An optimal scheduling method for park integrated energy system based on three-process coupling

Abstract

This invention discloses an optimized scheduling method for a park integrated energy system based on three-process coupling, belonging to the field of integrated energy system optimized scheduling technology. The method includes: a first process prediction, which inputs deterministic and uncertain prediction results of source loads, processes the uncertain prediction results, and then inputs the processed prediction results into a second process; a second process scheduling, which performs day-ahead and intraday optimized scheduling based on a two-layer collaborative mechanism, inputs the intraday scheduling results into a multi-twin system, and generates multiple twins for virtual operation according to different scheduling strategies; a third process evaluation, which uses a multi-time-period fusion evaluation method to evaluate the operating effects of multiple twins, selects the optimal twin's scheduling result, and executes it; after actual scheduling, the system operating results are evaluated again, achieving iterative operation of multiple twins through a combination of virtual and real evaluations.

Description

Technical Field

[0001] This invention belongs to the field of integrated energy system optimization scheduling technology, specifically involving an optimization scheduling method for a park integrated energy system based on three-process coupling. Background Technology

[0002] The integrated energy system of the industrial park achieves efficient energy utilization and reduces environmental pollution through the rational planning and scientific scheduling of various energy sources such as electricity, heat, and cooling, thereby improving the park's economic benefits while promoting its green and low-carbon development. However, the random fluctuations of wind and solar energy, along with the complex and diverse energy demands of the park, introduce multiple uncertainties into source load forecasting. Furthermore, the system's operating environment and the standards for understanding objective phenomena change over time; if the system model is not adjusted accordingly, model drift will occur. Currently, the "inadequate control" problem caused by multiple uncertainties in source load and model drift remains one of the technical bottlenecks in the management and control of integrated energy systems in my country and globally. Summary of the Invention

[0003] To address the problem of "inadequate control" caused by multiple uncertainties in source and load and model drift, this invention provides an optimized scheduling method for a park integrated energy system based on three-process coupling.

[0004] To achieve the above objectives, the present invention employs the following technical solutions:

[0005] An optimized scheduling method for a park integrated energy system based on three-process coupling includes:

[0006] The first process is source load prediction, which involves inputting the deterministic and uncertain prediction results of the source load, processing the uncertain prediction results of the source load, and then inputting the processed prediction results into the second process.

[0007] The second process is to optimize scheduling. Based on the two-layer collaborative mechanism, the day-ahead and intraday optimization scheduling is carried out. Then, the intraday scheduling results are input into the multi-twin system. Multiple twins are generated according to different scheduling strategies for virtual operation. The evaluation method of the third process is used to select the scheduling result of the optimal twin and execute it.

[0008] The third process is virtual-real evaluation, which uses a multi-time-segment fusion evaluation method to virtually evaluate the operational effects of multiple twins. Simultaneously, after the system is actually running, the system's operational results are evaluated again in real time, and the generation of multiple twins is adjusted based on the evaluation feedback.

[0009] Through the coupling effect of the three processes, the virtual and physical iterative operation of the park's integrated energy system is realized.

[0010] Furthermore, the source load uncertainty prediction results are processed in the first process, specifically by adopting the source load uncertainty processing method (1), which includes the following steps:

[0011] Step 1.1: Input the uncertainty prediction results of wind and solar power output and user load, that is, the probability density function f(x) of wind and solar power output and user load, and the probability density function follows a normal distribution;

[0012] Step 1.2: Discretize the probability density function of the source load prediction according to the standard deviation σ of the normal distribution, quantizing it into an interval form with a basic probability distribution. Generate three intervals around the mean of the normal distribution according to the magnitude of the standard deviation σ, and calculate the probability of each interval based on the probability density function. The calculation formula is as follows:

[0013]

[0014] In the formula, m(A) i ) is the i-th interval A i The probability of UB i It is the upper boundary of the i-th interval; LB i It is the lower boundary of the i-th interval; f(x) is the probability density function. At this time, m(A) i It is directly transformed into the basic probability distribution to participate in evidence fusion;

[0015] Step 1.3: After obtaining the basic probability distributions, the multiple basic probability distributions of each uncertain variable are fused according to the DS evidence synthesis rules to obtain the fused probability distributions of photovoltaic power output, wind power output, and user load, which serve as the basis for optimized scheduling. The classic evidence synthesis formula used is as follows:

[0016]

[0017] In the formula, m(B) is the basic probability assignment of proposition B; K is the conflict coefficient, which represents the magnitude of the conflict between two pieces of evidence.

