Bilevel Optimal Dispatch Strategy for a Multi-Energy System of Industrial Parks by Considering Integrated Demand Response

To combat energy shortage, the multi-energy system has gained increasing interest in contemporary society. In order to fully utilize adjustable multi-energy resources on the demand side and reduce interactive compensation, this paper presents an integrated demand response (IDR) model in consideration of conventional load-shedding and novel resource-shifting, due to the fact that participants in IDR can use more abundant resources to reduce the consumption of energy. In the proposed IDR, cooling, heating, electricity, gas and so forth are considered, which takes the connection between compensation and load reductions into consideration. Furthermore, a bilevel optimal dispatch strategy is proposed to decrease the difficulty in coordinated control and interaction between lower-level factories and upper-level multi-energy operators in industrial parks. In this strategy, resources in both multi-energy operator and user sides are optimally controlled and scheduled to maximize the benefits under peak shifting constraint. In the normal operation mode, this strategy can maximize the benefits to users and multi-energy operators. Particularly in heavy load conditions, compared to the conventional electricity demand response, there are more types of adjustable resources, more flexibility, and lower interactive compensations in IDR. The results indicate that optimal operation for factories and multi-energy operators can be achieved under peak shifting constraint and the overall peak power value in industrial park is reduced.


Introduction
With the increasing global energy crisis, the consensus is that low energy efficiency and high energy costs, due to separated planning and operation for cooling, heating, electricity, and gas systems, are the most important issues to be solved [1,2]. Due to its advantages in improving energy efficiency, reducing operational costs, and enhancing dispatching flexibility, the multi-energy system (MES) emerges as an attractive solution [3].
are few studies on multi-energy systems of industrial parks considering integrated demand response. Moreover, due to many interested parties and energy conversion devices in an industrial park, it is difficult to coordinate the multiple sources of energy. Therefore, a bilevel optimal dispatch strategy for a park-level multi-energy system considering integrated demand response is proposed in this paper.
The main contributions of this paper are as follows: I.
An integrated demand response model is built. In this model, the demand response for heating, cooling, and electricity is taken into consideration rather than single conventional electricity. There are more types of adjustable resources, more flexibility, and lower interactive compensations in the IDR program. II.
A bilevel optimal dispatch strategy is proposed to support the complex dispatch scheme and interaction of the industrial park. Resources in both multi-energy operator and factory sides are optimally controlled and scheduled with an economic objective under peak shifting constraints. The maximum interests of the lower-level factories and upper-level multi-energy operators can be ensured. A win-win situation for both multi-energy operator and factories can be created with this strategy. Meanwhile, computational difficulties and conflicts of interest can be eliminated. III. To evaluate the validity and practicality of the strategy proposed in this study, four cases are discussed. The results show that the maximum benefit of the lower-level factories and upper-level multi-energy operator can be ensured. In heavy load conditions, to handle emergencies in the power network, the most economical adjustable resources are chosen by the multi-energy operator to ensure the electricity balance. Moreover, the proposed model of integrated demand response and bilevel optimal dispatch strategy in this paper will be adopted by an actual multi-energy system demonstration project in China.
The rest of this paper is organized as follows: a device model of a multi-energy system is provided in Section 2. An integrated demand response model is established in Section 3. The bilevel optimal dispatch strategy model is proposed in Section 4. The case studies and discussion are described in Section 5. Finally, the main conclusions are summarized in Section 6.

