4, sopoint the TMCS of this taskofcontrolled the NSEA the to cut steel a speed pieces perpieces hour, operating 4, so the TMCS this task controlled NSE A toat cut steel atofa 1200 speed of 1200 per SE hour, the SEAAand to be be off, and the steel CEAA to to steelofat at2100 speed ofper 2100 pieces per hour. hour. During this the be off, theoff, CEand to the cut at cut acut speed pieces hour. During this time interval, A tothe A per hour, SE to CE steel aa speed of 2100 pieces per During this time interval, this task produced 3300 pieces of steel plate and required 13 kWh of electricity (Table 1). this task produced 3300 pieces of steel plate and required 13 kWh of electricity (Table 1). time interval, this task produced 3300 pieces of steel plate and required 13 kWh of electricity (Table 1). Table 6. 6. Determined Determined operating operating points points for for schedulable schedulable tasks tasks (STs) (STs) during duringeach eachtime timeinterval. interval. schedulable tasks (STs) Table during each time interval. Time Time Time Interval Interval Interval

ACS ACS ACS PPS#3 PPS#3 PPS#3 PPS#4 PPS#4 PPS#4 PPS#5 PPS#5 PPS#5

0

00

4

44 555 111 55

5

1

11

3

33 555 555 11

1

2

3

4

5

6

7

889

10 11 11 12 12 13 13 14 14 15 15 16 16 17 17 18 18 19 19 20 20 21 21 9910 10 11 12 13 14 15 16 17 18 19 20 21

22 22

55 3 55 5 55 1 5 55

33 4 55 1 11 5 5 55

443 115 551 5 55

334 555 111 5 55

443 555 115 1 55

44 555 555 11

22 3 3 4 4 5 5 6 6 7 7 8

5

55 555 555 11

1

1

11 5 55 5 55 55

5

5 5 5 5

55 55 55 55

5 5 5 5

33 4 55 1 55 5 5 11

44 11 55 55

3

4

3

5

5

5

3 3 4 4 3 3 5 55 1 11 1 11 1 11 5 55 5 55 5 5 5 5 5 5

1 5 5 1

11 55 55 11

1 1 1 1

11 11 11 11

1 5 1 1

11 55 11 11

3

33 5 55 5 55 11

1

4

44 5 55 5 55 11

1

3

33 555 111 55

5

3

33 555 555 11

1

22

23

23 23

4

1

3

33 555 555 11

1

Figure 10 10 shows the total electricity demand forfor STsSTs (without interaction of the the ESS) during each 10shows showsthe thetotal totalelectricity electricity demand (without interaction of ESS) the ESS) during Figure demand for STs (without interaction of during each time interval. interval. According to the theto dynamic changes in electricity electricity prices,prices, STs (including (including ACS and and PPS#3, each time interval. According the dynamic changes in electricity STs (including ACS and time According to dynamic changes in prices, STs ACS PPS#3, 4, and 5) were scheduled to operate at different operating points, as shown in Table 6. When the price PPS#3, and 5) were scheduled to at operate at different operating inWhen Table the 6. When 4, and 5)4,were scheduled to operate different operating points, aspoints, shownasinshown Table 6. price wasprice low, STs STs were were scheduled to operate operateto atoperate high-energy-consumption operating point, point, increasing the was low, scheduled STs were scheduled at a high-energy-consumption operating point, was low, to at aa high-energy-consumption operating increasing electricity demand. Consequently, more steel plate and spare parts were produced and stored during increasingdemand. electricity demand. Consequently, more steel plate andwere spare parts were and electricity Consequently, more steel plate and spare parts produced andproduced stored during these times, times, asthese shown in Figures Figures 11inand and 12, respectively. respectively. When the the price price was high, STs were were stored duringas times, as shown Figures 11 and 12, respectively. Whenwas the high, price was high, these shown in 11 12, When STs scheduled to operate operate at aa low-energy-consumption low-energy-consumption operating point, decreasing the electricity STs were scheduled to operate at a low-energy-consumption operating point, decreasingthe the electricity electricity scheduled to at operating point, decreasing demand. During these periods, STs produced fewer products, and previously stored products were During these these periods, periods, STs STs produced fewer products, and previously stored products were demand. During supplied to subsequent tasks, as shown in Figures 11 and 12, which eventually shifted the load from supplied to subsequent tasks, as shown in Figures 11 and 12, which eventually shifted the load from peak to to off-peak off-peak periods. periods. peak

Figure 10. Total electricity demand of STs (including ACS and andPPS#3, PPS#3,444and and5)5) 5)during duringeach each time interval. Figure 10. Totalelectricity electricitydemand demandof ofSTs STs (including (including ACS ACS time interval. Figure 10. Total and PPS#3, and during each time interval.

