A method for optimizing operation of a highway green energy system

Abstract

This invention relates to the field of energy system optimization technology, specifically a method for optimizing the operation of a green energy system for highways. To address the systemic challenge of handling multiple uncertainties in the source-load relationship, this method provides the following steps: "virtual-to-real" source-load prediction, "virtual-to-real" source-load prediction, multi-twin fusion decision-making for source-load prediction, "virtual-to-real" optimization control, "virtual-to-real" optimization control, and multi-twin fusion decision-making for optimization control. This method achieves optimized virtual-to-real operation of the system prediction, solving the problem of reduced prediction accuracy caused by a single prediction twin model's inability to adapt to environmental changes.

Description

Technical Field

[0001] This invention relates to the field of energy system optimization technology, specifically a method for optimizing the operation of a green energy system for highways. Background Technology

[0002] Optimizing the operation of green energy systems on highways is a key approach to promoting the integrated development of transportation and energy systems and achieving the "dual carbon" goal. However, the complex operation characteristics of multi-flow, wide-area highways make it difficult for traditional energy supply models to meet the needs of intelligent and green development. Simultaneously, the multiple uncertainties in renewable energy output and load demand within the highway network also pose challenges to the safe, stable, and efficient operation of the system. The impact of multiple uncertainties in renewable energy output-load demand (source-load) is mainly reflected in two core stages: first, the source-load forecasting stage, where distributed power generation such as wind and solar, and highway charging demand are affected by random factors such as weather and traffic flow, making it difficult for a single fixed model to achieve accurate forecasting; second, the disconnect between scheduling strategies and actual operating conditions caused by the transmission of forecast uncertainties, leading to risks such as inaccurate control and resource waste. Currently, handling multiple uncertainties in the source-load relationship remains one of the systemic challenges in optimizing the operation of green energy systems on highways. Summary of the Invention

[0003] In order to solve the systematic problem of handling multiple uncertainties in the source-load system, this invention provides a method for optimizing the operation of a green energy system for highways.

[0004] This invention is achieved using the following technical solution:

[0005] A method for optimizing the operation of a green energy system for highways includes the following steps:

[0006] Step 1: "Using virtual to predict real" source-load prediction, which involves generating multiple prediction twins based on the source-load deterministic prediction model. Each prediction twin performs a prediction task for a future time and outputs its own future deterministic prediction result. Then, the source-load uncertain prediction model is used to perform a prediction task for a future time and outputs the future uncertain prediction interval. The prediction accuracy of different prediction twins is judged based on the deviation between the future deterministic prediction result and the future uncertain prediction interval.

[0007] Step 2, “Real-source prediction using virtual methods”, involves inputting historical source-load operation data and using multiple source-load deterministic prediction twins to perform prediction tasks on the historical source-load operation data. The prediction accuracy of different source-load deterministic prediction twins is judged based on the deviation between the historical deterministic prediction results corresponding to the multiple source-load deterministic prediction twins and the historical actual values.

[0008] Step 3: Source-load prediction multi-twin fusion decision, that is, based on the prediction accuracy of the multi-twins according to the two prediction tasks of "predicting reality with virtual" and "performing reality with virtual", the optimal prediction twin is selected after fusion evaluation, and the deterministic prediction values ​​of renewable energy and load are obtained through the optimal prediction twin.

[0009] Step 4, "Optimization and control based on virtual reality", that is, generating multiple control twins according to different risk strategy combinations, each of the multiple control twins performs control tasks for the future and outputs its own optimized control results. Then, based on the source-load uncertainty prediction information, a source-load virtual scenario is generated, and the control results of the multiple control twins are used to perform virtual control on the virtual scenario and output the virtual control results.

[0010] Step 5: "Virtual simulation of real-world" optimization and control. Input historical system operation data, including historical prediction data and historical actual values. Based on the historical prediction data, use multiple control twins to perform control tasks and output their respective optimization and control schemes. Then, optimize and control the scenario under the historical actual values ​​according to the control schemes of each twin and output the historical control results.