Model of CCHP
In a CCHP unit, natural gas is consumed by GT to generate electricity, and natural gas is consumed by GB to generate heating. Waste heat is recovered by a heat recovery steam generator (HRSG) and absorption chiller (AC). CCHP is more able than a conventional thermal power plant to increase the energy efficiency and to cut costs [27]. CCHP can be classified into two types: (I) fixed heat to electricity ratio (back-pressure steam unit) (II) adjustable heat to electricity ratio (condensing steam type unit).
(i) The equivalent model of gas conversion: where F gas is the total input heat value of CCHP unit. F GT and F GB are the total input heat value of GT and gas boiler (GB), respectively. P GT and η GT represent the electric power generation and efficiency rate of GT, respectively. Q GB and η GB represent the thermal power generation and efficiency rate of GB, respectively.
(ii) The equivalent model of heat to electricity of CCHP The heat to electricity equivalent model of fixed heat to electricity ratio CCHP can be expressed as follows: where α is the fixed heat to electricity ratio. Q CCHP and P CCHP represent thermal and electricity power generation of CCHP, respectively. The heat to electricity equivalent model of adjustable heat to electricity ratio CCHP can be expressed as follows [28]: where Z is a fixed value. P con denotes the generated electricity in full condensing mode.
(iii) The equivalent model of waste heat recovery of GT (a) The equivalent model of HRSG Waste heat of GT can be reclaimed by HRSG to produce hot water and steam. The equivalent model of waste heat recovery of GT can be formulated as follows [29]: where Q out HRSG and Q in HRSG represent the output and input thermal power of HRSG, respectively. η HRSG denotes the efficiency rate of HRSG.
(b) The equivalent model of AC Waste heat of GT can be reclaimed by AC for refrigeration. The equivalent model of heat to cooling can be described as in Equation (7) [30]: where Q out AC and Q in AC represent the output refrigeration and input thermal power of AC, respectively. COP AC indicates the coefficient of performance of heat to cooling.

Model of Energy Storage
Energy storage (ES) is the key equipment in MES that can shift energy in the time dimension. ES is usually arranged to store energy during low tariff periods and discharge in high price hours to save on operational costs. With the advance of material technology, there are many types of ES on the market including battery storage (BS), thermal storage (TS), ice storage (IS), etc. [31]. There are various forms of energy storage, but the effect and constraints are similar. Generally, most energy storage devices can be expressed in the following model. The hourly remaining capacity in time t is calculated by Equation (8). The charge and discharge power of ES should not exceed its capacity limit, which can be described by Equations (9)- (11). Meanwhile, in the vast majority of case, ES cannot be charged and discharged at the same time t simultaneously, as expressed in Equation (12).
Energies 2018, 11, 1942 5 of 21 where W t S and W t+1 S are the level of ES in time t and t + 1, respectively. σ S is the loss ratio of ES. P S,ch and P S,dis are the storing power and releasing power of ES, respectively. Caps means the capacity of ES. γ S,ch and γ S,dis denote the maximal storing and releasing rate of ES. W S,min and W S,max are the minimal and maximal level of ES.

Model of Electric Refrigeration and Heating Device
Electric refrigeration and heating devices consume electricity to generate cooling and heating. The conversion model of electric refrigeration and heating device can be formulated as follows: Equations (13) and (14) denote the model of electric refrigeration device and electric heating device [29]. Q EC and P EC indicate the output refrigeration power and input electric power of electric refrigeration device. COP EC is the coefficient of performance of electricity to cooling. Similarly, Q EH and P EH indicate the output thermal power and input electric power of electric heating device. COP EH is the coefficient of performance of electricity to heating.

Model of Integrated Demand Response
Integrated demand response is established on the basis of conventional electricity demand response. The capability of energy complementation and integration of MES provide a basis for eliminating boundaries between electricity and other types of energy. In order to keep the energy balance at peak periods, not only conventional load-shedding but also novel resource-shifting should be involved in the IDR program [32]. IDR participants can use more abundant resources to reduce energy consumption. Conventional electricity demand response (EDR), heating demand response (HDR), cooling demand response (CDR), gas demand response, etc. are all included in the integrated demand response. P IDR = P EDR + P HDR + P CDR + P others (15) C IDR = C EDR + C HDR + C CDR + C others (16) Equations (15) and (16) present total load reduction and compensation of IDR, respectively.

