CCECE 2014 1569906291
Design and Implementation of a Rule-based Learning Algorithm Using Zigbee Wireless Sensors for Energy Management Azim Keshtkar, Student Member, IEEE, and Siamak Arzanpour School of Mechatronic Systems Engineering, Simon Fraser University Surrey, BC V3T OA3, Canada
[email protected] and
[email protected] Abstract-The
capabilities
of
wireless
sensors
temperature temperature.
networks
(WSNs) to measure different variables, could significantly improve the limitations of the existing energy management
capabilities of wireless sensors in reducing the electricity consumption without sacrificing thermal comfort that would help utilities in peak load curtailments. The method is applied to existing programmable thermostats (PTs) to add more intelligence to this device for better energy management in residential buildings. The simulation results demonstrate that the proposed rule-based wireless thermostat performs better than the PTs in various aspects, i.e., learning, electric energy conservation, and occupant comfort that could help utilities in peak load curtailment. Moreover, our method is implemented on a typical residential Air Conditioner (AC) by using of Xbee sensor
and
Arduino
Microcontroller.
Conducted
results show that the combination of WSNs capabilities and the rule-based method reduce the energy consumption by 33.5% compared to the similar existing AC system.
than
the
specified
The capabilities of sensor nodes to observe or monitor different variables of interest help to overcome the limitations of the existing energy management systems. Therefore, wireless sensor networks (WSN) afford a natural and potentially cost-effective mechanism for the monitoring, load control and energy management systems [4,10,15]. Load management can be divided into direct and indirect method [13]. Direct load management (control) is based on the technological measures and defines the load demand by directly switching different equipment on or off over the time. Indirect load control is based on the regulations or economic measures [13]. Different tariffs and pricing mechanisms such as fixed price, Time of Use (TaU), etc. are introduced in order to encourage customers to reduce load demand in peak periods [12]. The utility provides the electricity cost at different times of the day based on the TaU rates (e.g. On Peak, Mid Peak and Off Peak). In many cases it is important to combine indirect load control (pricing) and direct load control (demand shifting or shedding on HVAC systems, processes, etc.) in load management programs [8,9,13]. The idea of using FM radio signals for controlling of a residential air conditioner was proposed in [9]. They reported a median household load reduction of 0.77 kW over the five-hour critical peak period. In [10], the authors proposed a method based on rule-based
Keywords-Wireless Sensor Networks; Rule-Based techniques; Wireless Thermostat; HVAC Systems; Electricity Consumption.
I.
greater/smaller
Nowadays, PTs can be programmed to change temperature Set Points on a schedule and half of households actually use them in this way [1]. According to Energy Information Administration in USA in 2005, during the heating season, 60% of households with PTs used these devices to reduce temperature during the night time but only 45% reduced the temperature during the day time. During the cooling season, 55% of households with PTs, set them to increase temperature at night as well as during the day time [2]. Using a PT itself does not guarantee reduction in energy consumption and most of times the reduction depends on how the device is programmed and controlled by households. Hence, PTs have the potential to benefit both consumers and suppliers for the demand-side electricity management to automatically change inside temperature as well as fan operation modes.
systems. In this paper, we introduce a combination of rule based techniques and wireless sensors to demonstrate the
wireless
is
INTRODUCTION
Growing residential energy consumption and limited electrical power resources is now a global challenge. Governments and utilities have recently tried to enforce new policies to re-regulate electricity prices and load control strategies to handle the crisis. These strategies would result in decreasing wide-spread regional electrical power outages, automatic load reduction, and risk management [1,2,3,13]. Residential Heating, Ventilation, and Air Conditioning (HVAC) have the highest contribution to the world's energy consumption [2,11]. Thus, rising energy prices and the transition to dynamic electricity pricing from flat-rate tariffs to dynamic pricing will impact the consumers whose energy bills are highly related to HVAC systems. Programmable Thermostats (PTs) are used widely for automatic control of temperature, humidity and HVAC systems [1,2]. Generally, these devices use a local temperature in the level-crossing control logic with the specified temperature Set Point. The thermostat requests cooling/heating if the measured
1
fuzzy logic and implemented in Zigbee nodes, with the aim to reduce the on/off frequency of an air conditioner system. They used the temperature, humidity, fan speed, and engine speed as input variables. Their experiments showed promising results compared to a traditional control system which were based on discrete temperature values. The difficulty with this method is that it does not consider the thermal comfort. In [11], they have examined how much saving could have had if the users see their bill values on a deployed LCD screen in the house. Authors presented the application of WSN and adaptive learning techniques, in order to introduce an adaptable systemic solution in [14]. The advantage of their method is that, they used several subsystems that are able to share the knowledge between the subsystems. Their approach is able to learn and adapt by applying a rules-based system and adaptive learning methods. However, their method must be investigated further in terms of user comfort and implement practical measures. However, it has to be implemented on real HVAC systems in order to be more effective approach on energy consumption and occupant comfort.