Figure 11. 11. Storage Storage amount amount of of steel steel plate plate during during each each time time interval. interval. Figure

Energies 2016, 9, 650 Energies 2016, 9, 650 Energies 2016, 9, 650

14 of 18 14 of 18 14 of 18

Figure during each eachtime timeinterval. interval. Figure12. 12.Storage Storageamount amount of of spare spare parts parts during

Figure 1313shows shows the electricity ESS as as aa function function of ofthe theelectricity electricityprice priceduring during Figure13 showsthe theelectricity electricitydemand demand of of the the ESS Figure demand of the function of the electricity price during each time interval. For electricity demand, positive values indicate charging and negative values each time interval.For Forelectricity electricity demand, demand, positive positive values each time interval. values indicate indicatecharging chargingand andnegative negativevalues values indicate discharging. During low-price periods, the ESS charged from the power grid to store energy, indicatedischarging. discharging. During periods, the ESS from the power to store indicate Duringlow-price low-price periods, the charged ESS charged from the grid power gridenergy, to store while during high-price periods, the ESS discharged provide electrical energy for industrial while during high-price periods, the ESS discharged to provide electrical energy for industrial energy, while during high-price periods, the ESS discharged to provide electrical energy for industrial manufacturing. time interval, the maximum charging/dischargingamount amount the ESS manufacturing.During Duringeach eachtime timeinterval, interval, the the maximum maximum charging/discharging charging/discharging manufacturing. During each amountofof ofthe theESS ESS was 75 kWh, limited by the maximum charging/discharging rate (Table 5). During time interval was kWh, limitedbybythe themaximum maximumcharging/discharging charging/discharging rate 6, 6, was 7575 kWh, limited rate(Table (Table5). 5).During Duringtime timeinterval interval 6, the stored energy reached to 300 kWh, which was storage capacity of the ESS. And the ESS the stored energy reached to 300 kWh, which was the storage capacity of the ESS. And the ESS the stored energy reached to 300 kWh, which was the storage capacity of the ESS. And the ESS completely dischargedall allof energyduring duringtime time interval interval 18. 18. completely discharged all ofofenergy energy during time interval 18. The The total discharged electricitywas wasless less completely discharged The total totaldischarged dischargedelectricity electricity was less than the total charged electricity because of the energy loss during charging and discharging process. than the than the total total charged charged electricity electricity because because of of the the energy energy loss during during charging charging and and discharging discharging process. process.

Figure Electricitydemand demand of of the the ESS. ESS. Positive Positive values Figure 13.13.Electricity values indicate indicate charging chargingand andnegative negativevalues values Figure 13.discharging. Electricity demand of the ESS. Positive values indicate charging and negative values indicate indicate discharging. indicate discharging.