Model of Electricity Demand Response
Conventional electricity demand response program is categorized into time-based and incentive-based programs [33]. Interruptible load (IL) management is usually regarded as a vital implementation of the incentive-based program in an industrial park. The multi-energy operator signs an IL contract with large industrial customers that will cut power use to obtain a certain amount of compensation from a multi-energy operator.
Power outage costs on the user side will be caused by electricity load shedding in IL. Therefore, it is necessary for a multi-energy operator to offer interactive users reasonable compensations. In fact, power outage costs and compensations are determined by a customer's load characteristics and increased with the amount of load shedding quadratically [34,35].
where C EDR,i is the compensation provided by multi-energy operator to interactive factory i. P EDR,i,t indicates the amount of load shedding of interactive factory i in time t. β i and µ i are the coefficients of factory i, which are related to a customer's load characteristics. T is the participation period.

Model of Heating and Cooling Demand Response
When factories' electric load increases substantially in an industrial park, the electric power drawn from an external power network is likely to exceed maximum allowable value of a tie-line. Under such circumstances, the factories need to be advised to obtain more heating and cooling from the multi-energy operator. Meanwhile, a factory's own electric refrigeration and heating devices are advised to be halted. While using less electricity, more electricity will be produced as a result of the higher cooling and heating load of CCHP. The equivalence relationship can be expressed as follows: Equations (18) and (19) indicate that the total power of load shedding in heating (or cooling) response in factory i are from two sources: (1) the electricity replaced and saved by heating (or cooling) demand response; (2) the increased electricity generation of CCHP. ∆P i is the total power of replaced and saved electricity by heating or cooling demand responses in factory i. ∆P CCHP is the power of increased electricity generation of CCHP. P i,j is the power of device j in factory i. Equations (21) and (22) denote the increased electricity generation of CCHP by heating and cooling demand responses, respectively. Equation (23) denotes the performance coefficient of electricity heating or cooling.
The extra cooling and heating resources produced by CCHP can be consumed by the factory for free with additional compensation from the operator to encourage the factory's participation. The total compensation obtained by participants in heating demand response program can be expressed as in Equation (24): where C HDR,i is the compensation obtained by factory i in heating demand response program; T is the participation period. c e is the current electricity price; λ i is the corresponding coefficient of factory i. Similarly, the total compensation obtained by participants in a cooling demand response program can be expressed by Equation (25): where C CDR,i is the compensation obtained by factory i in a cooling demand response program.

Bilevel Optimal Dispatch Framework in IDR
Profits are obtained by a multi-energy operator in the industrial park from supplying factories with cooling, heating, electricity, and ancillary services. The multi-energy operator often owns a Energies 2018, 11,1942 7 of 21 substantial number of energy conversion devices such as photovoltaic (PV), CCHP, BS, etc. When its electricity generation is unable to meet the factory's demand, a multi-energy operator can purchase electricity from an external utility company under the constraint of the maximal permitting power value of the tie line. Electricity, cooling, and heating are generated simultaneously by CCHP owned by a multi-energy operator. The electricity is transmitted to factories via distribution lines, and the cooling and heating can be supplied to users via a transportation pipe. Various forms of devices such as MT, PV, WT, BS, EAC, GB, and so forth may be installed on the factory side [36]. Insufficient energy power on the user side is supplied by the multi-energy operator. An overall schematic diagram and logic flow chart for the bilevel optimal dispatch strategy considering IDR are presented in Figures 1  and 2, respectively.
where CCDR,i is the compensation obtained by factory i in a cooling demand response program.

