Demand side management with consumer clusters in cyber-physical smart distribution system considering price-based and reward-based scheduling programs

This study presents a demand side management (DSM) strategy for a cyber-physical smart distribution system (CPSDS). The proposed approach uses the price-based as well as reward-based DSM schemes as a part dual objective function. The objectives of the proposed scheduling approach comprise the profit maximisation objectives of demand response aggregator agent (DRAA) and network performance objectives of the distribution system operation agent. The same are achieved by providing incentives to the participating customers. The incentive information is communicated to responsive load agent (RLA) using cyber infrastructure (communication channels, sensors and cloud storage systems) and thus allowing customers to select the incentive program of their own choice as per their flexibility. The real-time implementation of the program is considered to have direct load control based once the event trigger acknowledgement is received by DRAA/utility from RLA for control action on responsive loads. The proposed price-based and reward-based DSM framework in CPSDS is evaluated using IEEE 37 bus test system. The simulation results of proposed dual objective price-based and reward-based mechanism are presented, discussed and compared with single objective price-based approach. The same demonstrates the improved tradeoff between techno-economic aspects of distribution system operation.


Introduction
Demand side management (DSM) in smart grid paradigm is identified as a key practice in realising various objectives of system reliability, operation efficiency, economic efficiency and so on [1].Effective management of load schedules under DSM programs needs an efficient and sophisticated physical and cyber infrastructure.Along with the infrastructure, the communication protocols and the financial transactions via cyber systems network comprises a key aspect of DSM [2].The operation policy agreements and strategies between different stakeholders influence the net benefit of the DSM [1].The main scope and focus of the DSM programs is expected to be at distribution level considering the financial and physical ease in carrying out DSM transactions [3].Therefore, this paper addresses a cyber-physical framework for financial and physical DSM transactions in smart distributed system enabled by cyber-physical systems (CPS).
The integration of demand response (DR) programs into power system operation has been investigated extensively for all levels, i.e. wholesale market, transmission and retail market level.The wholesale market provides a cyber platform for various market players including DR providers (DRPs), DR aggregators (DRAs), distribution companies and so on, to engage and negotiate the financial transactions.The initial research was focused on integrating the demand reserves in wholesale energy markets in the evolving smart grid paradigm with various objectives such as operational excellence, system reliability and sustainability and so on [4].The reliability improvement under intermittent wind energy is then extended to DR programs [5].In the subsequent models, the concept of DRA has been brought up.The DRA participates in wholesale energy market as a seller and self-schedules the load reduction aggregated from small customers based on price forecasts to maximise the profit [6].The scheduling of DR services in wholesale market was also carried out as an independent system operator exercise alongside the generation and storage scheduling in smart grid [7].The dynamic pricing scheme is then used to observe the relation between load reductions, uncertain renewable energy and locational marginal prices [8].The variable payments to responsive loads were also realised through dynamic incentive based approaches designed by load serving entity (LSE) through direct load control [9].The optimal scheduling of responsive loads to maximise cost savings through arbitrage intraday wholesale prices was also investigated [10].Recently, to include the congestion effect, the locational marginal prices are used in allocating the compensation cost to LSEs [11].The other dynamic approaches for clearing DR services at wholesale market includes the dynamic market mechanism where continuous negotiations between consumer companies and generation companies is carried until the verge of convergence [12].
The development of cyber-physical infrastructure and operation environment at distribution level marks a more challenging and vital aspect of DSM program implementation.The earliest methodologies in realising DR schemes are associated with load shedding objectives [13].The two-day digital communication between customers and utility through communication channel such as LAN is a key infrastructure aspect in distributed system's DR program [14].The communication and cloud based systems within the distribution system are investigated through communication protocols implemented in laboratory test systems [15].Looking at the lab test systems, the needs and taxonomy of intelligent, communication and control systems were proposed for large-scale deployment of DSM programs in distribution grids [16].Such paradigms can improve the flexibility of electricity purchase for customer via framework like online purchase electricity now schemes [17].The flexible DR programs enable the agent-based scheme for responsive loads to purchase electricity from distributed generators and multiple utilities [18].The objective of DSM scheduling later has been extended to multiobjective model where distributed system network preferences are met without compromising on customer satisfaction [19].The same can be attained by continuous and seamless negotiations between various stakeholders through a secure and reliable communication channel [20].The information flow between distribution system agent and DRA agent (DRAA) with DRP can be established using sensor networks enabled at personal premises of DRP [21].The information received by such sensors can be directed to household appliances through the home area network or local area network systems [22].Therefore, the cyber-physical infrastructure and operation comprises a vital component in the efficient and reliable management of the DR services in distribution network.
The introduction of advanced sensing, communication and monitoring technologies in smart grid infrastructure has led to the transformation of traditional grid into a cyber-physical modern power system.The application of cyber-physical paradigm is to improvise smart grid operation range from home energy management system to grid scale deployment [23].In [24], an intelligent use of information and communication system in residential energy management is presented through Markov decision process.The home energy management in DSM is also addressed using virtual energy provisioning in a supply chain through automation system [25].The coordinated DSM operation of utilities and customers is investigated through cyber-physical platforms such as DRsim [26].The intelligent and efficient management of electric vehicles (EVs) is performed using cloudassisted CPS in the context of DSM in smart grid operation [27].Similar to residential loads, the cyber and physical aspects of industrial processes, data centre operation performance and management were also considered for DR programs.[28].The cyber-physical infrastructure and computer-aided operation of overall smart grid paradigm including smart buildings and smart cities can reduce overall energy consumption [29,30].At wholesale market level, concepts like power integrated cyber marketing have been examined for power market reform [31].The interactions between utility and customer are modelled through game theory models to obtain equilibrium between preferences of both customer and utility [32].The cyber-physical operation along with the scheduling strategies can enhance the effectiveness of DSM in smart grid.The contributions of the work presented include: • Development of dual beneficiary framework for DR programs in a cyber-physical smart distribution system (CPSDS) using price and reward programs at various levels/stages.• Customer characterisation and dedicated incentives have been used to implement strategic DSM in distribution system.• Through comparative analysis of the dual-objective two-stage DSM framework for various performance attributes.
The rest of the paper is structured as follows.The proposed framework with customer incentive schemes and different scheduling scenarios with objectives and solution algorithm are presented in Section 2. The test system description with network and market specifications is presented along with simulation results in Section 3. The same presents the comparison of proposed incentive-based framework with scenario with no DR and scenario with time of use (TOU) only strategy for DR scheduling.Finally, Section 4 concludes the paper and underlines the possible future directions in extending the work carried out in this paper.

