SPECIAL SECTION ON SMART GRIDS: A HUB OF INTERDISCIPLINARY RESEARCH
Received October 9, 2015, accepted October 22, 2015, date of publication December 9, 2015, date of current version December 18, 2015. Digital Object Identifier 10.1109/ACCESS.2015.2507372
Understanding Customer Behavior in Multi-Tier Demand Response Management Program AQSA NAEEM1 , ALI SHABBIR1 , NAVEED UL HASSAN1 , (Senior Member, IEEE), CHAU YUEN2 , (Senior Member, IEEE), AYAZ AHMAD3 , AND WAYES TUSHAR2 , (Member, IEEE) 1 Electrical
2 Singapore
Engineering Department, Lahore University of Management Sciences, Lahore 54792, Pakistan University of Technology and Design, Singapore 487372 of Electrical Engineering, COMSATS Institute of Information Technology, Wah Cantonment 47040, Pakistan
3 Department
Corresponding author: N. Ul Hassan (
[email protected]) This work is supported in part by Lahore University of Management Sciences, Faculty Initiative Fund (FIF) grant and Singapore University of Technology and Design (SUTD) through the Energy Innovation Research Program (EIRP) Singapore NRF2012EWT-EIRP002-045 and IDC project IDG31500106.
ABSTRACT In this paper, we investigate the factors that influence the customer’s decision in subscribing to a particular demand response management (DRM) scheme. Based on these factors, we suggest a classification of customer types that include non-green comfort seeking behavior (NCSB) and green incentive seeking behavior (GISB). We use multi-tier DRM plans that clearly specify the incentive and inconvenience for NCSB and GISB customers. The grid operator can obtain maximum profit margin (after paying out the incentives to participating users) from the DRM only if a specific number of customers participate in the NCSB and GISB plans. Any deviation from an ideal subscription pattern is undesirable from the grid’s perspective. We also develop a mathematical framework that is based on logistic regression and considers the quantifiable as well as unquantifiable attributes of customer behavior. This model captures the probabilistic nature of customer preferences for different DRM plans. Simulation results reveal that the actual subscription of customers in NCSB and GISB plans significantly deviates from the ideal values. From these, we determine a compromise solution that lies between the ideal and the actual solutions. We also identify that along with economic factors, social factors, such as peer pressure and prompting green or caring behavior, could also be used as potential tools by the grid operator to influence the customer preference. This paper can also be used by the grid operator to design appropriate multi-tier DRM plans. INDEX TERMS Customer behavior, demand response management, incentive, inconvenience, logit model.
I. INTRODUCTION
Smart grid has become a significant component of the energy infrastructure, as it is conceived to facilitate power generation, distribution and management in an efficient way. It offers improved grid performance, reliability and security by using advanced computing facilities and allowing bidirectional flow of information and energy between the utility provider and customers [1]. The monitoring and regulation of the consumers’ intelligent devices enables smart grid to save energy, minimize cost and improve efficiency of the system by delivering power in a controlled way to the end-customers [2]. With a growing transition of conventional electric grids into smart grids, Demand Response Management (DRM) can be viewed as an essential tool to regulate the electricity VOLUME 3, 2015
demand and supply profiles [3]. DRM includes all activities that potentially enable the users to change their electricity consumption pattern either through electricity tariff variations or direct incentive payments by the grid operator [4], [5]. The true benefits of DRM programs for fine grained energy management requires significant and active customer participation [6], [7]. Empirical studies, however, show that customers, particularly in the residential sector, are hesitant to subscribe to DRM programs [8]. For example, the authors in [9] identify that the lack of clarity and awareness about the specifics of DRM programs are a major hindrance towards customer engagement. This observation is also supported by [10], which further elaborates that users tend to limit their power consumption when they are properly guided. Furthermore, the customers’ continued participation also depends on
2169-3536 2015 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
2613
A. Naeem et al.: Understanding Customer Behavior in Multi-Tier DRM Program
their satisfaction and previous experience with their service providers. A survey conducted by Opower, which is a software enterprise that provides a platform to utility companies to engage with their customers, shows that the customers’ satisfaction depends on the quality of communication that the energy provider establishes during the critical moments of their lives [11]. These critical moments may vary from receiving an unusually high bill to experiencing an unexpected power shutdown. Furthermore, the customers look forward to receiving customer support in the form of alerts and personalized information during these times. Therefore, the existence of a quality communication between the utility provider and the customer that clearly specifies the energy usage, billing and management content is necessary to ensure customers’ participation.
such programs [19]. Consequently, designing appropriately specified DRM programs and understanding customer behavior becomes necessary, which requires multidisciplinary research in engineering, social sciences and economics [20].
A. STATE OF THE ART
In [9], [10], [12], and [13], various authors have proposed user-friendly DRM programs for increased and active customer participation. In [9], the authors develop DRM plans that specify the inconvenience parameters for customers in terms of scheduling delays and temperature deviations of various flexible devices. The authors then determine the effectiveness of such plans for the peak load reduction. The authors in [10] proposed a DRM scheme, while considering the impact of time varying price and incentives on user behavior. In [12], the proposed DRM system employs a distributed, game theoretic approach in which each customer gets to select a daily schedule of household devices to reduce the energy consumption. López et al. [13] model the customer participation behavior using a non-cooperative game. In this game, the grid tries to reduce the peak load while users decide, based on other users’ participation, whether or not they want to participate in a given DRM program. Even though, efficient schemes have been proposed in the literature, most of them assume that users are willing to participate if their objective functions are optimized. However, the works in [14]–[16] suggest that the responses and choices of individuals tend to vary from the assumptions on which various DRM schemes are based. Various demand response research projects have also been initiated across the world. The adoption of DRM techniques in smart grid installments is dependent on several factors including state policies, load demand profiles and technological growth. Some case studies on the deployment of DRM schemes in the smart grid environment are available in [17]–[19]. These studies also discuss the barriers that inhibit the customers’ participation. Reference [18] further strengthens the observation that the lack of customer engagement plans is a major hindrance for residential customers to adopt the time varying rates, where they are given the option to participate. For this reason, the New York Department of Public Service (NYPSC) has identified the key approaches for maximizing the customer engagement by redefining the marketing strategies and emphasizing the need for considering the customer diversity while designing 2614
FIGURE 1. Contribution of this paper.