Bilevel Optimal Dispatch Framework in IDR
Profits are obtained by a multi-energy operator in the industrial park from supplying factories with cooling, heating, electricity, and ancillary services. The multi-energy operator often owns a substantial number of energy conversion devices such as photovoltaic (PV), CCHP, BS, etc. When its electricity generation is unable to meet the factory's demand, a multi-energy operator can purchase electricity from an external utility company under the constraint of the maximal permitting power value of the tie line. Electricity, cooling, and heating are generated simultaneously by CCHP owned by a multi-energy operator. The electricity is transmitted to factories via distribution lines, and the cooling and heating can be supplied to users via a transportation pipe. Various forms of devices such as MT, PV, WT, BS, EAC, GB, and so forth may be installed on the factory side [36]. Insufficient energy power on the user side is supplied by the multi-energy operator. An overall schematic diagram and logic flow chart for the bilevel optimal dispatch strategy considering IDR are presented in Figures 1  and 2, respectively.  Due to a large number of participants and potential conflicts of interest in park-level MES, it is difficult to coordinate and schedule multi-energy subsystems. As a result, the bilevel optimal dispatch strategy is necessary under such circumstances. Decision-makers at a lower level are factories' EMS and decision-makers at the operator level are a multi-energy operator's EMS [37]. Under normal operational conditions, the maximization of profit is set as a goal for both lower-level factories' EMS and multi-energy operator's EMS. In heavy load conditions, the multi-energy operator's EMS chooses the most economical multi-energy resources on its own side and controls the demand-side resources to keep the power balance and realize peak load shifting. Due to a large number of participants and potential conflicts of interest in park-level MES, it is difficult to coordinate and schedule multi-energy subsystems. As a result, the bilevel optimal dispatch strategy is necessary under such circumstances. Decision-makers at a lower level are factories' EMS and decision-makers at the operator level are a multi-energy operator's EMS [37]. Under normal operational conditions, the maximization of profit is set as a goal for both lower-level factories' EMS and multi-energy operator's EMS. In heavy load conditions, the multi-energy operator's EMS chooses the most economical multi-energy resources on its own side and controls the demand-side resources to keep the power balance and realize peak load shifting.

Distributed Dispatch Strategy for Lower-Level Factories
The objective for lower-level factories is to minimize operational costs by optimal dispatch of controllable devices according to the day-ahead load prediction and energy price. When the power of a tie line exceeds the maximum allowable value under peak shifting constraint, the IDR program will be started.

Objective Function
The objective for lower-level factories is to minimize daily operation costs. The objective function can be formulated in detail as follows: where f1 is the total operational costs. Celectricity, Cheating, Ccooling, and Cgas are the cost of purchasing electricity, heating, cooling, and gas, respectively.

Distributed Dispatch Strategy for Lower-Level Factories
The objective for lower-level factories is to minimize operational costs by optimal dispatch of controllable devices according to the day-ahead load prediction and energy price. When the power of a tie line exceeds the maximum allowable value under peak shifting constraint, the IDR program will be started.

Objective Function
The objective for lower-level factories is to minimize daily operation costs. The objective function can be formulated in detail as follows: where f 1 is the total operational costs. C electricity , C heating , C cooling , and C gas are the cost of purchasing electricity, heating, cooling, and gas, respectively.
where c t grid is the electricity price in time t. P t grid is the purchasing electricity power at time t. c heating is the price of heating. Q t heating is the purchasing heating power at time t. c cooling is the price of cooling. Q t cooling is the purchasing cooling power at time t. c gas is the heat value price of gas. P t MT and η MT are the power of electricity generation and efficiency ratio of a micro turbine. Q t GB and η GB are the thermal power generation and efficiency ratio of a gas boiler.

Constraints
(i) Electrical power balance P grid + P MT + P PV + P WT + P BS,dis + P IDR = P EL + P E AC + P HP + P ice DMME + P re f DMME + P BS,ch where P grid is the purchasing electricity power. P PV and P WT denote the power of electricity generation of photovoltaic and wind turbine, respectively. P BS,dis and P BS,ch indicate the discharging power and charging power of BS, respectively. P EAC and P HP present the input power of EAC and HP. P ice DMME and P re f DMME are the input power of double mode main engine in ice-making mode and refrigeration mode. P EL is the total power of the electric load.
(ii) Heat balance Q out HRSG + Q GB + Q TS,dis + Q HP = Q HL + ∆Q HL + Q TS,ch where Q out HRSG , Q GB , and Q HP are the output heat power of HRSG, GB, and HP, respectively. Q TS,D and Q TS,C represent the discharging power and charging power of TS, respectively. Q HL is the total power of the thermal load. (iii) Cooling balance Q cooling DMME + Q IS,dis + Q AC + Q EAC = Q CL + ∆Q CL + Q ice DMME (33) where Q cooling DMME , Q IS,dis , Q AC , and Q EAC are the output cooling power of double mode main engine in refrigeration mode, ice melting of IS, AC, and EAC, respectively. Q ice DMME is the cooling power of double mode main engine in ice-making mode. Q CL is the total power of the cooling load.