Figure 14 shows the comparison of total electricity demands involving three different test cases. Figure 14 shows shows the the comparison comparison of of total total electricity electricity demands involving involving three three different different test cases. cases. Figure Case a was14based on a fixed-price, which was constant demands over all time intervals and set to be test equal to Case a was on fixed-price, which was over allall time intervals and set set to be to the Case was based based onaadynamic fixed-price, which wasconstant constant over time intervals to equal be bequal the aaverage of the price, i.e., no DR function was applied to the and FEMS. Case onlyto average of the dynamic price, i.e., no DR function was applied to the FEMS. Case b only considered theconsidered average of thefunction dynamic price, i.e., no DR function was prices. applied to cthe FEMS. the Case only a DR based on day-ahead hourly electricity Case integrated ESSbinto a DR function based on day-ahead hourly electricity prices. Case c integrated the ESS into the DR considered a DR function based day-ahead electricity Case c integrated ESS into the DR function. Compared toon Case a, Case hourly b consumed moreprices. electrical energy duringthe low-price function. Compared to Case a, Case Case energy b more electrical energy during low-price periods theperiods DR function. Compared to a, consumed Caseduring b consumed more electrical energy low-price and consumed less electrical high-price periods. In Case c, the during energy demand and consumed less electrical energy during high-price periods. In Case c, the energy demand gap gap between maximum and minimum became larger. Among periods. the threeIn cases, consumed the periods and consumed less electrical energy during high-price CaseCase c, thec energy demand between maximum and minimum became larger. Among the three cases, Case c consumed the most most electrical energy during low-price periods and Among consumed least electrical energy during gap between maximum and minimum became larger. thethe three cases, Case c consumed the electrical energy during low-price periods and consumed the least electrical energy during high-price high-price periods, since the ESS stored periods energy when prices werethe lowleast and electrical supplied energy when most electrical energy during low-price and consumed during periods, since the ESS stored energy when prices were prices low and supplied energy whenenergy prices were prices were high. In Case b, the total costenergy was reduced by shifting demand from high-price periods high-price periods, since the ESS stored when were low and supplied when high. In Casehigh. b, the was reduced bywas shifting demand from high-price periods to demand low-price to low-price periods. In cost Case c, total the total reduced further not only by shifting prices were Intotal Case b, the costcost was reduced by even shifting demand from high-price periods periods. In Case c, the total cost was reduced even further not only by shifting demand but also also byperiods. managing 15 shows a cost comparison the three testby to but low-price In storage Case c, through the totalthe costESS. wasFigure reduced even further not only by for shifting demand managing storage through the ESS. Figure 15 shows a cost comparison for the three test cases in 1 cases 1 day. The total energythrough cost for Case c wasFigure 11.55%15 lower thanathat Case a. but alsoinby managing storage the ESS. shows costofcomparison for the three day. test The forenergy Case ccost wasfor 11.55% than that of Case a. of Case a. casestotal in 1 energy day. Thecost total Case clower was 11.55% lower than that

Energies 2016, 9, 650 Energies 2016, 2016, 9, 9, 650 650 Energies

15 of 18 15 of of 18 18 15

Figure Total electricity three test each Figure 14. 14. Total electricity demand demand (within (within the the interaction interaction of of the the ESS) ESS) for for three test cases cases during during each time time interval. interval: (a) without DR; (b) with DR only; and (c) with DR and ESS.

Figure 15. Cost test cases. cases. Figure 15. Cost comparison comparison for for three three test

7. 7. Conclusions Conclusionsand andFuture FutureWork Work As industrialmanufacturing manufacturing becomes more intelligent, many have factories already As industrial becomes more intelligent, many factories alreadyhave implemented implemented automatic production systems. Thus, process automation systems have infrastructure automatic production systems. Thus, process automation systems have infrastructure for adopting for adopting DR-based energy management systems. Against this background, introducing DR into DR-based energy management systems. Against this background, introducing DR into industrial industrial manufacturing has an influence effective influence on reducing consumption costs. The contributions manufacturing has an effective on reducing consumption costs. The contributions of this of this paper are mainly embodied in two aspects, the implementation of a STN representation to a paper are mainly embodied in two aspects, the implementation of a STN representation to a discrete discrete manufacturing system in terms of process operations and energy consumptions, and the manufacturing system in terms of process operations and energy consumptions, and the application of application a DR energy management scheme using affordable the computationally affordable MILP a DR energyof management scheme using the computationally MILP optimization algorithm optimization algorithm to discrete manufacturing processes in a SG. The aim of this scheme is to balance to discrete manufacturing processes in a SG. The aim of this scheme is to balance energy requirements energy requirements shifting electricity demands fromperiods peak to and off-peak periods reducingcost. the by shifting electricitybydemands from peak to off-peak reducing theand associated associated cost. To evaluate the performance of the DR scheme, a typical discrete manufacturing To evaluate the performance of the DR scheme, a typical discrete manufacturing system—automobile system—automobile manufacturing—was chosen as an example for simulation. manufacturing—was chosen as an example for simulation. The atat a The simulation simulation results resultsshowed showedthat thatwhen whenthe theprice pricewas waslow, low,STs STswere werescheduled scheduledtotooperate operate high-energy-consumption operating point, increasing electricity demand. Consequently, a high-energy-consumption operating point, increasing electricity demand. Consequently, intermediate intermediate products were produced and stored in greater quantities during these periods. When products were produced and stored in greater quantities during these periods. When the price was the price STs were scheduled operate at a low-energy-consumption point, high, STs was werehigh, scheduled to operate at a to low-energy-consumption operating point,operating decreasing the decreasing the electricity demand. Less of the corresponding intermediate products were produced electricity demand. Less of the corresponding intermediate products were produced and previously and previously products utilizedtasks in subsequent tasks duringInthese periods. In contrast, stored productsstored were utilized in were subsequent during these periods. contrast, the ESS charged the ESS charged from the power grid to store energy during low-price periods, and discharged to from the power grid to store energy during low-price periods, and discharged to provide energy for provide energy for industrial manufacturing during high-price periods. Comparing three different industrial manufacturing during high-price periods. Comparing three different test cases: Case a test cases:DR, Case a without with DR conly, withour DRresults and ESS, our results without Case b withDR, DR Case only, band Case withand DRCase and cESS, indicated thatindicated the total that the total costs can be reduced by shifting demand from peak to off-peak demand periods, and costs can be reduced by shifting demand from peak to off-peak demand periods, and reduced further reduced further by managing electricity storage through the ESS. Through managing industrial by managing electricity storage through the ESS. Through managing industrial facilities to work at facilities to work at these optimal operating points, the DR scheme provides a significant overall cost reduction without degrading production processes.