[0018] Step 1.4: Calculate the information entropy of the fusion probability distribution of the source and payload. Determine the weights of the uncertain variables of the source and payload based on the magnitude of the information entropy; the larger the information entropy, the larger the weight. The formulas for calculating information entropy and weights are shown below:

[0019]

[0020]

[0021] In the formula, Ed i w is the information entropy of the i-th variable; i It is the weight of the i-th variable.

[0022] Furthermore, in the second process, the day-ahead and intraday optimization scheduling based on the two-layer collaborative mechanism adopts the day-ahead-intraday optimization scheduling method based on the two-layer collaborative mechanism (2), which specifically includes the following steps:

[0023] Step 2.1: Select the comprehensive energy system optimization index and construct the optimization objective function as follows:

[0024]

[0025] In the formula, h and g are the equality and inequality constraints, respectively, which are set according to the actual operating characteristics of the park; The objective function that needs to be optimized after inputting deterministic predicted values;

[0026] Step 2.2: Input the deterministic prediction results of the source load, use the CPLEX solver to optimize the objective function, and obtain the IGDT optimization baseline value f0;

[0027] Step 2.3: Set the target deviation coefficient and calculate the IGDT optimization target value;

[0028] Step 2.4: In the first process, calculate the weights w1, w2, w3 of the uncertainty variables based on the source load uncertainty prediction results;

[0029] Step 2.5, construct the IGDT model. Two optimization models are established based on risk aversion and risk preference strategies respectively. The objective function threshold set for the risk aversion strategy is higher than f0, while the objective function threshold set for the risk preference strategy is lower than f0. For the variable deviation coefficients, the deviation coefficients of the three uncertain variables—wind power generation, photovoltaic power generation, and user load—are weighted and summed as the optimization object, resulting in the following two strategy models:

[0030]

[0031]

[0032] In the formula, and , respectively, are the deviation coefficients of the uncertain variables corresponding to wind power generation, photovoltaic power generation, and user load; w1, w2, and w3 are respectively and The corresponding weights; ξ is the target deviation coefficient, which is set according to the actual operation of the park;

[0033] Step 2.6: Use the particle swarm optimization algorithm to optimize the IGDT model and obtain the optimized scheduling results and the source load fluctuation range;

[0034] Step 2.7: Analyze the operating characteristics of energy storage and flexible load, and combine the scheduling instructions of the upper-level IGDT model to obtain the operating constraints of battery energy storage and flexible load.

[0035] Step 2.8: While ensuring that the scheduling instructions for other equipment in the system remain unchanged, only the cooperative game between energy storage and flexible loads is conducted; a cooperative game model for energy storage loads is constructed; the objective functions of the two stakeholders are added together to obtain the overall optimization objective function for the lower level as follows:

[0036]

[0037] In the formula, f 储能 It is the optimization objective function for energy storage; f 负荷 It is the objective function for optimizing the load;

[0038] Step 2.9: Solve the above optimization objective function and adjust the upper-level scheduling instructions based on the optimization results.

[0039] Furthermore, in the second process, the intraday scheduling results are input into the multi-twin system, and multiple twins are generated for virtual operation according to different scheduling strategies. This adopts a real-time optimization scheduling method based on multi-twin iterative operation (3), which specifically includes the following steps:

[0040] Step 3.1: Based on the risk aversion strategy model and the risk preference strategy model, adjust the target deviation coefficient in each model according to the real-time fusion evaluation results to generate multiple scheduling schemes;

[0041] Step 3.2: Generate a master twin and N sibling twins based on multiple scheduling schemes, with each twin virtually running one scheduling scheme;

[0042] Step 3.3: Multiple twins perform rolling optimization within the predicted time domain of intraday optimized scheduling, and use the optimization results for virtual operation, and perform multi-time period virtual fusion evaluation on the virtual operation results;

[0043] Step 3.4: Select the optimal sibling twin based on the evaluation results, output and execute the scheduling instructions.

[0044] Furthermore, in the third process, the multi-time period fusion evaluation method is used to evaluate the operational performance of multiple twins. This method is based on the improved DS evidence theory and includes the following steps:

[0045] Step 4.1: Based on the actual structure and working methods of the park, construct an evaluation index system for the park's integrated energy system.