Centralized Dispatch Strategy of Multi-Energy Operator
The objective of the multi-energy operator is maximizing the profit under the premise of meeting a factory's energy demand in the industrial park. When the power of the tie line exceeds the maximum allowable value under peak shifting constraint, battery storage installed in multi-energy operator side will be utilized to smooth load fluctuation. If necessary, participating factories will be asked to join the IDR program to reduce the electricity load or increase the cooling or heating load from CCHP at the multi-energy operator side.

Objective Function
The objective for multi-energy operator is to maximize the profit. The objective function can be formulated in detail as follows: where f 2 is the total profit. E electricity , E heating and E cooling are the profit of selling electricity, heating, and cooling. C grid , C gas are the expense of purchasing electricity and gas from external electricity and a gas utility company. C IDR is the compensation cost for integrated demand response.

Constraints
(i) Maximum permitted power value of tie line under peak shifting constraint: where P line,max is the maximum permitted power value of a tie line under peak shifting constraint. (ii) Total power value of IDR Equation (36) shows the total over-limit power shortage under peak shifting constraint. Accordingly, Equation (37) shows the requirement of participation in IDR program to solve the problem of overload.

Problem-Solving Method
The models of upper-level and lower-level are binary mixed integer linear programming problems. The binary variable is introduced to handle coupling variables in the constraints; as an example, ES devices cannot be in both a charging and a discharging state at the same time [29]. The commercial computational software of linear interactive and general optimizer (Lingo) has been employed to solve this binary mixed integer linear programming problem.

Case Studies and Discussion
In this paper, case studies are conducted in an actual multi-energy system demonstration project of an industrial park in China. There are 13 lower-level factories and one multi-energy operator in the industrial park. The structure of the case study is established on the basis of Figure 1. Three major factories are selected for analysis, and the specific configuration and parameters of the factories and operator are shown in the Appendix A. Full details of time of use are shown in Table 1. According to the local rate, the price of gas is 0.5391 $/m 3 , which is equivalent to 0.0545 $/kW·h for heat value. The price of cooling and heating are 0.1250 $/kW·h and 0.1016 $/kW·h, respectively.

Distributed Optimal Dispatch of Lower-Level Factories
Using the distributed optimal dispatch strategy, Figures 3-6 show that three factories achieve the objective of minimal operational costs. The total operational costs of Factories 1 to 3 are $2169.50, $3582.60 and $592.30, respectively. An operational costs comparison for the three factories with and without optimization is presented in Table 2.

Distributed Optimal Dispatch of Lower-Level Factories
Using the distributed optimal dispatch strategy, Figures 3-6 show that three factories achieve the objective of minimal operational costs. The total operational costs of Factories 1 to 3 are $2169.50, $3582.60 and $592.30, respectively. An operational costs comparison for the three factories with and without optimization is presented in Table 2.             From the optimal dispatch results of the three factories, the following conclusions can be obtained: I. The battery storage in Factory 1 can cut operational costs by charging and discharging according to the electricity price. Meanwhile, the peak of overall load curve of the whole industrial park is lowered. A mutual beneficial result for both Factory 1 and the operator is achieved by the  From the optimal dispatch results of the three factories, the following conclusions can be obtained: I. The battery storage in Factory 1 can cut operational costs by charging and discharging according to the electricity price. Meanwhile, the peak of overall load curve of the whole industrial park is lowered. A mutual beneficial result for both Factory 1 and the operator is achieved by the optimal dispatch. II.
The cooling load for Factory 2 can be met by electric air conditioner (EAC) and ice storage. Ice storage is similar to battery storage. The ice is made, stored, and melted based on the electricity price. Dual-model chiller units are not allowed to operate in refrigeration mode and ice-making mode at the same time, so an electric air conditioner is used to provide cooling for factories when dual-mode chiller units operate in ice-making mode. III.
MT and GB work together to meet the thermal load of Factory 3. When the electricity price is high, MT works to produce both thermal energy and electricity. Any electrical power shortage may be compensated for by the grid. When the electrical price is low, GB is used to meet all thermal loads.