side is affected by thermal resIstIvIty of materials K, temperature difference L1T, wall thickness L and area A, and is represented by equation (1) [5]: q =
L
=
(tl - t2)
( 1)
L
kA
(2)
Where L1T is the difference between inside and outside temperature, Ajlow the heated/cooled airflow, and cis Specific heat capacity of the air (J/kg K). For more details regarding the principles of house thermodynamic equations refer to [5,6]. III.
T HE PROPOSED RULE-BASED ALGORITHM
A rule-based algorithm represents the knowledge of the outside world and specifies how to react to input signals. For this purpose, the system makes decisions on the basis of a number of rules. In fact, the system constantly evaluates the inputs available (information from WSN) and makes decisions about the outputs of the system according to the rules. The rule-based consists of a number of simple/multiple if-then statements [7]. If AI, A2 , An represent the conditions of an environment that are sensed by sensors; and Bj, B2 Bn represent the actions to be taken if a particular condition(s) are true, then the rules can simply be expressed as shown in following: • • •
. . •
The rest of paper is organized as following. Section II describes the house thermodynamic model and WSN. We discuss the proposed rule-based algorithm in Section III. Section IV explains the simulation and results. In Section V we explain the implementation of Zigbee-based wireless thermostat. Section VI concludes the paper. HOUSE THERMODYNAMIC EQUATIONS AND
(tl - t2)A
And the heated/cooled air supply into the house is depicted as:
In this experimental research, we have developed the proposed method in [14] in terms of occupant comfort and implemented it on an Air Conditioner (AC) by utilizing a new rule-based algorithm and zigbee-based wireless sensors for load shifting/shedding to meet the concept of load management by load reduction (reducing temperature) without losing user thermal comfort during peak demand and/or peak price periods.
II.
k
IF (AI and A2 or A3 or A4 ... and A,J then (B,J Fig. 2 shows the concept of rule-based system and WSN. The system consists of a few subsystems that are sharing knowledge and information to achieve a better outcome. The WSN and rule-based technique enable the system to interact with sensor data and also using the existing KB with the new knowledge that being introduced.
WSNs
Fig. 1 shows the conceptual design of a house heating cooling system model that is included the outside temperature, converter, HVAC unit and a Thermostat [6]. We use WSN to obtain distributed sensor information and make decision based on our rule-based approach in terms of energy saving and thermal comfort.
A. Knowledge-Based Subsystem The Knowledge-Based (KB) contains information about the heating/cooling. For example, operation modes, different temperature ranges, house parameters, and characteristics of the house, can be the features of KB. In our case, the environment parameters are constantly observed via sensors which are able to detect the inside/outside temperature changes, airflow rate and the
As shown in Fig. 1 the thermostat consists of a controller and a temperature sensor that senses the inside temperature. The deployed sensor in thermostat receives the input from the Heating/Cooling system and transducer, and then computes the room temperature accordingly. The heating/cooling is generated by the heater/cooler and also the heat/cool loss generated based on the house thermodynamics. The Controller reacts when temperature values received from the inside temperature sensor are different than set point temperatures and provides the heated/cooled air with the constant air flow. Generally, heat flow through a house depends on a couple of factors, such as the difference in inside and outside temperature, conductivity of building materials, thickness of materials, etc. Heat transfer process from the warmer side to the colder
Heat Losses
Tout
Tin
·,·:"·::", LF ,
;....