[0046] Step 4.2: After the system is running, collect indicator data, calculate the indicator values ​​according to the evaluation indicator calculation formula, and normalize them; there are three indicator values ​​q1, q2, and q3.

[0047] Step 4.3: Set the indicator weights w1, w2, and w3. These weights can be set according to the importance of the indicators. The three weights can be set to 0.35, 0.35, and 0.3 respectively. Calculate the overall indicator evaluation result of the system using the following formula:

[0048] E t = w1×q1 + w2×q2 + w3×q3

[0049] In the formula, E t It is the overall system index evaluation result for time period t;

[0050] Step 4.4: Set 5 evaluation levels: Poor, Poor, Average, Good, and Excellent; construct fuzzy membership functions and calculate the membership degree of different indicators to different evaluation levels as the evaluation value of that indicator; the formula for calculating the fuzzy membership function is as follows:

[0051]

[0052] In the formula, A(E) ij ) is the membership degree of the i-th indicator value to the j-th evaluation level; P j These are the parameters for the j-th evaluation level, where P1 = 0; P2 = 0.25; P3 = 0.5; P4 = 0.75; P5 = 1;

[0053] Step 4.5: After normalizing the membership degrees of the indicator values ​​for all evaluation levels, these are directly used as evidence in evidence fusion. An improved evidence synthesis formula is used to fuse evidence from multiple time periods, and the final fusion evaluation result is output. There are four time periods for fusion evaluation. The improved evidence synthesis steps are as follows: First, the similarity between all pairs of evidence is calculated using the following formula:

[0054]

[0055] In the formula, Sim is the similarity between m1 and m2; after calculating all similarities, the credibility of each piece of evidence is calculated using the following formula:

[0056]

[0057] In the formula, Cred i It represents the credibility of the i-th piece of evidence;

[0058] Then, the weight of each basic probability is calculated using the following formula:

[0059] wij =Cred i ×m i (A j )

[0060] In the formula, w ij It is the basic probability assignment m i (A j The weight of );

[0061] Finally, multiple pieces of evidence are weighted and fused, and the calculation formula is as follows:

[0062]

[0063] In the formula, m(A) j This refers to the fusion probability assignment. Based on the virtual evaluation results, the optimal sibling twin is selected, and scheduling instructions are output and executed. Simultaneously, the system's operational results are evaluated in real time, and the generation of multiple twins is adjusted based on the evaluation feedback.

[0064] Compared with the prior art, the present invention has the following advantages:

[0065] (1) A two-layer collaborative optimization scheduling mechanism is proposed. The upper layer is the IGDT optimization scheduling method, which achieves uncertainty optimization of the system under the consideration of schedulable resources and constraints. The lower layer is a multi-agent interactive optimization method, which forms a constraint balance between the interests of energy storage and flexible loads through cooperative game, and achieves joint optimization scheduling. The two-layer collaborative optimization scheduling mechanism solves the optimization scheduling problem considering uncertainty and multiple agents in two layers, providing a new research approach for solving multi-objective optimization problems affected by uncertainty.

[0066] (2) A virtual-real iterative optimization mechanism is proposed to solve the model drift problem of the scheduling model. This mechanism is realized through the coupling of three processes: prediction, scheduling, and evaluation. The scheduling phase is divided into day-ahead optimization and intraday optimization. The scheduling scheme is virtually tested during the real-time optimization operation phase. In the prediction phase, the uncertainty prediction results of source loads are processed. In the evaluation phase, virtual-real evaluation is combined. The virtual-real iterative optimization mechanism completes the automatic optimization of the system scheduling model in actual operation, solves the model drift problem of complex scheduling models, and provides a theoretical basis for achieving safe and efficient system operation. Attached Figure Description

[0067] The present invention will now be further described with reference to the accompanying drawings.

[0068] Figure 1 This is a schematic diagram of the components of the present invention.

[0069] Figure 2 This is a block diagram of the source load uncertainty handling method in this invention.

[0070] Figure 3 This is a block diagram of the day-to-day optimization scheduling method based on two-layer collaboration in this invention.

[0071] Figure 4 This is a block diagram of the real-time optimization scheduling method based on multi-twin iterative operation in this invention.

[0072] Figure 5 This is a flowchart of the multi-time period fusion evaluation method based on the improved DS evidence theory in this invention.