B. CONTRIBUTIONS
In this paper, multiple DRM plans with clearly defined incentives are being offered as products to customers. As is shown in Figure 1, we identify and broadly categorize the customer behavior into two types 1) Non-Green Comfort Seeking Behavior (NCSB) and 2) Green Incentive Seeking Behavior (GISB). The grid operator then use these categories to specify the DRM plan parameters accordingly. Generally, an exact determination of customers preferences and behavioral patterns requires one to conduct an extensive survey for which numerous resources are needed. However, in the absence of these resources, appropriate mathematical models can be used for determining the customers’ behavior. Thus, the novelty of our work lies in the idea of designing different DRM plans to cater different categories of users and also, in using suitable models to understand the customer’s decision making process towards these plans, while considering the interests and objectives of the grid operator. By incorporating the customer behavior in the study, we are able to quantify the difference between the expected and actual benefits of the VOLUME 3, 2015
A. Naeem et al.: Understanding Customer Behavior in Multi-Tier DRM Program
grid operator that exist due to the deviation of the number of customers in the plans from the ideal one. To accommodate different customers’ behavioral patterns, we build on the multi-tier DRM plans proposed in our earlier work [9], [21]. The DRM plans for NCSB customers offer less inconvenience and less incentive while that for GISB customers offer relatively higher inconvenience and incentive. As Figure 1 shows, the customers subscribe to either NCSB plan or GISB plan. Their decision making process is influenced by a number of factors, including peer pressure, social standing etc., as well as the post-subscription behavior that accounts for the analysis of the DRM plan in its implementation. The number of customers subscribing to different plans will determine the profit margin for the grid operator. Grid operator can obtain maximum profit gain only for certain combination of NCSB and GISB customers. However, as human behavior is affected by lot of social and economic factors, obtaining ideal subscription in each plan could not be possible for the grid operator. We therefore develop a mathematical model based on logistic regression that takes into account the quantifiable as well as unquantifiable attributes of customer behavior and captures the probabilistic nature of customer preferences for different DRM plans. Simulation results show that the actual subscription of customers in NCSB and GISB plans significantly deviate from the ideal values. For this reason, we find the compromise solution which turns out to be an intermediate point at the intersection of actual and ideal curves. In contrast to the ideal solution, the compromise solution requires comparatively lesser number of customer to switch their behavior from NCSB to GISB or otherwise. We also identify that along with economic factors, social factors and prompting green or caring behavior can also serve the grid operator as potential tools to influence customers to switch to the plan that favors the grid’s objectives. Thus, this study can be used by the grid operator to its advantage by adapting appropriate measures to achieve the compromise solution. The rest of the paper is organized as follows. In Section II, we categorize customer behavior towards DRM in smart grids. In section III, we discuss multi-tier DRM plans and social and economic factors that influence customer decision making process. In section IV, we discuss two methods to determine actual customer preferences towards some specific DRM plan. In this section, we also develop a mathematical model of customer preferences. A case study and simulation results are presented in Section V, followed by the conclusion in Section VI. In the last, we present the derivation of logit probabilities in Section VII. II. CUSTOMER BEHAVIOR CATEGORIZATION TOWARDS DRM
In this paper, our objective is to categorize customer behavior towards DRM in smart grids. Further, as we propose multi-tier DRM plans, taking insights from social sciences and economics, we will also identify important factors that influence customer’s attitude towards a certain plan. VOLUME 3, 2015
These factors are not only directed by the environmental concerns and awareness but also by the economic concerns, such as price and financial incentives [22]. In order to categorize customer behavior towards DRM in smart grids, it is important to understand that a traditional grid has always operated by increasing electricity supply to meet customer demands. Customers in traditional grid over several decades have therefore acquired distinct electricity consumption patterns. Deviation from this consistent consumption routine is generally perceived as inconvenience (or discomfort) by the customers and hence the resulting reluctance to participate in DRM programs [23]. Furthermore, the authors in [24] have also identified biases that affect energy consumption, including, e.g., maintaining a status-quo and opting for the good instead of the best available option. Thus, customer behavior towards DRM, which requires a change in customer electricity consumption routine, not only depends on the amount of inconvenience but also on the demographics, monetary preferences, source of energy etc. In [25], the authors used data from the Dutch National Bank Household survey to understand energy conservation behavior of residential customers. The survey particularly focused on the impact of customer’s financial standing on the energy conservation related decision making process. They concluded that some customer’s attitude towards energy conservation is greatly linked to environmental concerns. On the other hand, financially rich or older customers tend to prefer comfort over conservation. Furthermore, the energy conservation behavior was found to be independent of the public awareness programs. Similarly, the work in [26] showed that the consumers using ‘‘green’’ (renewable) energy sources do not prefer to switch to other sources. On the contrary, a greater number of consumers who obtain their energy from the traditional energy sources such as, fossil fuels, are more willing to switch to the greener energy source. Based on these studies, we can broadly categorize customer behavior towards DRM into following two types1 : 1) Non-Green Comfort Seeking Behavior (NCSB): Customers with this behavior prefer comfort and are less willing to accept high inconvenience irrespective of the incentives being offered. 2) Green Incentive Seeking Behavior (GISB): Customers with this behavior prefer energy conservation and are more willing to accept high inconvenience but also require high incentive or reward. In order to attract customers with different behavior towards DRM, it is obvious that we would require multi-tier DRM plans. DRM plan for NCSB customers should generally offer less inconvenience and less incentive as compared to another DRM plan designed to attract GISB customers. It is important to note that customer selection of a DRM plan despite their behavioral tendencies strongly depend on the offered plan parameters, i.e., the values of inconve1 Note that there might be sub-categorizations of customer behavior but we do not consider them in this paper.