Problem formulation
The responsive load scheduling in past has been associated with price-based and incentive methodologies.The price-based schemes responsive load scheduling at distribution system level are criticised for the unreliable exposure of DRP to market price variations.The incentive-based schemes thus serve as better schemes for DSM compared with the price-based option from demand side.The generic cyber and physical operation of supplydemand chain in distribution system is shown in Fig. 1.In this system, utility participates in wholesale market level and purchases power at market price.The distribution system loads are offered the energy at fixed or differential tariff decided by utility.The same utility offers incentives to responsive loads in case of event triggered load shift/curtailment.The event trigger is followed by an acknowledge flag from the customer.Thereupon, the DRAA reschedules the available flexible loads to maximise the profit through the interactions with wholesale market (in the form of market price) and incentive minimisation.Then, DRAA communicates the schedules along with the incentive information to responsive load agents (RLAs) for actuation and in return receives the revised load schedule confirmation from RLAs.The same is represented in Fig. 2.
In this paper, RLAs are not considered/needed to schedule their loads in response to the price or tariff, but are encouraged/obligated to response to the event trigger by DRAA.The DRAA issues a load curtailment or load shift trigger for RLAs registered under incentive-based scheme.The user can select a particular incentive of his choice/by the nature of load as a part of the contract.An event trigger is a signal sent by LSE to user/customer indicating the amount of load and the time stamp of load shift/curtailment.Customer will receive the payments for all such load alterations on either daily or monthly basis.Alternatively, the incentives can be negated from the electricity bill at the time of payments by customers.

