ICT Application of DSM

S. Nithin

INTRODUCTION

Maximizing RE utilization is the ultimate goal for utilities in realizing the green energy concept. However, the uncertainties in RE generation force utilities to employ RE curtailment which weakens green energy initiatives. The scope of demand side management (DSM) strategies to handle the stochastic nature of demand and generation has attracted smart energy research efforts. The convergence of energy systems and ICT to enable smart operations has led to the development of smart versions of DSM like demand response (DR) and demand dispatch (DD). The ICT applications like WAMPAC and AMI have been discussed in the previous chapter. The DSM is yet another ICT application for smart grids.

This chapter introduces the DSM schemes of DR and DD and then reports in detail the development of a DD framework. Testing of the DD algorithm on the smart microgrid emulator as well as on a DC microgrid is also presented.

DEMAND RESPONSE

The demand for electric power is ever increasing and utilities are finding it difficult to manage the fluctuations in demand. The concept of distributed generation promotes interconnection of more RE sources, however the intermittency and variability in RE aggravates the existing demand-supply mismatch. The DSM strategies address the issues of demand-supply mismatch. DSM gained popularity owing to its efficacy in shaping the demand curve. Demand response is one such load management technique put in practice by the grid operator in order to regulate the demand as and when required. DR programs were originally proposed to curtail some of the bulk loads during peak times, via telephone calls. DR is usually associated with only a few consumers and is used infrequently. In a DR contract with the utility the consumer entrusts the utility to disconnect the contracted load when the latter feels the need. DR is an effective tool; however, its success depends on the identification of controllable loads. Utilities assign this task to demand side response aggregators who collect and control DR loads upon instruction from the utilities. The DR switches, w'hich receive control signals from utilities, are deployed on the consumer side and to control the loads. The DR programs could be primarily divided into two categories: (a) Dispatchable and (b) non-dispatchable/time-based. Detailed classification is given in Figure 4.1.

FIGURE 4.1 DR classification.

FIGURE 4.2 DSM control.

Communication and Control

The communication infrastructure of AMI aids the utility in deployment of DR operations. The introduction of smart appliances under IoT paradigm and smart home facilitates DSM by utilities. Typical DR communication is depicted in Figure 4.2.

The DSM strategy of DR was introduced by FERC in 2006 to handle demand during peak hours of consumption. The success of the early versions of DR motivated the development of more sophisticated versions, through which behind-the- meter resources - such as loads, RE generation, diesel generators, etc., are controlled to maintain supply-demand synergy.

DEMAND DISPATCH

The recently evolving DSM strategies can handle the stochastic nature of both demand and generation. The convergence of power systems and ICT to enable smarter operations lead to the development of the advanced version of DR, called demand dispatch (DD). Advanced communication technologies helped to extend DR schemes as DD, where the utility can aggregate consumer loads and dispatch them in tandem with the generation, thus forcing demand to follow generation. DD uses the concept of dispatchable loads (DLs). The DLs are loads that can deviate from their normal consumption pattern without affecting their operational constraints. For the success of DD, the utility needs to identify, aggregate and precisely control DLs. Traditional power system operation (i.e., where generation follows demand) is no longer effective in the context of variable RE. The availability and controllability of DLs are of prime concern for DD. The contract between the utility and the consumers ensures seamless control of DLs whenever needed by the utility.

Multiple demand side aggregators are needed for DD operations in a large utility grid. Since DD is a new management strategy, the operational framework and the various entities involved in operations are yet to be identified and standardized.

The DD requires real-time communication between appliances/loads and utility control centers. Latency of communication within the grid is of great importance as some loads should be dispatched first in cases of power system urgency. Internet is a promising medium through which DD could be achieved, as the infrastructure already exists. The utility can implement a dynamic tariff plan in conjunction with DD so that the change in price for electric power generation is reflected to the consumers as well. Dynamic tariffs will promote voluntary involvement of consumers in DD schemes. Moreover, the consumer can select the operation pattern of loads in such a way as to minimize his total consumption.

The major challenge to be addressed in DD is this customer participation aspect. The motivating factors for participating in a DD or DR program are the associated incentives. For DD/DR schemes to succeed, the consumer should be aware of dynamic tariff, time ratings and their own consumption patterns. To tackle these issues, the DD/DR schemes should be automated to incorporate consumer preferences in load connection/disconnection decisions w'hile maintaining synergy with grid operations. The utility has to aggregate and precisely control dispatchable loads to nullify the fluctuations induced by variable RE, like wind farms. The optimum selection of dispatchable loads based on spatial constraints is crucial as the changes in demand would affect voltage stability and power flow in the grid. The utility has to select appropriate communication technology and security measures conforming to various standards like ISO 27002, NIST SP800- 53, NERC Cl, ISA.