2615
A. Naeem et al.: Understanding Customer Behavior in Multi-Tier DRM Program
nience and incentive. Therefore, changing the plan parameters can lead to a different combination of customers in the two plans. For example, some GISB customers may select a low inconvenience plan if the incentive associated with high inconvenience plan is not good enough for their liking. By accepting a lower inconvenience value plan, such customers are still contributing towards the green initiative. Therefore, attracting a specific number of users in each plan for some desired grid objectives (e.g., a certain profit margin) by designing multi-tier DRM plans with appropriate incentive and inconvenience values is quite challenging. It is also important to note that several social and economic factors also influence the customer’s selection of a particular plan. In the next section, we describe our multi-tier DRM plans and discuss customer subscription pattern. III. MULTI-TIER DRM PROGRAM
As previously discussed, a major hindrance for customer participation in DRM plans is the lack of clarity in defining inconvenience and associated incentives, e.g., how many times a user will be required to change his/her routine electricity consumption pattern and how much compensation will be offered. Additionally, in order to maximize participation, different plan parameters are required to accommodate different customer behavior. We propose a multi-tier DRM program with two distinct DRM plans to accommodate NCSB and GISB customers. We consider a residential community comprising of K customers or homes, each of which subscribes to DRM program. Each home is supported by an interface home controller that provides the two-way communication link between the grid and home. Home controller is also responsible for an exchange of aggregate load profile with the grid operator and the execution of DRM plans. We assume that the flow of power is unidirectional i.e., from the grid to the home
via home controller. Figure 2 elaborates the concept of such residential community. We assume that every customer has a set of electrical appliances, which can be categorized into essential and flexible loads. For DRM, power consumption and scheduling of flexible devices can be controlled. Flexible devices are further divided into shiftable and thermal loads. As the name suggests, the operation of shiftable loads can only be delayed from the customer’s preferred time slot, e.g., clothes dryer (CD) can be operated half an hour later than usual but whenever it is operated, it will consume the same amount of power. On the other hand, thermal loads are generally used to regulate temperature and by changing the thermostat set point, power consumption of these devices can be changed (Air Conditioner (AC) is one typical example of thermal load). Figure 3 shows how shiftable and thermal loads operate with and without DRM. Delaying shiftable devices and changing the thermostat set point for a certain time duration creates inconvenience for customers. In our previous work [9], we proposed DRM plans, that specified inconvenience of various flexible loads by the following parameters: •
•
Shiftable loads: Scheduling delay in minutes from customer’s preferred starting time for the operation of the shiftable load, denoted by 1tˆ(j) (index j denotes j-th shiftable load). Thermal loads: Temperature deviation from customer’s defined thermostat set point calculated using two parameters θ ref (j) and α(j) and the total duration when the grid operator changes the set point, denoted by t˜max (j) (index j denotes j-th thermal load).
Further details of these parameters can be found in [21] and [27]. We assume that Is and It respectively denote the set of shiftable and thermal loads. In this paper, for NCSB and GISB customers, we define two different DRM plans such that
FIGURE 2. Residential community subscribed to DRM programs. 2616
VOLUME 3, 2015
A. Naeem et al.: Understanding Customer Behavior in Multi-Tier DRM Program
FIGURE 3. Shiftable and thermal loads.
ref (j) ≤ θ ref (j), ∀j ∈ I , 1tˆNCSB (j) ≤ 1tˆGISB (j), ∀j ∈ Is , θNCSB t GISB max (j) ≤ t˜max (j), ∀j ∈ I αNCSB (j) ≤ αGISB (j), ∀j ∈ It and t˜NCSB t GISB (the subscripts NCSB and GISB are used to denote the plan parameters and variables related respectively to NCSB and GISB customers). In this paper, we combine the inconvenience of various flexible loads into a single normalized value that is determined by the following equation [21]:
V˜ =
1
X
Vˆ max
j∈Is
1tˆ(j) +
X
t˜max (j)(1 + α(j))
(1)
j∈It
where, Vˆ max denotes the maximum inconvenience value (this value is set by the grid operator, details can be found in [21]). It is obvious that using (1), V˜ NCSB ≤ V˜ GISB . After defining inconvenience, the grid operator is also required to announce an incentive. Again as in [21], we assume a polynomial relationship between the incentive and inconvenience, i.e., h = aV˜ r , where, r = 1, 2, . . . and a is some constant, whose value is defined by the grid operator. The higher the value of exponent r, the higher will be the incentive. Similarly, the higher the value of a, the higher will be the incentive offered by the grid operator for a given value of inconvenience. If we assume a linear relationship between incentive and inconvenience, a can be viewed as the slope of the straight line. In the rest of the discussion a will be identified as the incentive-inconvenience ratio. Grid operator can obtain a certain financial advantage by controlling the flexible loads of participating customers (i.e., by creating inconvenience). We assume that this financial advantage is due to the peak load reduction resulting from DRM. However, it is important to note that the peak reduction and hence the financial gain to the grid operator starts to saturate (and the returns diminish) as the inconvenience values of DRM plans are further increased [9], [21]. VOLUME 3, 2015
TABLE 1. Multi-tier DRM plans for NCSB and GISB customers.
In a multi-tier DRM program (of given inconvenience values), it is therefore, more convenient for the grid operator to fix a certain profit margin (or a target peak load reduction) and find the ideal number of customers in each plan (without considering consumer behavior) in order to maximize the value of a. We used this approach in [21] and determined the incentive that the grid operator can associate with the DRM plans. A sample multi-tier DRM program for NCSB and GISB customers with the associated values of inconvenience and incentive is given in Table 1. In this table, we assume two shiftable loads, i.e., Clothes Dryer (CD) and Dish Washer (DW) and two thermal loads, i.e., Air Conditioner (AC) and Water Heater (WH). The value of Vˆ max is assumed to be 1440. Using (1), we find V˜ NCSB = 0.0979 and V˜ GISB = 0.2375, i.e., the inconvenience value for GISB customers is almost 2.42 times higher than NCSB customers. The incentive values associated by the grid operator for the linear and quadratic cases are also given in this table (more details can be found in [21]). In a community comprising of K customers, let KNCSB and KGISB respectively denote the number of customers subscribing to NCSB and GISB plans. In the ideal case, i.e., for an optimum number of users in each plan, the grid operator is able to reduce peak load and obtain a maximum profit margin after paying out incentives to the participating customers. 2617
A. Naeem et al.: Understanding Customer Behavior in Multi-Tier DRM Program
FIGURE 4. Five stages of customer’s selection process.