Energies 2016, 9, 650

16 of 18

these optimal operating points, the DR scheme provides a significant overall cost reduction without degrading production processes. In this study, we adopted a simulation model to validate the performance of the proposed DR algorithm. Because the validity of the DR algorithm was verified by the simulation results, one of our future works is to implement the DR algorithm to the actual discrete manufacturing system. Meanwhile, another kind of DERs, the energy generation systems (EGSs) such as solar, wind and waste heat power plants, can be integrated into the DR scheme to further reduce energy costs. Acknowledgments: This work was supported in part by the Human Resources Program in Energy Technology of the Korea Institute of Energy Technology Evaluation and Planning (KETEP); the Gyeonggi Regional Research Center (GRRC) program of Gyeonggi Province under Grant GRRC Hanyang 2015-B02 (Development of Industrial Communication-IoT Gateway). Author Contributions: Seung-Ho Hong conceived and designed the experiments. Zhe Luo performed the experiments and analyzed the data; Jong-Beom Kim contributed reagents/materials/analysis tools. All authors wrote and approved the manuscript. Conflicts of Interest: The authors declare no conflict of interest.

Nomenclature Variables crk,n,s Cess Cs,k,t dk,n dk,t Dt Dtess Eess,t Ec,t Ed,t i k Mc,t Md,t Mk,n,t n NST prk,n,s pt Ps,k,t s Ss max Ss min SAs,t ST t Tc,s Tp,s θc θd γc γd

Consumption rate of state s when schedulable task k operates at operating point n Energy capacity of the ESS Amount of state s consumed by task k within time interval (t, t + 1) Electricity demand of schedulable task k that operates at operating point n in one time interval Electricity demand of task k within time interval (t, t + 1) Electricity demand of whole manufacturing (excluding ESSs) within time interval (t, t + 1) Electricity demand of whole manufacturing (including ESSs) within time interval (t, t + 1) Electricity storage of the ESS at time boundary t Amount of electricity charged by the ESS within time interval (t, t + 1) Amount of electricity discharged by the ESS within time interval (t, t + 1) Index of system number Index of tasks If ESS charges electricity within time interval (t, t + 1), Mc,t = 1; otherwise Mc,t = 0 If ESS discharges electricity within time interval (t, t + 1), Md,t = 1; otherwise Md,t = 0 If schedulable task k operates at operating point n within time interval (t, t + 1), Mk,n,t = 1; Otherwise Mk,n,t = 0 Index of operating point supported by a schedulable task The set of non-schedulable task Production rate of state s when schedulable task k operates at operating point n Electricity price purchased from the utility within time interval (t, t + 1) Amount of state s produced by task k within time interval (t, t + 1) Index of state Maximum storage of state s Minimum storage of state s Storage amount of state s at time boundary (t, t + 1) The set of schedulable task Index of time boundary The set of tasks that consume state s The set of tasks that produce state s Maximum charge rate of the ESS Maximum discharge rate of the ESS Conversion coefficient of charge Conversion coefficient of discharge