[0073] Figure 1 In the middle: 1 is a source-load uncertainty handling method; 2 is a day-ahead-intraday optimization scheduling method based on two-layer collaboration; 3 is a real-time optimization scheduling method based on multi-twin iterative operation; 4 is a multi-period fusion evaluation method based on improved DS evidence theory. Detailed implementation method:

[0074] like Figure 1 As shown, the present invention provides an optimized scheduling method for a park integrated energy system based on three-process coupling, including a source-load uncertainty handling method (1), a day-ahead-intraday optimized scheduling method based on two-layer collaboration (2), a real-time optimized scheduling method based on multi-twin iterative operation (3), and a multi-period fusion evaluation method based on improved DS evidence theory (4).

[0075] The source load uncertainty handling method (1) is as follows: Figure 2 As shown, the specific implementation method is as follows:

[0076] 1) Input the uncertainty prediction results of wind and solar power output and user load, i.e. the probability density function f(x) of wind and solar power output and user load, and the probability density function follows a normal distribution.

[0077] 2) Discretize the probability density function of the source load prediction according to the standard deviation σ of the normal distribution, quantizing it into an interval form with a basic probability distribution. Generate three intervals around the mean of the normal distribution according to the magnitude of the standard deviation σ, and calculate the probability of each interval based on the probability density function. The calculation formula is as follows:

[0078]

[0079] In the formula, m(A) i ) is the i-th interval A i The probability of UB i It is the upper boundary of the i-th interval; LB i It is the lower boundary of the i-th interval; f(x) is the probability density function. At this time, m(A) i It is directly transformed into the basic probability distribution to participate in evidence fusion.

[0080] 3) After obtaining the basic probability distributions, the multiple basic probability distributions of each uncertain variable are fused according to the DS evidence synthesis rules to obtain the fused probability distributions of photovoltaic power output, wind power output, and user load, which serve as the basis for optimized scheduling. The classic evidence synthesis formula used is as follows:

[0081]

[0082] In the formula, m(B) is the basic probability assignment of proposition B; K is the conflict coefficient, which represents the magnitude of the conflict between two pieces of evidence.

[0083] 4) Calculate the information entropy of the fusion probability distribution of the source and load variables. Determine the weights of the uncertain variables of the source and load variables based on the magnitude of the information entropy; the higher the information entropy, the greater the weight. The formulas for calculating information entropy and weights are shown below:

[0084]

[0085]

[0086] In the formula, Ed i w is the information entropy of the i-th variable. i It is the weight of the i-th variable.

[0087] The day-to-day optimization scheduling method (2) based on two-layer collaboration is as follows: Figure 3 As shown, the specific implementation method is as follows:

[0088] 1) Select the comprehensive energy system optimization index and construct the optimization objective function as follows:

[0089]

[0090] In the formula, h and g represent the equality and inequality constraints, respectively, which are set according to the actual operating characteristics of the park; f(X,P) W,t ,P S,t ,P L,t ) is the objective function that needs to be optimized after the deterministic predicted values ​​are input.

[0091] 2) Input the deterministic prediction results of the source load, use the CPLEX solver to optimize the objective function, and obtain the IGDT optimization baseline value f0.

[0092] 3) Set the target deviation coefficient and calculate the IGDT optimization target value.

[0093] 4) In the first process, calculate the weights w1, w2, w3 of the uncertainty variables based on the source load uncertainty prediction results.

[0094] 5) Constructing the IGDT model. Two optimization models are established based on risk aversion and risk preference strategies, respectively. The objective function threshold set for the risk aversion strategy is higher than f0, while the objective function threshold set for the risk preference strategy is lower than f0. For the variable deviation coefficients, the deviation coefficients of the three uncertain variables—wind power generation, photovoltaic power generation, and user load—are weighted and summed as the optimization object, resulting in the following two strategy models:

[0095]

[0096]

[0097] In the formula, and , respectively, are the deviation coefficients of the uncertain variables corresponding to wind power generation, photovoltaic power generation, and user load; w1, w2, and w3 are respectively and The corresponding weights; ξ is the target deviation coefficient, which is set according to the actual operation of the park.

[0098] 6) The particle swarm optimization algorithm is used to optimize the IGDT model to obtain the optimized scheduling results and the range of source load fluctuations.