However, in practice, the number of users subscribing to each plan may substantially deviate from the ideal values. For this reason and as described in the previous sections, understanding customer behavior and response to energy management programs is crucial in the development and implementation of any DRM scheme. It is important then to outline some social and economic factors that influence a customer’s decision making process. Social and Economic Factors that influence Customer’s Decision Making Process: Customer’s preference for a given DRM plan is influenced by factors such as the environmental and social constraints, past experiences, accumulated judgment and peer pressure. These elements reflect the subjective nature of the customer’s selection within the limitations of certain parameters (such as, the type or income of the user). Generally, a consumer’s decision making process to purchase some commodity is viewed in five stages [28]. These stages for a multi-tier DRM program are as shown in Figure 4. Below we explain these stages: 1) Demand Realization: In the first stage, a customer realizes his/her electricity demand and consumption pattern, e.g., the dish washing time, thermostat set point of AC, number of hours AC is turned ON in a day etc. The customer also realizes that changing the electricity consumption pattern by subscribing to a DRM plan can cause inconvenience but will also be economically beneficial. 2) Gathering Information: The customer in the second stage thoroughly studies the offered plan and understands which devices are associated with each plan. Consumer also tries to understand what distinguishes one plan from another within the multi-tier DRM program and gather information about any other offers available in the market. Customer might also obtain feedback from his peers about the DRM plans and also about their subscription preference. 3) Assessment of Alternatives: In the third stage, the customer evaluates all the information to assess the pros and cons of each plan. 4) Decision - To subscribe or not to subscribe: Once the consumer has made a decision, he informs the grid about his decision and completes all the necessary formalities. 5) Post-Subscription Evaluation: The final stage for the customer is to analyze the program in its implementation. This is the last stage where the answer to the question ‘‘Was the decision correct?’’ gets answered. The user might be bounded for a fixed period or may have a leverage to switch in case of dissatisfaction. That, however, will depend on the policy adopted by the grid. In case, more customers select a plan that goes against the 2618
grid operator’s expectations, an understanding of these stages of decision process will assist the grid operator to come up with viable strategies to change the customer’s decision by offering counteractive measures. For example, in this selection process, we can see that at the information gathering stage, customer’s decision can be influenced by the peers or by the general perception of the plan. Additionally, the grid operator can also draw more attention to specific plans by giving them specific labels e.g., ‘‘green plan’’ or ‘‘caring plan’’ in order to attract more participating in such plans. We will discuss these issues in more detail in later sections. In the next section, we discuss different ways to determine actual preference of a customer towards a DRM plan. IV. DETERMINING CUSTOMER PREFERENCE IN MULTI-TIER DRM PROGRAM
In a multi-tier DRM program, there are two ways to determine the preference of a customer towards a certain plan 1) customer preference survey 2) customer preference modeling. Both methods have their advantages and disadvantages. A. CUSTOMER PREFERENCE SURVEY IN MULTI-TIER DRM PROGRAM
In customer preference survey, every customer is asked to answer a certain set of questions in order to determine his behavior and preference for some particular DRM plan. In this method the design of a questionnaire is a challenging task. The survey form can be categorized into fundamental and detailed questions. The fundamental questions may aim at collecting the background information and social behavior of the user. For instance, questions like Are you aware of the parameters announced by the grid? or How much inconvenience level are you willing to tolerate? Answer to these questions can be marked on a scale of 1-10 with 1 being the lowest and 10 being the highest. The responses to this section will help in understanding the reasons, which prompt the users to select some specific plan. The second module of the survey can include some detailed questions on the understanding of the respondents and they may assist in reviewing and revising the current DRM policies to adapt to the demands of users and to the grid. For instance, questions like What do you understand by DRM in Smart Grids? or How well are you aware of the functions of Advanced Metering Devices installed at your household? To assess the user’s response, certain weights can also be assigned by finding keywords in the answers provided by the consumer. For the former question, the terms like peak demand, peak hours or ancillary generation will show a higher level of familiarity to the concept. Generally, an exact determination of customers preferences and behavioral patterns requires one to conduct an extensive survey, which requires lot of resources. A survey conducted on a reduced sample size has a certain probability of error. The survey would be conducted by the grid operator. When the multi-tier DRM programs that are designed in this paper will be implemented or offered to customers by the grid operators, such detailed surveys would be conducted. VOLUME 3, 2015
A. Naeem et al.: Understanding Customer Behavior in Multi-Tier DRM Program
In the absence of such data, we can make use of mathematical models to determine customer preferences in such DRM programs. B. CUSTOMER PREFERENCE MODELING IN MULTI-TIER DRM PROGRAM
In this paper, we use the most widely used discrete choice method, ‘‘logit model’’, to determine the customer’s preference for a certain DRM plan. It is a logistic regression model that is used to understand the customer choice for a certain product. This model requires a product set, some quantifiable attributes of the products and unquantifiable attributes that are mostly related to customer social and physiological behavior. Based on the given product set, quantifiable and unquantifiable parameters, a utility function is associated with each product and the model then determines a certain probability of acceptance for a particular product. To use this model for our multi-tier DRM program, we have the following parameters: • Product set: In our paper, there are two DRM plans and we assume that the customers have to select one of these plans. With this assumption we require that the selection of plans must be mutually exclusive, i.e., customers must either subscribe to NCSB plan or GISB plan. • Quantifiable attributes: The noticeable quantifiable parameters of DRM plans are the incentives (hNCSB , hGISB ) and inconveniences (V˜ NCSB , V˜ GISB ). We assumes that the customers are completely aware of these plan parameters. Also note that in this paper we assume a polynomial relationship between incentive and inconvenience, i.e., incentive and inconvenience are dependent on each other. • Unquantifiable attributes related to customer behavior: This model also assumes a class of unquantifiable variables associated with customer behavior that affects his/her choice. It takes into account the user’s past experiences, peer pressure and general perception of any plan. All those factors that influence a person’s daily also life have a direct impact on his/her behavior as a customer of a particular product or service [16]. This observation is also supported by psychologists who explain that the pattern of customer decision and perception of products is formulated by their behavior and conscious experience. These unquantifiable attributes are a major reason that hinders exact predictability of human behavior. It is also important to note that the surveys are mainly designed to quantify these attributes. The unquantifiable attributes related to NCSB and GISB customers are lumped together and denoted respectively by NCSB and GISB . Based on the quantifiable and unquantifiable parameters, we define a utility function for NCSB and GISB customers, Ui = αi hi + i ,