[0011] Step 6: Optimize the decision-making of the multi-twin system for regulation. Based on the regulation results of the multi-twin system in the two regulation tasks of "predicting reality with virtual" and "performing reality with virtual", the optimal regulation twin is selected after fusion evaluation, and the twin is used to carry out the actual regulation task of the system.

[0012] Explanation of principles: The source-load prediction and optimization control in this application both adopt two methods: "virtual prediction of reality" and "virtual simulation of reality". The core idea of ​​"virtual prediction of reality" is to guide the selection of the current prediction and control twin by predicting the uncertainty of the "future". The core idea of ​​"virtual simulation of reality" is to guide the selection of the current prediction and control twin by using the "historical" operating data. By integrating the operating results of the twin in the "future" and "historical" times, the optimal control master twin can be selected for the current decision, thereby providing a more reliable basis for the accurate prediction and optimization control of the system.

[0013] Furthermore, the specific steps of step one are as follows:

[0014] 11. Select and collect data for deterministic prediction of influence sources and loads, and preprocess the data;

[0015] 12. Select the input features for source-load deterministic prediction from the data preprocessed in step 11, and construct multiple source-load deterministic prediction models using various methods (such as long-short-term memory network, LSTM, and convolutional neural network, CNN) to obtain multiple source-load deterministic prediction twins corresponding to the multiple source-load deterministic prediction models.

[0016] 13. Select the best input features for source-load uncertainty prediction from the data preprocessed in step 11, and construct a source-load uncertainty prediction model (e.g., a source-load uncertainty prediction model can be constructed using natural gradient boosting or NGBoost).

[0017] 14. Using the multiple source-load deterministic prediction twins obtained in step 12 and the source-load uncertainty prediction model constructed in step 13, the source-load output at future times is predicted, and the future deterministic prediction results corresponding to the multiple source-load deterministic prediction twins and the probability distribution function corresponding to the source-load uncertainty prediction model are obtained. The future uncertainty prediction interval is obtained based on the probability distribution function corresponding to the source-load uncertainty prediction model.

[0018] 15. Calculate the prediction accuracy of the future deterministic prediction result relative to the future uncertain prediction interval for each source-load deterministic prediction twin, thereby obtaining the uncertainty prediction accuracy of each source-load deterministic prediction twin.

[0019] In practice, the uncertainty prediction accuracy P of each source-load deterministic prediction twin is calculated using the PICP index. The calculation formula is as follows:

[0020] ;

[0021] Where N is the number of deterministic predictions for the future. This indicates whether the future deterministic prediction result corresponding to each source-load deterministic prediction twin is within the future uncertainty prediction interval. If the future deterministic prediction result corresponding to the source-load deterministic prediction twin is within the future uncertainty prediction interval, then... ,otherwise .

[0022] Furthermore, the specific steps of step two are as follows:

[0023] 21. Input the historical operation data of the source-load, and use the multiple source-load deterministic prediction twins obtained in step 12 to predict the historical operation data of the source-load, so as to obtain the historical deterministic prediction results corresponding to the multiple source-load deterministic prediction twins;

[0024] 22. Calculate the prediction accuracy of the historical deterministic prediction results relative to the historical true values ​​for each source-load deterministic prediction twin, and obtain the deterministic prediction accuracy of each source-load deterministic prediction twin;

[0025] Furthermore, the specific steps of step three are as follows:

[0026] 31. Construct a source-load deterministic prediction evaluation index system and a source-load uncertainty prediction evaluation index system based on the PICP index and RMSE index, respectively;

[0027] 32. Calculate the index values ​​of the multiple source-load deterministic prediction twins in the different evaluation index systems constructed in step 31;

[0028] 33. Normalize the index values ​​of each source-load deterministic prediction twin in the different evaluation index systems constructed in step 31 to obtain the basic probability assignment of different evaluation index evidence;

[0029] 34. The basic probability assignments of different evaluation index evidence for each source-load deterministic prediction twin are fused using the classical evidence combination formula. The final fused probability assignment for each source-load deterministic prediction twin is output, and the prediction twin with the highest fused probability assignment is selected as the prediction master twin. The classical combination formula is as follows:

[0030] ;

[0031] In the formula, This represents the final fusion probability assignment of the source-load deterministic prediction twins; Represents the source-load deterministic prediction twin of the i-th source-load. The basic probability assignment of the corresponding indicator evidence; This represents the j-th source-load deterministic prediction twin. The basic probability assignment of the corresponding indicator evidence; K is the conflict coefficient; where express and The corresponding twins when they are the same; φ represents and Different situations at the same time.