Centralized Optimal Dispatch of Multi-Energy Operator
The maximum permitted power value of a tie line in the industrial park under peak shifting constraint is set to 11 MW, so a multi-energy operator needs to sign IDR program contracts with factories in order to ensure the energy balance in the industrial park. Specific compensation tables based on the actual situation are shown in Tables 3 and 4. In normal mode, the system operation is illustrated in Figures 7-9 using the centralized optimal dispatch strategy. In this case, the energy balance can be kept by optimal and coordinated dispatch of CCHP and battery storage directly owned by a multi-energy operator. The total profit of multi-energy operator is $72,509.70 in Case 1.
In normal mode, the system operation is illustrated in Figures 7-9 using the centralized optimal dispatch strategy. In this case, the energy balance can be kept by optimal and coordinated dispatch of CCHP and battery storage directly owned by a multi-energy operator. The total profit of multienergy operator is $72,509.70 in Case 1.

Case 1. Optimal dispatch in normal operation
In normal mode, the system operation is illustrated in Figures 7-9 using the centralized optimal dispatch strategy. In this case, the energy balance can be kept by optimal and coordinated dispatch of CCHP and battery storage directly owned by a multi-energy operator. The total profit of multienergy operator is $72,509.70 in Case 1.   According to the day-ahead forecast, the total electricity load power will be increased by 3.5 MW at 14 p.m., and the cooling and thermal loads remain unchanged. The optimal dispatch of electric resources in Case 2 is shown in Figure 10. The optimal dispatch of cooling resources and thermal resources are as in Figures 8 and 9. The load fluctuation effect in the industrial park is mainly  According to the day-ahead forecast, the total electricity load power will be increased by 3.5 MW at 14 p.m., and the cooling and thermal loads remain unchanged. The optimal dispatch of electric resources in Case 2 is shown in Figure 10. The optimal dispatch of cooling resources and thermal resources are as in Figures 8 and 9. The load fluctuation effect in the industrial park is mainly alleviated by discharging of the battery directly owned by the multi-energy operator. The total profit of multi-energy operator is $72,228.40 in Case 2.