Energies 2016, 9, 650

17 of 18

Abbreviations ACS AMPS CE DER DR DSM EMS ESS FEMS IAN ICT MILP NSE PPS SE SG STN TOU TEMA TMCS WAN

Automatic cutting system Automobile manufacturing process system Controllable equipment Distributed energy resource Demand response Demand side management Energy management system Energy storage system Factory energy management system Industrial area network Information and communication technology Mixed integer linear programming Non-shiftable equipment Parts production system Shiftable equipment Smart grid State task network Time of use Task energy management agent Task monitoring and control system Wide area network

References 1.

2. 3. 4.

5.

6. 7. 8.

9.

10. 11. 12.

Ettlinger, M.; Gordon, K. The Importance and Promise of American Manufacturing. Available online: http://cdn.theatlantic.com/static/mt/assets/business/CAP_Importance_and_Promise_of_American_ Manufacturing%20.pdf (accessed on 6 August 2016). Key World Energy Statistics 2015. Available online: https://www.iea.org/publications/freepublications/ publication/KeyWorld_Statistics_2015.pdf (accessed on 6 August 2016). Samad, T.; Kiliccote, S. Smart grid technologies and applications for the industrial sector. Comput. Chem. Eng. 2012, 47, 76–84. [CrossRef] Zhang, Q.; Grossmann, I.E. Planning and scheduling for industrial demand side management: Advances and challenges. In Alternative Energy Sources and Technologies; Martín, M., Ed.; Springer International Publishing AG: Basel, Switzerland, 2016; pp. 383–414. Endo, M.; Nakajima, H.; Hata, Y. Simplified Factory Energy Management System Based on Operational Condition Estimation by Sensor Data. In Proceedings of the 2012 IEEE International Conference on Automation Science and Engineering (CASE), Seoul, Korea, 20–24 August 2012; pp. 14–19. Yan, Y.; Qian, Y.; Sharif, H.; Tipper, D. A Survey on Smart Grid Communication Infrastructures: Motivations, Requirements and Challenges. IEEE Commun. Surv. Tutor. 2013, 15, 5–20. [CrossRef] Shariatzadeh, F.; Mandal, P.; Srivastava, A.K. Demand response for sustainable energy systems: A review, application and implementation strategy. Renew. Sustain. Energy Rev. 2015, 45, 343–350. [CrossRef] Koliou, E.; Eid, C.; Chaves-Ávila, J.P.; Hakvoort, R.A. Demand response in liberalized electricity markets: Analysis of aggregated load participation in the German balancing mechanism. Energy 2014, 71, 245–254. [CrossRef] Rahimi, F.; Ipakchi, A. Overview of Demand Response under the Smart Grid and Market Paradigms. In Proceedings of the 2010 Innovative Smart Grid Technologies (ISGT), Gaithersburg, MD, USA, 19–21 January 2010; pp. 1–7. Neves, D.; Silva, C.A. Optimal electricity dispatch on isolated mini-grids using a demand response strategy for thermal storage backup with genetic algorithms. Energy 2015, 82, 436–445. [CrossRef] Li, S.; Zhang, D.; Roget, A.B.; O’Neill, Z. Integrating home energy simulation and dynamic electricity price for demand response study. IEEE Trans. Smart Grid 2014, 5, 779–788. [CrossRef] Giacone, E.; Mancò, S. Energy efficiency measurement in industrial processes. Energy 2012, 38, 331–345. [CrossRef]

Energies 2016, 9, 650

13. 14.

15.

16.

17.

18.

19. 20. 21. 22. 23. 24. 25. 26. 27.

28. 29. 30. 31.