[0099] 7) Analyze the operating characteristics of energy storage and flexible load, and combine the scheduling instructions of the upper-level IGDT model to obtain the operating constraints of battery energy storage and flexible load.

[0100] 8) While ensuring that the scheduling instructions for other equipment in the system remain unchanged, only the cooperative game between energy storage and flexible loads is conducted. An energy storage load cooperative game model is constructed. The objective functions of the two stakeholders are added together to obtain the overall optimization objective function at the lower level as follows:

[0101]

[0102] In the formula, f 储能 It is the optimization objective function for energy storage; f 负荷 It is the objective function for optimizing the load.

[0103] 9) Solve the above optimization objective function and adjust the upper-level scheduling instructions according to the optimization results.

[0104] The real-time optimization scheduling method (3) based on multi-twin iterative operation is as follows: Figure 4 As shown, the specific implementation method is as follows:

[0105] 1) Based on the risk aversion strategy model and the risk preference strategy model, the target deviation coefficient in each model is adjusted according to the real-time fusion evaluation results to generate multiple scheduling schemes.

[0106] 2) Generate a primary twin and N sibling twins based on multiple scheduling schemes. Each twin virtually runs one scheduling scheme.

[0107] 3) Multiple twins are used for rolling optimization within the predicted time domain of intraday optimized scheduling, and the optimization results are used for virtual operation. The virtual operation results are evaluated through multi-time period virtual fusion.

[0108] 4) Select the optimal sibling twin based on the evaluation results, output and execute the scheduling instructions.

[0109] The multi-time fusion evaluation method based on the improved DS evidence theory (4) is as follows: Figure 5 As shown, the specific implementation method is as follows:

[0110] 1) Based on the actual structure and working methods of the park, construct an evaluation index system for the park's integrated energy system.

[0111] 2) After the system is running, collect indicator data, calculate the indicator values ​​according to the evaluation indicator calculation formula, and normalize them. There are three indicator values: q1, q2, and q3.

[0112] 3) Set the indicator weights w1, w2, and w3. These weights can be set according to the importance of the indicators, and can be 0.35, 0.35, and 0.3 respectively. Calculate the overall indicator evaluation result of the system using the following formula:

[0113] E t = w1×q1 + w2×q2 + w3×q3

[0114] In the formula, E t It is the overall system index evaluation result for time period t.

[0115] 4) Set five evaluation levels: Poor, Poor, Average, Good, and Excellent. Construct a fuzzy membership function and calculate the membership degree of different indicators to different evaluation levels as the evaluation value of that indicator. The formula for calculating the fuzzy membership function is as follows:

[0116]

[0117] In the formula, A(E) ij ) is the membership degree of the i-th indicator value to the j-th evaluation level; P j These are the parameters for the j-th evaluation level, with P1 = 0; P2 = 0.25; P3 = 0.5; P4 = 0.75; and P5 = 1.

[0118] 5) After normalizing the membership degrees of the indicator values ​​for all evaluation levels, these are directly used as evidence in evidence fusion. An improved evidence synthesis formula is used to fuse evidence from multiple time periods, outputting the final fusion evaluation result. There are four time periods for fusion evaluation. The improved evidence synthesis steps are as follows: First, calculate the pairwise similarity between all pieces of evidence using the following formula:

[0119]

[0120] In the formula, Sim represents the similarity between m1 and m2. After calculating all similarities, the credibility of each piece of evidence is calculated using the following formula:

[0121]

[0122] In the formula, Cred i It represents the credibility of the i-th piece of evidence.

[0123] Then, the weight of each basic probability is calculated using the following formula:

[0124] w ij =Cred i ×m i (A j )

[0125] In the formula, w ij It is the basic probability assignment m i (A j The weight of ).

[0126] Finally, multiple pieces of evidence are weighted and fused, and the calculation formula is as follows:

[0127]

[0128] In the formula, m(A) j This represents the fusion probability assignment. Based on the evaluation results, the optimal sibling twin is selected, and the scheduling instructions are output and executed. Simultaneously, the system's operational results are evaluated in real time, and the generation of multiple twins is adjusted based on the feedback from the evaluation results.

[0129] In the process of three-process coupling, the second process IGDT model can be optimized not only by using swarm optimization algorithms, but also by using CPLEX solvers, etc.; the evaluation index system and fuzzy membership function of the third process multi-time fusion evaluation can be constructed according to different actual needs.