Case 2. The load fluctuation alleviated by battery storage
According to the day-ahead forecast, the total electricity load power will be increased by 3.5 MW at 14 p.m., and the cooling and thermal loads remain unchanged. The optimal dispatch of electric resources in Case 2 is shown in Figure 10. The optimal dispatch of cooling resources and thermal resources are as in Figures 8 and 9. The load fluctuation effect in the industrial park is mainly alleviated by discharging of the battery directly owned by the multi-energy operator. The total profit of multi-energy operator is $72,228.40 in Case 2. Case 3. The load fluctuation alleviated by heating demand response According to the day-ahead forecast, total electricity load power will be increased by 4.5 MW at 14:00, and the cooling and thermal loads remain unchanged. Over-limit power of a tie line under peak shifting constrain at 14:00 is 471 kW, as calculated by Equation (36). To maintain the energy balance, a multi-energy operator needs to turn off electric heating devices installed in IDR participants' factories and guide factories in using heating from the operator's CCHP. The total power of replaced electric heating is 125.6 kW; in addition, the total heating power increased by CCHP from the operator side is 502.3 kW. The total compensation expenses spent by a multi-energy operator on participating factories are $38.10, and the total profit of the multi-energy operator is $72,213 in this case. The optimal dispatch is shown in Figures 11 and 12. Case 3. The load fluctuation alleviated by heating demand response According to the day-ahead forecast, total electricity load power will be increased by 4.5 MW at 14:00, and the cooling and thermal loads remain unchanged. Over-limit power of a tie line under peak shifting constrain at 14:00 is 471 kW, as calculated by Equation (36). To maintain the energy balance, a multi-energy operator needs to turn off electric heating devices installed in IDR participants' factories and guide factories in using heating from the operator's CCHP. The total power of replaced electric heating is 125.6 kW; in addition, the total heating power increased by CCHP from the operator side is 502.3 kW. The total compensation expenses spent by a multi-energy operator on participating factories are $38.10, and the total profit of the multi-energy operator is $72,213 in this case. The optimal dispatch is shown in Figures 11 and 12. Energies 2018, 11, x FOR PEER REVIEW 15 of 21 Figure 11. Optimal dispatch of electric resources in Case 3. Figure 11. Optimal dispatch of electric resources in Case 3.

Case 4. The load fluctuation alleviated by electricity demand response
According to the day-ahead forecast, total electricity load power will be increased by 4.72 MW at 14:00, but the cooling and thermal load remain unchanged. Over-limit power of tie line under peak shifting constrain at 14:00 is 695 kW as calculated by Equation (36). To maintain the energy balance, the operator needs to turn off electric heating devices installed in IDR participants' factories and guide factories in using heating from an operator's CCHP. In addition, the electricity demand response participants come to the rescue by cutting 236.7 kW electrical power. The total electricity DR compensation is $472. The total compensation cost for an operator in a heating demand response program is $40.30. The total profit of the multi-energy operator is $71,766.90. The optimal dispatch is shown in Figures 13 and 14. The profit comparison for a multi-energy operator with and without optimization is presented in Table 5. According to the day-ahead forecast, total electricity load power will be increased by 4.72 MW at 14:00, but the cooling and thermal load remain unchanged. Over-limit power of tie line under peak shifting constrain at 14:00 is 695 kW as calculated by Equation (36). To maintain the energy balance, the operator needs to turn off electric heating devices installed in IDR participants' factories and guide factories in using heating from an operator's CCHP. In addition, the electricity demand response participants come to the rescue by cutting 236.7 kW electrical power. The total electricity DR compensation is $472. The total compensation cost for an operator in a heating demand response program is $40.30. The total profit of the multi-energy operator is $71,766.90. The optimal dispatch is shown in Figures 13 and 14. The profit comparison for a multi-energy operator with and without optimization is presented in Table 5.    From the optimal dispatch results in the four case studies, the following conclusions can be reached: The thermal load can be satisfied by fixed heat to electricity ratio CCHP. The CCHP operates in the following thermal load (FTL) mode and supplies electricity, heating and cooling simultaneously for users in an industrial park. The shortage of electricity can be compensated for by an external power network with a tie line. II. In normal operation, the battery storage directly owned by the operator stores cheap energy in flat and valley periods of electricity tariff and discharges when the electricity price is high to save on operational costs. Observing the maximum permitted power value of a tie line, the battery storage will discharge to eliminate the load peak when the power value of a tie line exceeds maximum the allowable value. III. Due to the high compensation costs, IDR participants are only required to adjust their energy usage when the power of the tie line exceeds the maximum allowable value. Due to the low impact on the comfort level and satisfaction ratio, the compensation cost of the heating demand  From the optimal dispatch results in the four case studies, the following conclusions can be reached: The thermal load can be satisfied by fixed heat to electricity ratio CCHP. The CCHP operates in the following thermal load (FTL) mode and supplies electricity, heating and cooling simultaneously for users in an industrial park. The shortage of electricity can be compensated for by an external power network with a tie line. II.
In normal operation, the battery storage directly owned by the operator stores cheap energy in flat and valley periods of electricity tariff and discharges when the electricity price is high to save on operational costs. Observing the maximum permitted power value of a tie line, the battery storage will discharge to eliminate the load peak when the power value of a tie line exceeds maximum the allowable value. III. Due to the high compensation costs, IDR participants are only required to adjust their energy usage when the power of the tie line exceeds the maximum allowable value. Due to the low impact on the comfort level and satisfaction ratio, the compensation cost of the heating demand response program is lower than the cost of the electricity demand response program. In practice, the heating demand response is an operator's favorite means of load shaving in an IDR program. IV. In an electricity demand response program, the compensation cost is in proportion to the square of the load shedding amount. To save on compensation costs, the total load shedding amount may be averaged for the three participating factories.