18 of 18

Wang, Y.; Zeng, P.; Yu, H.; Zhang, Y.; Wang, X. Energy tree dynamics of smart grid based on industrial internet of things. Int. J. Distrib. Sens. Netw. 2013, 2013. [CrossRef] Ikeyama, T.; Watanabe, H.; Isobe, S.; Takahashi, H. An Approach to Optimize Energy Use in Food Plants. In Proceedings of the 2011 SICE Annual Conference (SICE), Tokyo, Japan, 13–18 September 2011; pp. 1574–1579. Georgiadis, M.C.; Levis, A.A.; Tsiakis, P.; Sanidiotis, I.; Pantelides, C.C.; Papageorgiou, L.G. Optimisation-based scheduling: A discrete manufacturing case study. Comput. Ind. Eng. 2005, 49, 118–145. [CrossRef] Chong, C.S.; Sivakumar, A.I.; Gay, R. Simulation-Based Scheduling for dynamic Discrete Manufacturing. In Proceedings of the 2003 Winter Simulation Conference, New Orleans, LA, USA, 7–10 December 2003; pp. 1465–1473. Cannata, A.; Karnouskos, S.; Taisch, M. Energy Efficiency Driven Process Analysis and Optimization in Discrete Manufacturing. In Proceedings of the 35th Annual Conference of IEEE Industrial Electronics (IECON ’09), Porto, Portugal, 3–5 November 2009; pp. 4449–4454. Sohail, M.; Florea, A.; Lastra, J.L.M. A Case Study of Share of ICT Infrastructure in Energy Consumption of Discrete Manufacturing Facility. In Proceedings of the 2014 IEEE 15th Workshop on Control and Modeling for Power Electronics (COMPEL), Santander, Spain, 22–25 June 2014; pp. 1–6. Mohagheghi, S.; Raji, N. Managing industrial energy intelligently: Demand response scheme. IEEE Ind. Appl. Mag. 2014, 20, 53–62. [CrossRef] Karwan, M.H.; Keblis, M.F. Operations planning with real time pricing of a primary input. Comput. Oper. Res. 2007, 34, 848–867. [CrossRef] Mitra, S.; Grossmann, I.E.; Pinto, J.M.; Arora, N. Optimal production planning under time-sensitive electricity prices for continuous power-intensive processes. Comput. Chem. Eng. 2012, 38, 171–184. [CrossRef] Shrouf, F.; Ordieres-Meré, J.; García-Sánchez, A.; Ortega-Mier, M. Optimizing the production scheduling of a single machine to minimize total energy consumption costs. J. Clean. Prod. 2014, 67, 197–207. [CrossRef] Pulkkinen, P.; Ritala, R. TMP production scheduling under uncertainty: Methodology and case studies. Chem. Eng. Process. Process Intensif. 2008, 47, 1492–1503. [CrossRef] Castro, P.M.; Harjunkoski, I.; Grossmann, I.E. New continuous-time scheduling formulation for continuous plants under variable electricity cost. Ind. Eng. Chem. Res. 2009, 48, 6701–6714. [CrossRef] Castro, P.M.; Harjunkoski, I.; Grossmann, I.E. Optimal scheduling of continuous plants with energy constraints. Comput. Chem. Eng. 2011, 35, 372–387. [CrossRef] Kondili, E.; Pantelides, C.C.; Sargent, R.W.H. A general algorithm for short-term scheduling of batch operations—I. MILP formulation. Comput. Chem. Eng. 1993, 17, 211–227. [CrossRef] Ding, Y.M.; Hong, S.H. A Model of Demand Response Energy Management System in Industrial Facilities. In Proceedings of the 2013 IEEE International Conference on Smart Grid Communications (SmartGridComm), Vancouver, BC, Canada, 21–24 October 2013; pp. 241–246. Ding, Y.M.; Hong, S.H.; Li, X.H. A demand response energy management scheme for industrial facilities in smart grid. IEEE Trans. Ind. Inform. 2014, 10, 2257–2269. [CrossRef] Futurelook’s Wireless Communication Module in Automotive Assembly Line. Available online: http://www.plcremote.com/P/p74.html (accessed on 6 August 2016). Chen, Y.; Li, G.; Ping, C.; Lu, L.; Li, W. Information model of processing for discrete manufacturing based on granular information. WIT Trans. Eng. Sci. 2016, 113, 106–113. US Ameren Illinois Utility Company Real-Time Pricing. Available online: https://www.powersmartpricing. org/prices/?date=20140618 (accessed on 6 August 2016). © 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).