,
..........,, ..............
t__________ �p_"-t.!..O}_�iJI.'2�!____________ �
Fig. 1: Conceptual design of house heating/cooling system
2
Fig. 2: Conceptual design of rule-based system
activity of occupants. The system collects input data from WSNs, and detects when the user's schedules change. The decision is wirelessly sent to actuators to control the airflow and provide comfort temperature. The airflow rate is computed in KB subsystem based on the algorithm in Fig. 3 for different inside and outside temperatures.
Fig. 3: Flowchart of Fan speed adjustment for heating operation
consumer such that could conserve energy when the home is unoccupied or providing comfort when the user is home. To implement the method we assume three different offsets as the weights associated with heating/cooling temperature for each particular day. The rule-based system is used in decision making process, in order to learn and predict new habits. Thus, there are three different weights which can be assigned to any daily temperature schedule. Learning structure elements of each daily schedule such as a Set Point are: {Heat/Cool Set Point, Start Time, Stop Time}. Therefore, for any change in learning structure elements three weights are defined to learn and predict occupant habit. The intervals of weights are WL [0 1] °c for low change, WM [1 3] °c for Medium change and WH [3 5] °c for high change in scheduled Set Points. We have the same weights for Start time and Stop time when any change occurs at a time different from defined schedules in Table II (WL 1 hour, WM 3 hours and WH 5 hours). In the beginning, initial weights of each element are zero that means no change observed/happened yet. Once it changes the system records the amount and the time of that occurred event and assigns 1 to it. The algorithm works as following:
In Fig. 3 we assume the system has three speeds (i.e. Low, Medium, and High). We also assume T states the difference between inside and outside temperature, and P indicates the total consumption of HVAC system for each month. If PI is normal threshold consumption of HVAC in residential buildings (38% of total house electricity consumption [1, 2]) and TI, T2, T3 are the boundaries for temperature, therefore, the KB after processing the information, based on the algorithm depicted in Fig. 3, returns the recommended fan speed to rule-base system. In Fig. 3, it is also taken into consideration the user presence to provide user thermal comfort.
=
=
B. System Model and Rule-Based Algorithm The rule-based algorithm is used to provide a smart device that does not require to be constantly programmed by consumers for participating in demand-side load management and adjusting thermal comfort. Therefore, to have the reasonable savings, and manage the peak load demand; WSN is able to communicate with the main unit and process input variables. Since, normally, wireless sensors/actuators are limited in memory and power [4, 15]. Thus, rule-based techniques can be used to enhance the performance of the system by activating the sensor nodes to act as smart system and predict the future peak load events after receiving the potential functions and time of a day interval from smart meters which are deployed in houses. The aim of rule-based system is to optimize the user comfort with respect to energy consumption by learning occupancy habits and providing thermal comfort. •
=
=
=
=
Suppose Olnew, 02new, and 03new are three new consecutive occurrences which are sensed by sensors for the set points once the user changes the Set Point at different times. We use information in Table II to train the system. The following rules are executed to predict the new habit for the fourth day. If the difference between new change and old change (defmed schedule) for each occurrence has the same weight or different weights, therefore, the average of new changes is the predicted habit for forth event (see equation 3). The average of the new changes in temperature set point and the time that it started and ended is returned as a new schedule. If two of changes were in the same interval, we would disregard another one and the new habit is computed by equations (4), (5) (6), or (7). Therefore, the new pattern for changing in temperature set point and the time that those happened by considering the user comfort (±PMV) is computed as following. There are totally 18 rules and some of them come in below. IF IOlnew - 0lold I � WL and IOznew - 0zold I � WL and I03new - 0301d � WL OR
Predicted Mean Vote (PMV) index for thermal comfort
In order to provide occupant thermal comfort, instead of only using air temperature as a thermal comfort index, we can consider the more global PMV index selected by international standard ISO 7730 [5]. In this way, we are able to conserve more energy but maintaining user thermal comfort according to PMV index for thermal comfort. In [14], they just used air temperature as an index in their method for predicting user habits. In our algorithm we take into account PMV index after predicting new habit of
3
WL