Results Analysis and Discussion
Without optimization, the operational plan of the device cannot be properly arranged. Operational costs are much higher in the three factories, while profits are lower for a multi-energy operator. The overall peak power value in an industrial park is large. The total operational costs of Factories 1 to 3 are $2498.60, $3637.40 and $637. 30 Simulation results have validated the effectiveness of the optimal dispatch considering IDR under peak shifting constraints. The optimal operation of factories and the multi-energy operator is achieved under peak shifting constraints and the overall peak power value in an industrial park is significantly reduced. In the normal operation mode, a device is reasonably operated according to the energy price. For instance, the energy storage on the factory side and multi-energy operator side is usually designed to store energy during low tariff periods and discharge in high price hours to maximize the benefits. The MT on the factory side works to produce both thermal energy and electricity in high price hours to minimize operational costs. With optimization, the total operational costs of Factories 1 to 3 are $2169.50, $3582.60, and $592.30, respectively. The total profit of the multi-energy operator in Cases 1 to 4 is $72,509.70, $72,228.40, $72,213 and $71,766.90, respectively. The power of the tie line in Cases 1 to 4 is 11 MW. Compared with the results without optimization, the factories and multi-energy operator gain more profits and the overall peak power value in an industrial park is reduced. Particularly in heavy load conditions, factories can consume more heating or cooling from CCHP to generate more electricity. Furthermore, the computing time is on the millisecond level, which can meet the engineering demand.
Although there are important discoveries revealed by these studies, there are also limitations. The current algorithm is relatively preliminary. The presented binary mixed integer linear programming problem is solved by Lingo. More attention will be given to intelligent algorithm analysis. Then a more accurate and comprehensive IDR model needs further study and exploration.

Conclusions
With the development of multi-energy systems, IDR has proven to be a new demand response form that can ensure MES' friendly interactions with the power network. In this paper, an integrated demand response model containing various types of flexible resources is established to fully utilize the adjustable multi-energy resources on the demand side and reduce the costs of compensation. Moreover, to decrease computational difficulty and conflict of interest in MES, a bilevel optimal dispatch strategy is proposed. The maximum profits of factories and multi-energy operators can be ensured via the bilevel optimal dispatch strategy. Four cases are analyzed to verify and validate the proposed strategy. The results show that using a distributed optimal dispatch strategy and multiple energy resources owned by lower-level factories can be coordinated to minimize operating costs. Similarly, the maximal profit of a multi-energy operator can be achieved and the overall peak power value in an industrial park can be reduced using a centralized optimal dispatch strategy. Particularly in heavy load conditions, the battery storage, heating demand response, and electricity demand response will be selected and implemented in turn to smooth the load fluctuation. A multi-energy operator has more choices at the peak time than ever, and a multi-energy operator is inclined to choose the most economical flexible resources in an industrial park. There are more types of flexible resources and lower interactive compensations in the IDR program. The steady-state dispatch of a multi-energy system based on the day-ahead prediction is mainly discussed in this paper. In the future, the intra-day dispatch strategy with an ultra-short-term load forecast and uncertainty of renewable energy resources and energy market will also be studied further.