[0130] Contents not described in detail in this specification are prior art known to those skilled in the art. Although illustrative specific embodiments of the invention have been described above to facilitate understanding by those skilled in the art, it should be understood that the invention is not limited to the scope of the specific embodiments. Various modifications are readily apparent to those skilled in the art as long as they fall within the spirit and scope of the invention as defined and determined by the appended claims, and all inventions utilizing the concept of this invention are protected.

Claims

1. An optimized scheduling method for a park integrated energy system based on three-process coupling, characterized in that, include: The first process is source load prediction, which involves inputting the deterministic and uncertain prediction results of the source load, processing the uncertain prediction results of the source load, and then inputting the processed prediction results into the second process. The second process is to optimize scheduling. Based on the two-layer collaborative mechanism, the day-ahead and intraday optimization scheduling is carried out. Then, the intraday scheduling results are input into the multi-twin system. Multiple twins are generated according to different scheduling strategies for virtual operation. The evaluation method of the third process is used to select the scheduling result of the optimal twin and execute it. The third process is virtual-real evaluation, which uses a multi-time period fusion evaluation method to virtually evaluate the operational effects of multiple twins; at the same time, after the system is actually running, the system operation results are evaluated again in real time, and the generation of multiple twins is adjusted according to the evaluation feedback; Through the coupling of three processes, the virtual and real iterative operation of the park's integrated energy system is realized. In the first process, the uncertainty prediction results of source load are processed, specifically using a source load uncertainty processing method, which includes the following steps: Step 1.1: Input the uncertainty prediction results of wind and solar power output and user load, that is, the probability density function f(x) of wind and solar power output and user load, and the probability density function follows a normal distribution; Step 1.2: Discretize the probability density function of the source load prediction according to the standard deviation σ of the normal distribution, and quantize it into an interval form with a basic probability distribution; generate three intervals around the mean of the normal distribution according to the magnitude of the standard deviation σ, and calculate the probability of each interval based on the probability density function; the calculation formula is as follows: ； where m(A i ) is the probability of the ith interval A i ; UB i is the upper boundary of the ith interval; LB i is the lower boundary of the ith interval; and f(x) is the probability density function; at this time, m(A i ) is directly converted into the basic probability distribution to participate in evidence fusion; Step 1.3: After obtaining the basic probability distributions, the multiple basic probability distributions of each uncertain variable are fused according to the DS evidence synthesis rules to obtain the fused probability distributions of photovoltaic power output, wind power output, and user load, which serve as the basis for optimized scheduling. The classic evidence synthesis formula used is as follows: ； In the formula, m(B) is the basic probability assignment of proposition B; K is the conflict coefficient, which represents the magnitude of the conflict between two pieces of evidence. Step 1.4: Calculate the information entropy of the fusion probability distribution of the source and payload. Determine the weights of the uncertain variables of the source and payload based on the magnitude of the information entropy; the larger the information entropy, the larger the weight. The formulas for calculating information entropy and weights are shown below: ； ； In the formula, Ed i w is the information entropy of the i-th variable; i It is the weight of the i-th variable; In the second process, the day-ahead and intraday optimization scheduling based on the two-layer collaborative mechanism adopts a day-ahead-intraday optimization scheduling method based on two-layer collaboration, which specifically includes the following steps: Step 2.1: Select the comprehensive energy system optimization index and construct the optimization objective function as follows: ； In the formula, h and g are the equality and inequality constraints, respectively, which are set according to the actual operating characteristics of the park; The objective function that needs to be optimized after inputting deterministic predicted values; Step 2.2: Input the deterministic prediction results of the source load, use the CPLEX solver to optimize the objective function, and obtain the IGDT optimization baseline value f0; Step 2.3: Set the target deviation coefficient and calculate the IGDT optimization target value; Step 2.4: In the first process, calculate the weights w1, w2, w3 of the uncertainty variables based on the source load uncertainty prediction results. Step 2.5: Construct the IGDT model; establish two optimization models based on risk aversion and risk preference strategies respectively. The objective function threshold set for the risk aversion strategy is higher than f0, while the objective function threshold set for the risk preference strategy is lower than f0. For the variable deviation coefficients, the deviation coefficients of the three uncertain variables—wind power generation, photovoltaic power generation, and user load—are weighted and summed as the optimization object, resulting in the following two strategy models: ； ； In the formula, , and , respectively, are the deviation coefficients of the uncertain variables corresponding to wind power generation, photovoltaic power generation, and user load; w1, w2, and w3 are respectively , and The corresponding weights; The target deviation coefficient is set according to the actual operation of the park. Step 2.6: Use the particle swarm optimization algorithm to optimize the IGDT model and obtain the optimized scheduling results and the source load fluctuation range; Step 2.7: Analyze the operating characteristics of energy storage and flexible load, and combine the scheduling instructions of the upper-level IGDT model to obtain the operating constraints of battery energy storage and flexible load. Step 2.8: While ensuring that the scheduling instructions for other equipment in the system remain unchanged, only the cooperative game between energy storage and flexible loads is conducted; a cooperative game model for energy storage loads is constructed; the objective functions of the two stakeholders are added together to obtain the overall optimization objective function for the lower level as follows: ； In the formula, f 储能 It is the optimization objective function for energy storage; f 负荷 It is the objective function for optimizing the load; Step 2.9: Solve the above optimization objective function and adjust the upper-level scheduling instructions according to the optimization results; In the second process, the intraday scheduling results are input into the multi-twin system. Multiple twins are generated for virtual operation based on different scheduling strategies. This employs a real-time optimization scheduling method based on multi-twin iterative operation, specifically including the following steps: Step 3.1: Based on the risk aversion strategy model and the risk preference strategy model, adjust the target deviation coefficient in each model according to the real-time fusion evaluation results to generate multiple scheduling schemes; Step 3.2: Generate a primary twin and N sibling twins based on multiple scheduling schemes; each twin virtually runs one scheduling scheme. Step 3.3: Multiple twins perform rolling optimization within the predicted time domain of intraday optimized scheduling, and use the optimization results for virtual operation; perform multi-time period virtual fusion evaluation on the virtual operation results; Step 3.4: Select the optimal sibling twin based on the evaluation results, output and execute the scheduling instructions.