Role of Aggregators

The utility operators can employ DD through direct load control; however, it is desirable for huge power systems to use aggregator services. The architecture of DD is evolving; therefore, DD architecture, which is similar to that used for DR, is discussed here. The architecture comprises of three layers: The first layer involves the utility, demand dispatch provider (DDP) and independent power producers (IPP); these entities interact in the electricity market. The IPPs forecast their net generation on a day-ahead basis and tender their price accordingly. Utilities in the first layer bid for IPP forecasted generation and request DDP service in anticipation of forecasting errors. The DDP relies on several demand dispatch aggregators (DDA) for collecting information on DL from consumers, controlling DL and providing incentives to consumers for their participation in DD. The DDAs can be broadly classified into: (1) Commercial DDA, (2) industrial DDA, and (3) residential DDA. The role played by each DDA is summarized in Table 4.1.

The second layer of the architecture consists of the numerous DDAs performing DD operation based on the requirements of DDP. Dispatchable loads under the control of commercial, industrial, and residential DDAs form the third layer of the architecture.

The DDP requests DDA to collect DL information and announces the dispatch capacity required for each time slot of operation. DL owners belonging to each sector enroll their DLs on a day-ahead basis against the incentives offered by DDAs. Finally, DDAs

TABLE 4.1

Types of DD Aggregators

FIGURE 4.3 DD architecture CDDA: Commercial DDA; IDDA: Industrial DDA; RDDA: Residential DDA; CDL: Commercial DL; IDL: Industrial; RDL: Residential DL.

submit bids to DDP for service cost against dispatch capacity. The DD architecture is depicted in Figure 4.3 and information flow in the architecture is depicted in Figure 4.4.

Each aggregator announces the extent of demand they can change (increase/ decrease) for each operational time slot on a day-ahead basis. Aggregators do this on the basis of DLs enrolled by the consumers and priorities assigned by the consumers for each DL. Aggregators collect such information through an online portal, where consumers can enroll their DLs, set minimum operational time of each DL in a day and assign a priority for each DL. It is required that aggregators are equipped with such data, have assigned a priority rating for each DL considering consumer preferences and have announced the service charge for each time slot.

The incentives offered to consumers vary across the day and hence the service costs of aggregators vary with time. The service cost per kW for each operational

FIGURE 4.4 Information flow in DD architecture.

time slot is also announced by the aggregators on a day-ahead basis. This variable incentive scheme attracts more consumers to participate in DD operations. The aggregators benefit as they are paid by the utility for the DD service. The utility has to compensate for the services offered by the aggregator and this further increases the cost of utility operation. At high levels of RE penetration, the goal of the utility is to dispatch more RE and maximize net revenue. The utility deploys DD to absorb the variable RE power injected into its system.

The role of aggregators is to facilitate DD and thereby maximize their profit, without sacrificing consumer priorities. A single aggregator might not be sufficient to facilitate DD in a large grid; when multiple aggregators are present, the utility needs to precisely estimate the volume of demand each aggregator has to dispatch as depicted in Figure 4.5. Among the numerous demand-side aggregators existing in the grid, the utility needs to judiciously assign a dispatch share, so that each aggregator can adjust DLs under its control, thus leading to optimum DD. This process is termed aggregator dispatch share allocation (ADSA). The ADSA should consider grid constraints with respect to voltage stability and feeder overloading.

FIGURE 4.5 Schematic of ADSA.

ADSA Formulation

The potential of DD to reshape the demand curve in real time is utilized here to adjust to changes in RE generation. A utility can employ DD to absorb RE generation directly or as a tool to account for unscheduled RE generation. Though the operational aspects are similar in both scenarios, the key factor is the availability of DLs in large numbers. It is assumed here that DD is assigned with the former task while DLs are available in surplus numbers from various DDAs.

Consider a utility grid with n buses where n_P4 buses feed loads. Let P_K be the instantaneous RE penetration at the time instant of t and AP_K be the change in RE injection with respect to the previous instant. This is expressed as,

where P_R(t) is the RE injection at the instant of t and P_R(t - 1) is the injection at the previous instant. It is assumed that the aggregator has the net DL capacity at each of the n_pq buses represented by Rdis_Patch(')e{l,2...n_M}, and let P_DD be the aggregate dispatch needed on each of the n_P4 buses at that instant. P_DD of all the buses, when aggregated, equal ДP_R that means,

where 0 < P_DD(i) < P_dispatch(/).

Let the utility procure RE at a rate of Rs. C_R/kWh, and pay Rs. C_DD/kWh as dispatch service cost to m number of aggregators. Let the utility sell energy to consumers at a rate of Rs. Q/kWh, where C_R < C, . The incentive given to the consumers will be Rs. C,/kWh. These power and cash flows are depicted in Figure 4.6. The net income (N1) of the utility can be expressed as,

The utility has to maximize the net revenue through DD; the corresponding maximization function is,

subject to constraints, C = V'_min(t')/(/)< У_тах(г), V ie{l,2,...,л} where Vj^and V_max are the lower and upper limits respectively of the bus voltages in per unit (p.u.).