[0032] Furthermore, the specific steps of step four are as follows:

[0033] 41. For multiple microgrid systems in the highway green energy system, optimization objective functions, equipment mechanism models, and operational constraints are constructed respectively, resulting in the following optimization problem for each microgrid system:

[0034]

[0035] In the formula, and Let X1 be the deterministic prediction of the i-th renewable energy source and load obtained from the master twin, i = 1 - N, and N be the number of deterministic predictions of renewable energy source and load. Let h and g be the equality and inequality constraints, respectively. The objective function is to optimize the system based on deterministic predictions of the future.

[0036] 42. The system optimization objective function based on deterministic prediction values ​​in step 41 is optimized using an optimization algorithm to obtain the future optimization objective value f for each microgrid system. 10 ;

[0037] 43. Based on the optimization target value f0 obtained in step 42, set the target deviation coefficient and the weight of the uncertainty variable, and construct the IGDT optimization model under different risk strategies;

[0038] Based on the future optimization target value f obtained in step 42 10 We set the target deviation coefficient and the weights of uncertainty variables, and constructed IGDT optimization models under different risk strategies.

[0039] Among them, the constructed IGDT optimization models under different risk strategies are the IGDT models under risk aversion strategy and risk preference strategy, and their expressions are as follows:

[0040]

[0041]

[0042] In the formula, L is the deviation factor; and These are the step sizes for renewable energy and load, respectively; and All are target deviation coefficients, among which The avoidance coefficient; This is the preference coefficient; and These are the future real values ​​of the i-th renewable energy source and the load, respectively; Optimize the objective function for a system based on future true values; and These are the target thresholds in the IGDT model for risk aversion strategy and the IGDT model for risk preference strategy, respectively. The optimization objective of the IGDT model for risk aversion strategies is to maximize the bias factor. The optimization objective of the risk preference strategy IGDT model is to minimize the bias factor;

[0043] 44. The IGDT model constructed in step 43 is solved using a two-layer structure optimization. The lower layer uses an optimization algorithm (such as particle swarm optimization) to solve the system optimization objective function given the deviation factor L. The optimization objective function is consistent with the optimization objective function in step 41. The upper layer uses a bisection iteration method. By judging the relationship between the target value solved in the lower layer and the target value of IGDT, the possible interval of the deviation factor is continuously bisected. The error precision of the deviation factor is set. The iteration ends when the interval size of the deviation factor meets the error precision. The optimized control results and deviation factor of a single microgrid under independent operation are obtained, and the fluctuation range of the uncertainty variable is determined.

[0044] 45. For each microgrid system, different risk strategies are formulated. In step 43, multiple target deviation coefficients are formulated. Each target deviation coefficient represents a risk strategy. Then, different IGDT optimization models are constructed and solved according to the different target deviation coefficients to obtain the risk decision scheme for each microgrid system under different risk strategies.

[0045] 46. ​​Based on the risk decision schemes of each microgrid system under different risk strategies in step 45, construct a risk strategy set for the overall highway green energy system. In the risk strategy set, each risk strategy corresponds to a risk decision scheme for the overall microgrid system. Finally, generate multiple control twins based on the different risk decision schemes.

[0046] 47. Within the regulatory twin, multiple microgrid systems perform joint optimization scheduling under a risk strategy. The joint optimization scheduling uses the same optimization algorithm as the single microgrid optimization scheduling. Unlike the single microgrid optimization scheduling, the joint optimization scheduling incorporates energy trading between different microgrids. After optimization, the multi-microgrid optimization scheduling results of each regulatory twin are output.