i ∈ {NCSB, GISB}
(2)
where αi is a constant. The unquantifiable attributes NCSB and GISB are user specific and are modeled as independent VOLUME 3, 2015
and identically distributed (iid) random variables of extreme value of type-1 and are drawn from Gumbel distribution with variance π 2 /6). It means that the unobserved factor for PlanNCSB gives no information about that of Plan-GISB. Another way to think of this is to consider that the observed factor is sufficient to specify the utility while the remaining factors simply constitute the white noise. This leads to simplification of the expression of choice probability that takes the closed form and can be easily interpreted. The probability density function and the cumulative distribution functions are given by the following equations [29], f (i ) = exp−i exp(− exp−i ), F(i ) = exp(− exp
−i
),
i ∈ {NCSB, GISB}
i ∈ {NCSB, GISB}
(3) (4)
In this model, a customer subscribes to a plan that provides the greatest utility in equation (2). Therefore, customer preference for a DRM plan can be neatly summarized as follows: Subscribe to NCSB Plan if and only if UNCSB > UGISB , otherwise subscribe to GISB plan. With this characterization, the probability of a user subscribing to NCSB plan can be computed as [30], pNCSB = Prob(UNCSB > UGISB ) = Prob(GISB − NCSB < αNCSB hNCSB − αGISB hGISB ) = Prob(1 < αNCSB hNCSB − αGISB hGISB )
(5)
where 1 = GISB −NCSB . Note that the difference between two extreme value parameters follows a logistic distribution [30]. As NCSB and GISB are iid extreme value random variables, their difference given by 1 has the following distribution: e1 (6) 1 + e1 Using (6) the probability that a customer prefers NCSB Plan over GISB Plan is given by, F(1) =
eαNCSB hNCSB (7) eαNCSB hNCSB + eαGISB hGISB Equation (7) is known as the ‘‘logit probability’’ for choosing NCSB Plan. The derivation of logit probability using (5) can be found in [30] but is also provided in the appendix for the sake of completeness. By further simplifying equation (7), we get the probability that a customer prefers NCSB Plan over GISB Plan, 1 (8) pNCSB = 1 + eαGISB hGISB −αNCSB hNCSB Similarly, the probability that the user prefers GISB Plan over NCSB Plan is given by, pNCSB =
1 (9) 1 + eαNCSB hNCSB −αGISB hGISB We can easily verify that, pNCSB + pGISB = 1. Once the probabilities are obtained, the actual number of participants in a DRM plan can be determined as, pGISB =
Ki = pi × K
such that
i ∈ {NCSB, GISB}
(10) 2619
A. Naeem et al.: Understanding Customer Behavior in Multi-Tier DRM Program
where K is the total number of customers in a given community. In the next section, we present a case study to evaluate the difference between ideal and actual customer subscription behavior. We also give some policy recommendations to prompt a change in the behavior of customers. V. CASE STUDY
We consider a residential community consisting of 1000 customers. We assume that each customer uses a maximum of four flexible loads, two of which are thermostat loads and the rest are shiftable loads. The remaining household appliances form the essential base load. This allows the community to have a specific aggregated load demand curve with a distinct peak. We assume that the residential community has a peak load demand of 3.7 MW as given in [27]. We assume that the grid operator is aware of the peak demand and its primary objective is to reduce the peak load by regulating the power consumption pattern of the customers flexible loads. To control the customer’s power consumption profile, the grid operator offers Multi-tier DRM plans to customers and disseminates the information as given in Table 1. According to this table, two plans are offered to the customers to accommodate NCSB and GISB behavior. The amount of inconvenience for GISB customers is 2.42 times higher than NCSB customers. Similarly, the incentive offered by the grid operator for GISB customers is also higher than the NCSB customers. In our case study, we restrict our analysis to the linear and quadratic relationships between the incentive and inconvenience. We use αNCSB = αGISB = 100, while computing the values predicted by the model.
FIGURE 6. Ideal and actual number of GISB customers (linear case) for different values of a.
actual cases. We can see that as the value of a increases, the grid operator requires more and more NCSB customers, while customer preference model predicts otherwise. The trend is opposite for GISB customers, which is evident from Figure 6 as more customers would wish to subscribe to GISB plan for higher and higher incentive values. We can also see that the ideal and actual curves intersect approximately at KNCSB = 300 and KGISB = 700, where a = 0.07. This intersection point can be seen as a compromise between the expectations of grid operator and customer preferences. The solution provided by the intersection point is termed as the ‘‘compromise solution’’.
FIGURE 7. Ideal and actual number of NCSB customers (quadratic case) for different values of a.
FIGURE 5. Ideal and actual number of NCSB customers (linear case) for different values of a.
In Figures 5 and 6, we vary the value of a (i.e., the incentive-inconvenience ratio), while keeping the same values of inconvenience as given in Table 1: i.e., V˜ NCSB = 0.0979 and V˜ GISB = 0.2375. We use linear relationship between incentive and inconvenience. For example, if a = 0.05 then, hNCSB = 0.49¢ and hGISB = 1.19¢. Thus as the value of a increases, the incentives also increase. In Figure 5, we plot NCSB customers for the ideal and 2620
In Figures 7 and 8 we repeat the simulations for the quadratic relationship between the incentive and inconvenience. However, in these simulations we use higher value of a as compared to the previous case. This choice of values of a allows us to see the intersection point (a is defined as a ratio). Again it is evident that as the incentive payments increase, grid wants more customers in NCSB plan, while the customers prefer GISB plan. Interestingly, we can see that at the intersection point again KNCSB = 300 and KGISB = 700. The intersection point is roughly achieved around a = 0.2 i.e., 0.1916¢ for NCSB customers and 1.128¢ for GISB customers. VOLUME 3, 2015
A. Naeem et al.: Understanding Customer Behavior in Multi-Tier DRM Program
TABLE 2. Cost saving for the grid operator and total incentive payment for linear and quadratic relationship for the multi-tier DRM program of Table 1.
the NCSB plan, while its inconvenience is only 2.42 times higher than NCSB. The model predicts that the quadratic case will attract 157 customers in NCSB plan and 843 customers in GISB plan. Thus there is a huge gap between the ideal and actual number of participants particularly in the quadratic case. TABLE 3. Ideal, actual and compromise number of NCSB and GISB customers (linear case).
FIGURE 8. Ideal and actual number of GISB customers (quadratic case) for different values of a.
TABLE 4. Ideal, actual and compromise number of NCSB and GISB customers (quadratic case).