Addition of extra demand on the grid has adverse effects on bus voltages and hence this is modeled as a penalty factor P, which is a vector of size n and can be expressed as,

The addition/removal of demand to maintain voltage within these limits is a constraint and hence P is used as a factor in calculating the cost. Penalty values are assigned as a function of AP_K, so that voltages at all buses are maintained within the limits by controlling the dispatched load. This is modeled as a cost, by multiplying the quantum of demand to be modified on the respective bus by a voltage penalty of C_vv. Here C_v represents the aggregate cost of voltage deviation on all the buses as,

The line flow limits are determined through Newton-Raphson (NR) iterative load flow analysis. The non-convergence of NR load flow is also modeled as a higher cost denoted as O.F,

From (3.3), (3.5), (3.6) and (3.7) the fitness F for each particle can be evaluated as,

The objective function is given in (4.4), fitness function is in (4.8) and constraints are given in (4.5) and (4.7). The particle swarm optimization (PSO) method is adopted here. Particles are generated in increments of P_DD at each bus. Their fitness is evaluated using (4.8) through NR load flow. Global best and personal best fitness values of particles are computed and particles are updated based on velocity. The PSO retire particles from the iteration once tolerance has been met; final particles indicate the demand to be dispatched on each bus and so the aggregators in charge of buses can control DLs, respectively.

Selection of Loads

The DD framework discussed here relies on DDAs for identification and aggregation of DLs from among commercial, industrial and residential consumers. This is a cumbersome task as such participants will be copious. An efficient way to address this is to rely on energy management systems (EMS) to be used in all the sectors. Home-EMS, building area network under smart grid paradigm and industrial IoT (IIoT) concepts could be utilized for the development of such EMS. Such an attempt is reported here to develop an EMS architecture irrespective of the operational sector. DLs can be smart appliances with built-in intelligence and communication. However, a low-cost solution can be adopted to transform existing appliances to be compatible with smart operations. Such solution may include a load control unit (LCU), a central EMS unit and necessary communication links. The schematic of EMS is shown in Figure 4.7.

FIGURE 4.7 Schematic of EMS.

The EMS depicted in Figure 4.7 ensures the enrollment of DLs for DD via mobile phone application. The mobile phone application establishes a bidirectional medium through which consumers and DDA can interact. Such interactions result in announcement of incentives and enrollment of DLs in a day-ahead basis. DDAs make use of these data and formulate their bids. The EMS could be equipped with smart operations devoid of DD, too. The suggested system will also enable remote control of non-DL, as well. The LCU modules could be equipped with multiple communication capabilities depending on the deployment requirements. LCU modules are capable of power measurement and control of DLs. Once the user registers a DL, its power consumption is fed to DDA by the LCU. LCUs, depicted in Figure 4.8, keep track of the power consumption of DLs and update the state of operation in the DDA server/cloud. DDAs can collect DL information and prioritize these based on consumer inputs such as minimum operational hours, DL schedule, etc.

Several criteria indices of DL prioritization, suggested for DR in the literature, can be used by DDAs in the case of DD as well. These include appliance priority index, appliance flexibility index, appliance satisfaction index, power similarity index, high power consumption index, etc. Furthermore, DDAs can inform each EMS unit of the amount of DLs to be connected/disconnected during the DD operation. EMS will

FIGURE 4.8 Schematic of LCU.

follow the instructions from DDA, and if the dispatch share is less than the available DL capacity, then EMS can prioritize DL operation considering consumer preferences as well.

The load selection process by the aggregator is a combinatorial optimization problem, where the aggregator should find DL matching the ADS. In fact, the DL selection problem exhibits similarities to 0-1 Knapsack problem, where a knapsack/ backpack is to be filled with items to maximize profit without exceeding the knapsack’s weight carrying capacity. Each item available for selection has a respective weight and value; the algorithm should select the best items so that their combined values are the highest and yet are well within the weight capacity of knapsack. Here, DLs available for DD are considered to be items, the rated power of each DL is taken as its weight and the priority assigned to each DL by the aggregator is taken as item value. The objective is to select high priority (value) DLs where the aggregate of selected DLs (weights) equals the required ADS. Selection of DLs modeled as a 0-1 knapsack problem, is given in equations 4.9 and 4.10, and solved through dynamic programming. Assuming nl number of DLs, ADS is taken as the knapsack’s capacity, DL ratings are taken as item weights and priority assigned to each DL is taken as item value.

subject to

where DLp_riorjlyj is the priority assigned to each DL, DL_rating. is the power rating of each DL and S, denotes if a DL is selected or not. Dynamic programming divides a problem into sub problems, solves each of these and combines the solutions. The recursion formula for the knapsack is,

where к e {l, 2g = ADS, V_k = net value considering Ath DL, W_k = power rating of A'th DL.

Equation (4.9) represents the objective function, where DLs with high priorities are selected. The selection of high priority DLs should be performed with a constraint that the aggregate of selected DLs’ power consumption should not be higher than the required power to be dispatched; hence it is subjected to the constraint given in (4.10). The recursive operation in (4.11) checks each DL’s priority and rated power based on dynamic programming, then decides to select or reject it based on (4.9) and (4.10).