[0047] 48. Based on the optimized control results of a single microgrid operating independently obtained in step 44 and the optimized control results of multiple microgrids obtained in step 47, determine the cooperative mode of the multi-microgrid system within the control twin.

[0048] 49. Based on the cooperative mode of the multi-micro-energy network system in the twin, and using the Shapley value method in cooperative game theory, calculate the proportion of marginal contribution of the members in the alliance to distribute the benefits, and finally output the twin's optimized control scheme.

[0049] 410. Based on the source-load uncertainty prediction model constructed in step 13, obtain the probability distribution function, and randomly generate virtual source-load scenarios using the Monte Carlo method or the Latin hypercube sampling method. Each control twin performs virtual control on the virtual source-load scenario according to the control scheme obtained in step 49, and outputs the virtual control results.

[0050] Furthermore, the specific steps of step five are as follows:

[0051] 51. For multiple microgrid systems in the highway green energy system, optimization objective functions, equipment mechanism models, and operational constraints are constructed respectively, resulting in the following optimization problem for each microgrid system:

[0052]

[0053] In the formula, and X1 and X2 are the deterministic predictions of the i-th renewable energy source and load obtained from the prediction master twin, respectively, based on historical data; X3 is the historical variable to be optimized; and h and g are the equality and inequality constraints, respectively. The objective function for system optimization is based on deterministic predictions of historical data.

[0054] 52. Optimize the system optimization objective function based on deterministic prediction values ​​in step 51 using an optimization algorithm (such as particle swarm optimization algorithm) to obtain the optimization objective value f0 for each microgrid system;

[0055] The optimization algorithm is used to optimize the system optimization objective function based on deterministic prediction values ​​in step 51, and the historical optimization objective value f for each microgrid system is obtained. 20 ;

[0056] 53. Based on the optimization target value f obtained in step 52 20 We set the target deviation coefficient and the weights of uncertainty variables, and constructed IGDT optimization models under different risk strategies.

[0057] Based on the historical optimization target value f obtained in step 52 20 We set the target deviation coefficient and the weights of uncertainty variables, and constructed IGDT optimization models under different risk strategies.

[0058] Among them, the constructed IGDT optimization models under different risk strategies are the IGDT models under risk aversion strategy and risk preference strategy, and their expressions are as follows:

[0059] ;

[0060] ;

[0061] In the formula and These are the historical true values ​​of the i-th renewable energy source and the load, respectively; Optimize the objective function for a system based on historical true values; and These are the target thresholds in the IGDT model for risk aversion strategy and the IGDT model for risk preference strategy, respectively.

[0062] 54. The IGDT model constructed in step 53 is solved using a two-layer structure optimization. The lower layer uses an optimization algorithm (such as particle swarm optimization) to solve the system optimization objective function given the deviation factor L. The optimization objective function is consistent with the optimization objective function in step 51. The upper layer uses a bisection iteration method. By judging the relationship between the target value solved in the lower layer and the target value of IGDT, the possible interval of the deviation factor is continuously bisected. The error precision of the deviation factor is set. The iteration ends when the interval size of the deviation factor meets the error precision. The optimized control results and deviation factor of a single microgrid under independent operation are obtained, and the fluctuation range of the uncertainty variable is determined.

[0063] 55. For each microgrid system, different risk strategies are formulated. In step 53, multiple target deviation coefficients are formulated. Each target deviation coefficient represents a risk strategy. Then, different IGDT optimization models are constructed and solved according to the different target deviation coefficients to obtain the risk decision scheme for each microgrid system under different risk strategies.

[0064] 56. Based on the risk decision schemes of each microgrid system under different risk strategies in step 55, construct a risk strategy set for the overall highway green energy system. In the risk strategy set, each risk strategy corresponds to a risk decision scheme for the overall microgrid system. Finally, generate multiple control twins based on the different risk decision schemes.