Customer incentive programs
This paper considers DSM implementation through the incentivebased programs.In which, RLAs are presented with various categories of incentives and underlying operating conditions offered by DRAA for event-based DSM mechanism.Thereupon, customers are encouraged to select incentive category of their own depending on their load consumption pattern.Other approach is to group the customers to match the requirements of a particular incentive program.The different types of incentive categories along with the interrelations are presented as follows.The customer grouping based on incentive-based program is shown in Fig. 3.

Group I:
The loads of the customers in this group are designated as priority loads and are given least preference for load shifting/curtailment purpose by DRAA or distribution system operation agent (DSOA).Therefore, the DRAA/DSOA issues a DSM event trigger to these loads only when all the other responsive loads are scheduled to their maximum possible flexibility levels.In the proposed incentive-based scheme, the least rescheduling of priority loads is realised by offering high compensation/incentive for shift in per unit of energy consumption.The total incentive thus received by customers grouped in this section (or) customers who choose to maintain their loads with priority can be expressed as follows: where I denotes the group of customers who opt for being designated as priority loads, I represents the monetary incentive offered by DRAA per shift in unit energy consumption ($/MWh) for group I, Δ k represents the change in load during event and k I represents the sum of incentives received for event trigger response of kth RLA over the total scheduling period (24 h in this case).

Group II:
The group of customers whose loads are noncritical and deferrable in nature are grouped in this category of incentive.The compensation demanded by these non-priority loads is comparatively less than that of priority loads.In this paper, the lower compensation to the deferrable non-priority loads is realised by specifying lower incentive range which would enable the DRAA/DSOA to issue the DSM event trigger with utmost preference.The total incentive thus received by customers grouped in this section (or) customers whose loads are non-critical and are also deferrable nature can be expressed as follows: where II denotes the group of customers who are designated/ choose to behave as non-priority/non-critical loads, II represents the monetary incentive offered by DRAA per shift in unit energy consumption ($/MWh) for group II, Δ k represents the change in load during event and k II represents the sum of incentives received for event trigger response of kth RLA over the total scheduling period (24 h in this case).

Group III:
The group of customers who want to compromise on their load schedules/criticality when the incentives are relatively high are grouped into this third section.Therefore, customers of this group tend to reschedule their loads upon event trigger, only if considerable monetary incentive is given.The same aspect is realised in this by assigning a comparative monetary incentive to group III customers.The total incentive thus received by customers grouped in this section (or) customers whose loads can be altered more at higher incentives can be expressed as follows: where III denotes the group of customers who are designated/ choose to behave as non-priority/non-critical loads for higher incentives, III represents the monetary incentive offered by DRAA per shift in unit energy consumption ($/MWh) for group II, Δ k III represents the change in group III load during event and k

III
represents the sum of incentives received for event trigger response of kth RLA over the total scheduling period (24 h in this case).The preferences (or) sequence of the event triggers to the RLAs depends on the amount of incentives to be paid for customers grouped under various incentive programs.The order of incentives assigned for three groups as per the incentive programs are defined as follows: In a typical distribution grid with residential, commercial and industrial loads, the proposed approach of different incentives can be fit using the load criticality and monetary preferences of customers.The residential loads being non-critical are set to have lowest compensation cost/lowest incentive ( II as given in ( 4)).
Whereas, the industrial being designated as non-deferrable and critical loads are assigned with highest compensation and thus choose to register under highest incentive ( III as given in ( 4)) program.Finally, the commercial customers with both critical and deferrable loads are assigned with medium values of incentives ( I as given in ( 4)) which balances the choice between criticality and monetary incentives.It would be an interesting insight to allow the customers from one to another incentive program.For simplification, this study assumes that the customer select a particular incentive program at the start of day.However, the concept presented in the work can be extended with customer shifting from one to other incentive schemes.The shifting of customers between incentives schemes can be performed either interday or intraday (through moving time horizon approach) for real-time application.In the moving time horizon control scheme, consumers may switch between incentives at any time t ∈ T to select one of their own choice and aggregator evaluates the overall objective (including customer benefit as well as system benefit) and load reschedules for time period (t, T] can communicate to consumer.The consumer can either accept the load alterations by comparing the benefits of the previous schedule and the updated reschedules.The moving time control horizon may continue until the scheduling time horizon is reached to final time slot.