2. The optimized scheduling method for a park integrated energy system based on three-process coupling as described in claim 1, characterized in that: The evaluation of the operational performance of multiple twins in the third process using a multi-time period fusion evaluation method is based on an improved DS evidence theory. The specific steps include: Step 4.1: Based on the actual structure and working methods of the park, construct an evaluation index system for the park's integrated energy system; Step 4.2: After the system is running, collect indicator data, calculate the indicator values ​​according to the evaluation indicator calculation formula, and normalize them. There are three indicator values: q1, q2, and q3. Step 4.3: Set the indicator weights w1, w2, and w3 according to the different levels of importance of the indicators. The three weights are set to 0.35, 0.35, and 0.3 respectively. Calculate the overall indicator evaluation result of the system using the following formula: ； In the formula, E t It is the overall system index evaluation result for time period t; Step 4.4: Set 5 evaluation levels: Poor, Poor, Average, Good, and Excellent; construct fuzzy membership functions and calculate the membership degree of different indicators to different evaluation levels as the evaluation value of that indicator; the formula for calculating the fuzzy membership function is as follows: ； In the formula, A(E) ij ) is the membership degree of the i-th indicator value to the j-th evaluation level; P j These are the parameters for the j-th evaluation level, with P1=0; P2=0.25; P3=0.5; P4=0.75; P5=1; Step 4.5: After normalizing the membership degrees of the indicator values ​​for all evaluation levels, these are directly used as evidence in evidence fusion. An improved evidence synthesis formula is used to fuse evidence from multiple time periods, and the final fusion evaluation result is output. There are four time periods for fusion evaluation. The improved evidence synthesis steps are as follows: First, the similarity between all pairs of evidence is calculated using the following formula: ； In the formula, Sim is the similarity between m1 and m2; after calculating all similarities, the credibility of each piece of evidence is calculated using the following formula: ； In the formula, Cred i It represents the credibility of the i-th piece of evidence; Then, the weight of each basic probability is calculated using the following formula: ； In the formula, w ij It is the basic probability assignment m i (A j The weight of ); Finally, multiple pieces of evidence are weighted and fused, and the calculation formula is as follows: ； In the formula, m(A) j ) represents the fusion probability assignment; Based on the evaluation results, the optimal sibling twin is selected, and the scheduling instructions are output and executed. At the same time, the system operation results are evaluated in real time, and the generation of multiple twins is adjusted according to the feedback of the evaluation results.

Images