[0065] 57. Within the regulatory twin, multiple microgrid systems perform joint optimization scheduling under a risk strategy. The joint optimization scheduling uses the same optimization algorithm as the single microgrid optimization scheduling. Unlike the single microgrid optimization scheduling, the joint optimization scheduling incorporates energy trading between different microgrids. After optimization, the multi-microgrid optimization scheduling results of each regulatory twin are output.

[0066] 58. Based on the optimized control results of a single microgrid operating independently obtained in step 54 and the optimized control results of multiple microgrids obtained in step 57, determine the cooperative mode of the multi-microgrid system within the control twin.

[0067] 59. Based on the cooperative mode of the multi-micro-energy network system in the twin, and using the Shapley value method in cooperative game theory, calculate the proportion of marginal contribution of the members in the alliance to distribute the benefits, and finally output the twin's optimized control scheme.

[0068] 510. Input the historical source-load true values ​​of the system. Each control twin optimizes and controls the historical source-load scenario according to the control scheme obtained in step 59, and outputs the historical control results.

[0069] Furthermore, the specific steps of step six are as follows:

[0070] 61. Select multiple evaluation indicators (such as operating costs and energy utilization efficiency) to construct a comprehensive evaluation index system for highway green energy systems;

[0071] 62. Calculate the virtual control results of different control twins under different evaluation indicators in the two control tasks of "predicting reality with virtuality" and "performing reality with virtuality".

[0072] 63. Normalize the index values ​​of various regulatory twins under different evaluation indicators calculated in step 62 to obtain the evidence and basic probability assignments corresponding to different evaluation indicators. Each regulatory twin has its corresponding basic probability assignments for different indicator evidence.

[0073] 64. Calculate the similarity between different evaluation index evidences corresponding to each regulatory twin, and then calculate the credibility of each index evidence. The formulas for calculating similarity and credibility are shown below:

[0074]

[0075]

[0076] In the formula, Let m1 and m2 represent the similarity between them; m1 and m2 represent indicator evidence 1 and indicator evidence 2, respectively; n is the number of twins; N is the number of pieces of evidence. Let be the credibility of the i-th piece of evidence; Let be the credibility weight of the i-th piece of evidence;

[0077] 65. A game-theoretic-based combinatorial weighting method is adopted to calculate combinatorial weights, achieving a unification of subjective weights, objective weights, and credibility weights. Specifically, in subjective weighting, the analytic hierarchy process (AHP) is used to assign subjective weights to different evaluation indicator evidence. In objective weighting, credibility entropy is used to assign entropy weights to different evaluation indicator evidence; the lower the credibility entropy, the greater the indicator weight, and vice versa. The formula for calculating credibility entropy Ed is as follows:

[0078] ;

[0079] In the formula, The index values ​​are normalized values ​​for all twins;

[0080] 66. After calculating the combined weights according to step 65, calculate the reward / penalty factor based on the combined weights to distinguish the importance of the basic probability assignments of different twins under different indicators. The calculation formula for the reward / penalty factor is as follows:

[0081]

[0082] In the formula, Let i be the combined weight of the i-th piece of evidence; The average of the combined weights of all evidence; α is the reward / penalty coefficient; Assign a basic probability value to the j-th twin based on the i-th piece of evidence; for Corresponding reward and punishment factors;

[0083] 67. Based on the reward and punishment factors obtained in step 66, perform evidence fusion on the indicator evidence; the main framework of evidence fusion is as follows: fuse every two pieces of evidence once, then calculate the average of all fusion results to obtain the corrected evidence, and finally fuse the corrected evidence using the classic evidence combination formula, and output the fusion probability assignment corresponding to each regulatory twin.

[0084] When evidence and evidence During the fusion process, the trust obtained by assigning a basic probability value to each piece of evidence is first calculated using a unified trust allocation formula. Here, only m is used as an example. i (A1) and m j Taking the trust obtained by (A1) as an example, the trust obtained by the basic probability assignment of other twins is similar.

[0085] ;

[0086] ;

[0087] in, and They represent m respectively i (A1) and m j (A1) The trust gained, for Reward and punishment factors; for Reward and punishment factors; for Reward and punishment factors; for Reward and punishment factors;

[0088] get and Then, calculate and The basic probability assignment of twins after the fusion of two pieces of evidence is illustrated here, using twin A1 as an example. Its expression is

[0089]

[0090] Then, all fusion results corresponding to the twin are normalized to obtain the corrected basic probability assignment for twin A1. .