Demand response aggregation agent
The DRAA aggregates the RLAs to represent a large customer in the distribution network that offers various services to DSOA.In this paper, the utility and DRAA are interchangeably used considering the wholesale market and retail market operations of utility in the distribution system.The DRAA schedules the available load under responsive load penetration to maximise its operational profit.However, the same is constrained by distribution system performance.The profit maximisation depends on the total load scheduled, market clearing price of wholesale market, retail tariff structure, losses of the network, incentives given to RLAs.
The profit maximisation and customer electricity bill minimisation are often conflicting as LSE revenue is generated by bill payments of customer.However, customer adjusts its loads according to tariff in the first stage of DR scheduling and safeguards its own payments through selecting appropriate incentive range/scheme.If LSE wants to schedule DR loads with an agenda of maximising its profit, it has to pay the incentive to customers which acts as an additional cost/limiting factor to LSE and stabilises the DR profiles from being largely deviated from customer preferences.Thus, the incentive schemes considered in this paper can help customers in preserving their own benefits.In case of scenario 2, the relationship between load flattening objective and profit maximisation is not always conflicting.The conflicting nature comes into play under certain conditions like, if the difference (positive) between tariff and market price is high and load level is closer to average load (low variance), then increment in profit any further will deteriorate (increase) the load variance objective.In this paper, both the objectives have been unified as a single objective (minimisation) by negating the profit maximisation objective (8).Also, the expression of 'performance loss' as profit degeneration can also be used to unify the objective into a single minimisation objective.This paper examines two scenarios based on techno-economic performance distribution system.

Scenario 1: DRAA profit maximisation:
The profit maximisation objective of DRAA considering the interaction with wholesale market and retail market can be expressed as follows: The first part of equation equals the profit difference between retail market revenue and wholesale market participation cost.Whereas the second part comprises the incentives paid by DRAA to RLAs which represent the additional costs of DRAA in distribution system operation with RLAs.In ( 5), W t is the estimated/predicted market clearing price of wholesale market, t is the time step of operation (equal to 1 h in this paper), R t represents the retail tariff and this paper considers a differential tariff as follows: where R t is the differential tariff at hour t, R OFP t , R MP t and R ONP t , respectively, represents the off peak, mid peak and peak tariff rates over a day.

Scenario 2: DRAA profit maximisation and load levelling:
The supply-demand balance in distribution system is constrained by system performance aspects such as total system losses, voltage level and load profile.Therefore, besides profit maximisation, DRAA should also consider system performance aspects in order to not get penalised by DSOA for poor system performance.This paper considers load flattening aspect for accounting the distribution system performance.The objective includes the minimisation of variance in the load profile for each hour.The basic idea behind the objective is to flatten the load profile.The same will improve the peak (reduction) and valley (increase) conditions of load profile.The flattened load profile is known to have improved voltage profile and for the same amount of power, current flows will get reduced in the peak periods, which in turn reduce the losses of the network.Though, increment in current flows in valley period increase, the increment in losses (varying as a square of current flowing) at valley period is less compared with reduction in losses during peak periods of the day.Therefore, the overall impact of load flattening objective on total network losses over the day is positive (reduced losses).The overall scheduling objective of DRAA/LSE can be expressed as follows: where k t is the rescheduled load (after DSM program) of kth RLA for tth hour, mn . is the mean power consumption over the day, which can be expressed as follows: where base t is the base case power (without DSM) at hour t; k I , k II , k III , respectively, represents the base case loads of customers in group I, group II and group III.