We now thoroughly investigate the impact of customer participation in different plans and its relation to the profit margin of the grid operator. Using the incentive and inconvenience values of the DRM program in Table 1, and the algorithm proposed in [21], in Table 2, we determine the %age peak load reduction, cost saving for the grid operator, the total incentive payment (i.e., KNCSB hNCSB + KGISB hGISB ) and the net profit to the grid operator for the linear and quadratic cases for various combinations of NCSB and GISB customers. We can observe that as the number of GISB customers increases, the cost saving as well as the total incentive payments by the grid operator also increases. The net profit of the grid operator thus changes. Notice that these calculations are done by the grid operator without any consideration of customer preferences. From this table, we can observe that for the linear case, the grid operator can gain a maximum profit of 18$ only if 200 customers subscribe to NCSB plan and 800 to GISB plan. For the quadratic case, again the grid operator can gain a maximum profit of 18$ but that requires 800 customers in NCSB plan and only 200 in GISB plan. On the other hand, customer preference model predicts that (for the given multi-tier DRM plans) the linear case will attract 292 customers in NCSB plan and the remaining 708 customers will be in GISB plan. For the quadratic case, the incentive value of GISB plan is 5.88 times higher than
In Tables 3 and 4, we summarize the ideal, actual and compromise solutions for the linear and quadratic cases. Note that compromise solution is the number of NCSB and GISB customers obtained at the intersection point of the curves in Figures 5, 6, 7 and 8. We can see that for the linear case, actual solution and compromise solutions are very close and they are also not far away from the ideal solution. This shows that for the linear case, obtaining a compromise solution would be much easier for the grid operator. The grid operator can also try to obtain the ideal solution as it is not that far away from the actual solution and would require convincing 92 NCSB customers to opt for GISB plan that is more green and also provides more incentive. On the other hand, for the quadratic case, actual as well as the compromise solutions are very very far away from the ideal solution. In this case, we can again observe that the compromise and actual solutions though far but are still close enough as compared to the ideal solution. Therefore it would be comparatively easier for the grid operator to attain the compromise solution that would require converting 143 GISB customers into NCSB customers. However, achieving the ideal solution would be
VOLUME 3, 2015
2621
A. Naeem et al.: Understanding Customer Behavior in Multi-Tier DRM Program
FIGURE 9. Using social influence to shift customer behavior.
extremely difficult as it would require 643 green customers to switch from green to non-green behavior. In the following, based on our study, we give some useful recommendations that could be helpful in influencing customer behavior: • Social Influence: The grid operator can use social influence to prompt behavioral change as is suggested by Figure 9. Social influence can cause a change in customer’s feelings, attitude and behavior towards a specific plan. If the grid operator wants more GISB users, green behavior could be emphasized and given more recognition. Similarly, peer pressure can also be used for the grid’s advantage as some individuals tend to change their beliefs as a result of either an interaction with another social group or an attempt to fit in. On the other hand, if more NCSB customers are required to obtain the ideal or the compromise solutions, the grid operator could instead give more recognition to the caring behavior for the elderly or comfort seeking customers. Apart from these, awareness campaigns through the social media, celebrity endorsements or an expert’s advice can also be used to switch the customers to the desired plan. It is also important that the grid operator should understand the customer base and if required launch plan switching campaigns. Furthermore, if the grid operator wants to retain the current number of customers in any plan, then it is essential for him to ensure the customer’s satisfaction. It can be done by delivering not only the promised services but also by conducting focus groups to show the utility companies’ concern for what their customers think. Moreover, the grid operator could also leverage the critical moments of the customers’ lives (i.e. when they experience unprecedented power outages or when they require billing information etc.) by providing appropriate customer support and 2622
•
information alerts. By enriching the customers’ experience when they seek information, the energy provider can ensure the customer’s satisfaction and continued participation in the program. Economic Influence: The grid operator can also come up with altogether different multi-tier DRM plans with different incentive and inconvenience values. The grid operator can also increase or decrease its net profit margin for certain time durations to prompt behavior switching. For example, grid operator can decrease its profit margin and increase incentives to attract more GISB customers.
VI. CONCLUSION
In this paper, we categorized customer behavior into two broad categories, NCSB and GISB. We proposed multi-tier DRM plans clearly specifying the incentives and inconveniences of various flexible devices in order to attract significant customer participation. We also developed a mathematical model to determine the actual preference of a customer towards a DRM plan of given parameters using logit Model. Taking insights from social sciences and economics, we also identified factors that could possibly influence customer decision making process. We then used a case study to further understand the ideal and actual number of customers that subscribe to various plans depending on the DRM plan parameters. We also identified a compromise solution that lies between the ideal and actual solutions. In some cases, it is thus easier for the grid operator to attain a compromise solution, which requires convincing lesser number of customers to switch their behavior. We also identified that along with economic factors, social factor such as peer pressure and prompting green or caring behavior could also be used as potential tools by the grid operator to influence customer VOLUME 3, 2015
A. Naeem et al.: Understanding Customer Behavior in Multi-Tier DRM Program
preferences to meet some of its desired objectives. This study can also be used by the grid operator to design appropriate multi-tier DRM plans. APPENDIX
In this section, we derive equation (7) from equation (5), following [30]. For the ease of notation, we refer to NCSB as Plan-A and GISB as Plan-B in the following derivation. The probability that the customer selects Plan-i where, i 6 = j and i, j ∈ {A, B} is given by, pi = Prob(Ui > Uj ;
Here we make change of variables. We consider u = e−i where −e−i di = du. The limits also change as follows; i −→ −∞, u −→ ∞ and i −→ ∞, u −→ 0 which gives us, Z 0 X e−(αi hi −αj hj ) (−du); pi = exp −u ∞
= 0
∀j 6 = i and i, j ∈ {A, B})
= Prob(j < i + αi hi − αj hj ;
This expression becomes the cumulative distribution for j evaluated at i + αi hi − αj hj , if i is given. In this case, according to equation (4), the cumulative distribution for j can be written as, (11)
Since ’s are independent, the probability for selecting Plan-i is given as, Y pi | i = exp(−exp − (i + αi hi − αj hj ));
exp −u
X
e−(αi hi −αj hj ) du;
j
∀j 6 = i and i, j ∈ {A, B})
F(j ) = exp(−exp − (i + αi hi − αj hj ))
j
∞
Z
P
e−(αi hi −αj hj )
exp −u j = P − j e−(αi hi −αj hj ) eαi hi pi = P α h ; j j je
∞ 0
∀j and i, j ∈ {A, B}
(16) (17)
Thus, the probability of selecting Plan-A is given as, pA =
eαA hA ; + eαB hB
eαA hA
(18)
which is the required expression.
j6=i
∀j 6 = i and i, j ∈ {A, B} (12) However, as i is not given, we take the integral of the above expression over all values of i times its probability density to get the probability, pi : Z ∞ Y −(i +αi hi −αj hj ) e−e e−i e−e−i di ; pi = −∞
j6=i
∀j 6 = i and i, j ∈ {A, B}
(13)
To evaluate this integral, we remove the restriction of i 6 = j from the product term. Note that if j = i, then αi hi −αj hj = 0. Thus, by adding ith term to the product and multiplying − by ee i , we get, Z ∞ Y −(i +αi hi −αj hj ) e−e e−i e−e−i ee−i di ; pi = −∞
j
∀j and i, j ∈ {A, B}
(14)
The last two terms in the product of the above equation cancels out and we get, Z ∞ Y −(i +αi hi −αj hj ) e−e e−i di ; pi = −∞
Z
∞
= −∞
Z
∞
= −∞
j
exp −
X
e−(i +αi hi −αj hj ) e−i di ;
j
exp −e−i
X
e−(αi hi −αj hj ) e−i di ;
j
∀j and i, j ∈ {A, B} VOLUME 3, 2015
(15)
REFERENCES [1] Y. Yan, Y. Qian, H. Sharif, and D. Tipper, ‘‘A survey on smart grid communication infrastructures: Motivations, requirements and challenges,’’ IEEE Commun. Surveys Tuts., vol. 15, no. 1, pp. 5–20, Feb. 2013. [2] P. Siano, ‘‘Demand response and smart grids—A survey,’’ Renew. Sustain. Energy Rev., vol. 30, pp. 461–478, Feb. 2014. [3] Y. Liu, C. Yuen, S. Huang, N. Ul Hassan, X. Wang, and S. Xie, ‘‘Peakto-average ratio constrained demand-side management with consumer’s preference in residential smart grid,’’ IEEE J. Sel. Topics Signal Process., vol. 8, no. 6, pp. 1084–1097, Dec. 2014. [4] J. S. Vardakas, N. Zorba, and C. V. Verikoukis, ‘‘A survey on demand response programs in smart grids: Pricing methods and optimization algorithms,’’ IEEE Commun. Surveys Tuts., vol. 17, no. 1, pp. 152–178, Mar. 2015. [5] N. Ul Hassan, X. Wang, S. Huang, and C. Yuen, ‘‘Demand shaping to achieve steady electricity consumption with load balancing in a smart grid,’’ in Proc. IEEE PES Innovative Smart Grid Technol. (ISGT), Feb. 2013, pp. 1–6. [6] C. W. Potter, A. Archambault, and K. Westrick, ‘‘Building a smarter smart grid through better renewable energy information,’’ in Proc. IEEE/PES Power Syst. Conf. Expo., Mar. 2009, pp. 1–5. [7] N. Ul Hassan et al., ‘‘Framework for minimum user participation rate determination to achieve specific demand response management objectives in residential smart grids,’’ Int. J. Elect. Power Energy Syst., vol. 74, pp. 91–103, Jan. 2016. [8] J. Fahey. (2013). Power Companies Dangle Free Nights and Weekends. [Online]. Available: http://www.poconorecord.com/apps/pbcs.dll/article? AID=/20130816/NEWS90/130819795/, accessed Dec. 7, 2015. [9] N. Ul Hassan, Y. I. Khalid, C. Yuen, and W. Tushar, ‘‘Customer engagement plans for peak load reduction in residential smart grids,’’ IEEE Trans. Smart Grid, vol. 6, no. 6, pp. 3029–3041, Nov. 2015. [10] P. T. Baboli, M. Eghbal, M. P. Moghaddam, and H. Aalami, ‘‘Customer behavior based demand response model,’’ in Proc. IEEE Power Energy Soc. General Meeting, Jul. 2012, pp. 1–7. [11] Survey Shows Utility-Customer Relationship Hinges on a Few Key Moments. [Online]. Available: https://opower.com/company/news-press/ press_releases/122, accessed Dec. 7, 2015. [12] A.-H. Mohsenian-Rad, V. W. S. Wong, J. Jatskevich, R. Schober, and A. Leon-Garcia, ‘‘Autonomous demand-side management based on gametheoretic energy consumption scheduling for the future smart grid,’’ IEEE Trans. Smart Grid, vol. 1, no. 3, pp. 320–331, Dec. 2010. 2623
A. Naeem et al.: Understanding Customer Behavior in Multi-Tier DRM Program
[13] M. A. López, S. de la Torre, S. Martín, and J. A. Aguado, ‘‘Demandside management in smart grid operation considering electric vehicles load shifting and vehicle-to-grid support,’’ Int. J. Elect. Power Energy Syst., vol. 64, pp. 689–698, Jan. 2015. [14] S. Gunasekara, ‘‘Analysis and mathematical modeling of consumer behavior in mobile telecommunications industry,’’ Int. J. Sci. Technol. Res., vol. 4, no. 6, pp. 333–343, Jun. 2015. [15] M. Khan, Consumer Behaviour and Advertising Management. New Delhi, India: New Age International, 2007, p. 4. [16] D. L. Loudon and A. J. Della Bitta, Consumer Behavior: Concepts and Applications. New York, NY, USA: McGraw-Hill, 1993, p. 8. [17] U.S. Department of Energy. (2014). Smart Grid System Report. [Online]. Available: http://energy.gov/sites/prod/files/2014/08/f18/SmartGridSystemReport2014.pdf, accessed Dec. 7, 2015. [18] Federal Energy Regulatory Commission Staff Report. (2014). Assessment of Demand Response and Smart Metering. [Online]. Available: http:// www.ferc.gov/legal/staff-reports/2014/demand-response.pdf, accessed Dec. 7, 2015. [19] New York Department of Public Service (NYPSC) Staff Report and Proposal. (2014). Reforming the Energy Vision. [Online]. Available: http://www3.dps.ny.gov/W/PSCWeb.nsf/All/CC4F2EFA3A23551585257 DEA007DCFE2?OpenDocument, accessed Dec. 7, 2015. [20] S. Uludag, P. Sauer, K. Nahrstedt, and T. Yardley, ‘‘Towards designing and developing curriculum for the challenges of the smart grid education,’’ in Proc. IEEE Frontiers Edu. Conf., Oct. 2014, pp. 1–8. [21] A. Shabbir, N. Ul Hassan, C. Yuen, A. Ahmad, and W. Tushar, ‘‘Multi-tier incentive scheme for residential customer participation in demand response management programs,’’ in Proc. IEEE ISGT ASIA, Nov. 2015, pp. 1–6. [Online]. Available: https://www.researchgate.net/ publication/283724728_Multi-tier_Incentive_Scheme_for_Residential_ Customer_Participation_in_Demand_Response_Management_Programs [22] S. Owens and L. Driffill, ‘‘How to change attitudes and behaviours in the context of energy,’’ Energy Policy, vol. 36, no. 12, pp. 4412–4418, Dec. 2008. [23] A. B. Jaffe and R. N. Stavins, ‘‘The energy paradox and the diffusion of conservation technology,’’ Resour. Energy Econ., vol. 16, no. 2, pp. 91–122, May 1994. [24] E. R. Frederiks, K. Stenner, and E. V. Hobman, ‘‘Household energy use: Applying behavioural economics to understand consumer