Constraints
The operational objectives explained in above section are constrained by following limits, equality and inequality constraints.

Energy consumption constraint:
The overall energy consumption over the day for each RLA should be left unaltered by DSM programs as follows: where t, k req .represents the kth RLA's energy consumption/ requirement for tth hour.

Load shift limits of RLA:
The load shift limits depend on the maximum flexibility of the loads of RLAs.The deviation in base load during rescheduling is explained as follows: where ζ k is the flexibility limit of load for kth RLA.The solution procedure of the proposed scheduling methodology is presented in Algorithm 1 (see Fig. 4).

Results and discussion
The proposed DSM framework for CPSDS is examined using IEEE 37 bus test system with three types of loads categorised as shown in Fig. 5.The market price and the differential tariff of the system considered is represented in Fig. 6.The three areas/types of loads include residential, commercial and industrial customers, designated under groups I, II and III, respectively, with respect to the incentive payments during load alteration event of DSM program.The total aggregated load of the system over the day is 43939.23 kW.Case 1 represents the base test case without any DSM programs an overall system peak of 2773 kW occurring at 6 PM of the day.The peaks of residential, commercial and industrial loads are 841.8,640 and 1343.2 kW occurring at 7PM, 7 PM and 12 PM, respectively.The overall loss in supplying the total load over the scheduling period for case 1 is evaluated to be 740.20 kW.The additional cost of losses is appended to the utility, which has to be balanced by purchasing additional energy from wholesale energy market.The additional cost of losses in case 1 is observed to be 2581.31rupees (Rs.).The same brings down the profit of the IET Cyber-Phys.Syst., Theory Appl., 2017, Vol. 2 Iss. 2, pp.75-83 This is an open access article published by the IET under the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0/)utility/LSE from Rs. 79,269 to Rs. 76,688.Whereas, the electricity payments made by all the customers are evaluated to be Rs.234,377 (residential = Rs.61,020 + commercial = Rs.48,596 + industrial = Rs.124,759).The load factor of the system under no DSM programs is observed to be 0.65.
The second case (case 2) of this paper considers the dual objective problem, i.e. profit maximisation of the utility/ DRAA/LSE alongside the incentive minimisation paid to various types of customers.The network performance objectives such as load levelling/loss minimisation are not considered in this work.The incentives are paid to customers for both peak shaving and valley events.Therefore, customers in this scheme can avail dual benefit, i.e. during load reduction (during peak shaving) as well as load recovery (during valley filling).As a result, the profit (including the cost of system losses) of utility has increased by 5.28% (Rs.76,688 to Rs. 80,744) compared with case 1.The total electricity bill payments of the customers with and without incentives are observed to be Rs.231,368 and Rs.235,277, respectively.Therefore, the payments are reduced by 1.2% due to the incentives when compared to case 1.The peak loads of residential, commercial and industrial sectors are increased by 14.4, 15.1 and 6.9%, respectively, when compared with case 1.Consequently, the overall system peak in case 2 has been increased by 12.4% as compared to overall system peak load in case 1. Whereas, the total system losses in case 2 are increased by 3.4% (740-765 kW) as compared to case 1.Similarly, the load factor of the system for case 2 is also deteriorated by 11% (0.659-0.586) when compared with case 1.Despite the improvement in utility/LSE/DRAA profit and customer's electricity payments, the peak load and load factor of the system witnessed deterioration.
It can be observed from case 2 that, by mere consideration of economic perspectives of the system, the technical performance of the network may get worsened.Therefore, the deterioration of technical performance of network is considered alongside economic benefits of utility and customers.In case 3, the same is accompanied by using a multi-objective function modelled by utility/LSE/DRAA which includes profit maximisation, incentive minimisation and load factor minimisation objectives as explained in (7).The hourly load schedules for cases 1-3 are presented in Fig. 7a.The overall profit of the utility after deducting cost of network loss (Rs.2333) and incentives (Rs.3502) in case 3 is observed to be Rs.79,090.The same is 3.1% higher than the base case (case 1) with no DSM programs.The hourly variation of utility/LSE/DRAA profit across all the formats is presented in Fig. 7b.The scheduling objective of the LSE/utility/DRAA is to maximise the profit over the whole time of horizon, and negative profit instances may be observed when the market price is higher compared to tariff.The load schedule during the period can be reduced to maximise the profit.However, customer flexibility and utilisation constraints will not allow it to be completely zero, thus resulting in negative profits.Whereas, the customer payments in case 3 are reduced by 1.52% (Rs.234,377-Rs.230,810) when compared with case 1.However, the improvement in profit and electricity payments from cases 1-3 is understandably lower when compared with the improvement from cases 1-2 with case 2 favouring the economic incentives only.Compared with cases 1-2, the overall system losses in case 3 are reduced by 2.77 and 6%, respectively.The same can be attributed to the inclusion of system performance related objective in the overall objective.Consequently, the cost of loss for utility/LSE/DRAA in case 3 is The variation in other techno-economic aspects across different cases is presented in Figs. 9 and 10.The variation in hourly payments of residential, commercial and industrial loads is presented in Figs.9a-c, respectively.It can be observed that the difference in payments between various cases is minimum for industrial customers (Fig. 9c) compared with residential and commercial customers/loads.This can be attributed to the fact that industrial customers are distinguished as critical loads and are assigned to group III with highest incentive.Therefore, the load changes and incentives received by industrial customers are the lowest among all the customers, which in turn shows up in lower difference/impact between various DSM programs.
Similarly, the variation in minimum system voltage for the three cases is observed and is presented in Fig. 10.The range of variation for minimum system voltage is represented using range