decisionmaking and behaviour,’’ Renew. Sustain. Energy Rev., vol. 41, pp. 1385–1394, Jan. 2015. [25] D. Brounen, N. Kok, and J. M. Quigley, ‘‘Energy literacy, awareness, and conservation behavior of residential households,’’ Energy Econ., vol. 38, pp. 42–50, Jul. 2013. [26] D. Pichert and K. V. Katsikopoulos, ‘‘Green defaults: Information presentation and pro-environmental behaviour,’’ J. Environ. Psychol., vol. 28, no. 1, pp. 63–73, Mar. 2008. [27] Y. I. Khalid, N. Ul Hassan, C. Yuen, and S. Huang, ‘‘Demand response management for power throttling air conditioning loads in residential smart grids,’’ in Proc. IEEE Int. Conf. Smart Grid Commun., Nov. 2014, pp. 650–655. [28] P. Kotler, Marketing Management. Englewood Cliffs, NJ, USA: Prentice-Hall, 2003. [29] M. Ben-Akiva and M. Bierlaire, ‘‘Discrete choice methods and their applications to short term travel decisions,’’ in Handbook of Transportation Science. New York, NY, USA: Springer, 1999, pp. 5–33. [30] K. E. Train, Discrete Choice Methods With Simulation. Cambridge, U.K.: Cambridge Univ. Press, 2003.
AQSA NAEEM received the M.S. degree in electrical engineering from the Lahore University of Management Sciences (LUMS), Pakistan, in 2015, where she is currently pursuing the Ph.D. degree. She has been a Research Assistant with the Advanced Communication Laboratory, LUMS, for about two years, and has contributed to research on indoor localization. Her current research interests include various aspects of smart grids.
2624
ALI SHABBIR received the B.Sc. degree in electrical engineering from the University of Central Punjab, in 2010. He is currently pursuing the M.Sc. degree in electrical engineering with the Lahore University of Management Sciences, Lahore, with a specialization in embedded and electronics. His research interests include DRM in smart grids and IoT. Besides that he has also initiated two startups for providing affordable solar energy solutions and automation/control for households and commercial activities.
NAVEED UL HASSAN (M’08–SM’15) received the B.E. degree in avionics engineering from the College of Aeronautical Engineering, Risalpur, Pakistan, in 2002, and the M.S. and Ph.D. degrees in electrical engineering, with a specialization in digital and wireless communications, from the Ecole Superieure d’Electricite, Gif-sur-Yvette, France, in 2006 and 2010, respectively. In 2011, he joined as an Assistant Professor with the Department of Electrical Engineering, Lahore University of Management Sciences, Lahore, Pakistan. Since 2012, he has been a Visiting Assistant Professor with the Singapore University of Technology and Design, Singapore. He has several years of research experience and has authored/co-authored numerous research papers in refereed international journals and conference proceedings. His major research interests include cross layer design and resource optimization in wireless networks, demand response management, and integration of renewable energy sources in smart grids, indoor localization, and heterogeneous networks.
CHAU YUEN (S’02–M’08–SM’12) received the B.Eng. and Ph.D. degrees from Nanyang Technological University, Singapore, in 2000 and 2004, respectively. In 2005, he was a Post-Doctoral Fellow with Lucent Technologies Bell Labs, Murray Hill, NJ, USA. In 2008, he was a Visiting Assistant Professor with Hong Kong Polytechnic University, Hong Kong. From 2006 to 2010, he was a Senior Research Engineer with the Institute for Infocomm Research, Singapore, where he was involved in an industrial project developing an 802.11n wireless local area network system and actively participated in the third generation Partnership Project Long-Term Evolution (LTE) and LTE-A standardization. In 2010, he joined the Singapore University of Technology and Design, Singapore, as an Assistant Professor. He has authored over 200 research papers in international journals or conferences. He holds two U.S. patents. He received the IEEE Asia-Pacific Outstanding Young Researcher Award in 2012. He serves as an Associate Editor of the IEEE Transactions on Vehicular Technology, and was awarded as Top Associate Editor from 2009 to 2014.
VOLUME 3, 2015
A. Naeem et al.: Understanding Customer Behavior in Multi-Tier DRM Program
AYAZ AHMAD received the B.Sc. degree in electrical engineering from the NWFP University of Engineering and Technology, Peshawar, Pakistan, in 2006, and the master’s degree in wireless communication systems and the Ph.D. degree in telecommunications from Ecole Superieure d’Electricite, Gif-surYvette, France, in 2008 and 2011, respectively. He is currently an Assistant Professor with the Department of Electrical Engineering, COMSATS Institute of Information Technology, Wah Cantonment, Pakistan. His research interests include resource optimization in wireless communication systems and energy management in smart grid.
VOLUME 3, 2015
WAYES TUSHAR (S’06–M’13) received the B.Sc. degree in electrical and electronic engineering from the Bangladesh University of Engineering and Technology, Bangladesh, in 2007, and the Ph.D. degree in engineering from Australian National University, Australia, in 2013. He was a Visiting Researcher with National ICT Australia, ACT, Australia. He was also a Visiting Student Research Collaborator with the School of Engineering and Applied Science, Princeton University, NJ, USA, during summer 2011. He is currently a Research Scientist with the SUTD-MIT International Design Center, Singapore University of Technology and Design, Singapore. His research interest includes signal processing for distributed networks, game theory, and energy management for smart grids. He was a recipient of two best paper awards, both as the first author, in the Australian Communications Theory Workshop in 2012 and the IEEE International Conference on Communications in 2013.
2625