Conclusion and future scope
A multi-objective approach based DSM framework for CPSDS is presented in this paper.The same considers the technical as well as economic performance of various stakeholders of smart distribution network.The DSM program considered in this paper uses the event trigger response based framework with consumer incentives.The overall objective is modelled to maximise the profit of utility/DRAA/LSE alongside customer incentive payment minimisation and load factor maximisation.The proposed methodology is implemented on IEEE 37 bus system considered to have residential, commercial and industrial loads/customers with various load preferences.The simulation results conclude the effectiveness of proposed multi-objective approach in terms of techno-economic performance.The improved economic performance is observed using reduction in electricity payments of customers and improved profit of utility/LSE/DRAA.Whereas, reduced network losses, improved load factor and voltage profile of the system demonstrate the improved technical performance of smart distribution system under proposed multi-objective approach.The future scope of the work may include the scheduling of other energy resources such as EVs alongside the DSM programs in CPSDS.The appliance operational characteristics and constraints can be used in future extensions of this work for modelling the dynamic characteristics of the loads and thus designing the incentive schemes.Apart from the day ahead scheduling, future works of this work may consider the dynamic real-time scheduling with moving horizon time control considering the customer shift between different incentive schemes available.

Acknowledgment
This study was funded in part by the DST Sponsored Project 'Control of grid interfaced solar photovoltaic energy with networked electric vehicles to enable vehicle a smart grid (project no: RP03200)'.

Fig. 3
Fig. 3 Customer grouping under incentive programs

Fig. 4 Fig. 6
Fig. 4 Algorithm I: energy management algorithm using Newton method

Fig. 8 Fig. 9
Fig. 8 Hourly variation in (a) Total system loss, (b) Cost of loss across three scheduling cases

( a )
Residential, (b) Commercial, (c) Industrial customer groups across three scheduling cases IET Cyber-Phys.Syst., Theory Appl., 2017, Vol. 2 Iss. 2, pp.75-83 This is an open access article published by the IET under the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0/)bands.In Fig. 10, V min represents the minimum system voltage, V min L and V minU are the lower and upper bounds of the minimum system voltage, respectively.It can be observed that, difference between lower and upper bounds for minimum system voltage (V case d, min ) is minimum for case 3, when compared with cases 1 and 2 (highest deviation in minimum system voltage).

Fig. 10
Fig. 10 Hourly variation in hourly minimum system voltage