[0091]

[0092] The corrected basic probability assignments for other twins are obtained using the same method, resulting in corrected evidence after fusion of the two pieces of evidence. When the number of pieces of evidence is greater than or equal to 3, evidence is fused once for every two pieces of evidence (the order of fusion does not affect the fusion result). The basic probability assignments of all corrected evidence are added together, and the average value is taken to obtain the final corrected evidence, the expression of which is:

[0093]

[0094] The final revised evidence is fused N-1 times using the classic evidence combination formula in step 34 to obtain the final fusion result, which is the fusion probability assignment for each twin.

[0095] 68. After outputting the fusion probability assignment for each twin in step 67, analyze the twins corresponding to the fusion probability assignment. The twin with the highest fusion probability assignment is the final selected control master twin.

[0096] 69. The optimal control twin selected in step 68 is applied to the actual operation of the system to achieve optimized system operation.

[0097] Further, in step 22, the deterministic prediction accuracy R of each source-load deterministic prediction twin is calculated using the RMSE index, and the calculation formula is as follows:

[0098] ;

[0099] in, This represents the historical deterministic prediction results corresponding to each source-load deterministic prediction twin; represents the historical true value of the source-load; n represents the number of deterministic prediction results based on historical data.

[0100] The beneficial effects of this invention are as follows: (1) A source-load prediction method based on multi-twin fusion decision-making is proposed. By using the "virtual-to-real" approach, uncertainty prediction information is introduced before the actual operation of the system to guide the selection of the source-load prediction model of the system; by using the "virtual-to-real" approach, historical real operation data is introduced before the actual operation of the system to guide the selection of the source-load prediction model of the system. Based on the prediction accuracy of the multi-twin prediction in the two prediction tasks of "virtual-to-real" and "virtual-to-real", the optimal prediction twin is selected after fusion evaluation, and the twin is used to carry out the actual prediction task of the system. The virtual-to-real optimization operation of the system prediction is realized through the above methods to solve the problem that the single prediction twin model cannot adapt to environmental changes and other factors that lead to a decrease in model prediction accuracy.

[0101] (2) An optimization control method based on multi-twin fusion decision-making is proposed. Through the "virtual-to-real" approach, uncertainty prediction information is introduced before the actual operation of the system to guide the optimization control model of the selected system; through the "virtual-to-real" approach, historical real-time operating data is introduced before the actual operation of the system to guide the optimization control model of the selected system. A system optimization control method based on a comprehensive risk decision-making-cooperative game mechanism is proposed. While realizing the coordinated control of multiple microgrids, the system's "virtual-to-real" and "virtual-to-real" multi-twin fusion decision-making mechanisms are applied to realize the comprehensive risk decision-making process of the system, ultimately achieving optimized virtual-to-real operation of the system, thereby improving the optimization control performance of the highway green energy system under the influence of multiple uncertainties in the source-load relationship. Attached Figure Description

[0102] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0103] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0104] Figure 1 This is an overall optimized operation block diagram of the operation method described in this invention;

[0105] Figure 2 This is a block diagram of the source-load prediction method based on multi-twin fusion decision-making in this invention.

[0106] Figure 3 This is a block diagram of the optimization and control method based on multi-twin fusion decision-making in this invention. Detailed Implementation

[0107] To better understand the above-mentioned objectives, features, and advantages of the present invention, the solutions of the present invention will be further described below. It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other.

[0108] Many specific details are set forth in the following description in order to provide a full understanding of the invention, but the invention may also be practiced in other ways different from those described herein; obviously, the embodiments in the specification are only some embodiments of the invention, and not all embodiments.

[0109] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0110] like Figure 1 , 2 As shown in Figure 3, an optimized operation method for a green energy system on a highway includes the following steps:

[0111] Step 1: "Using virtual to predict real" source-load prediction, which involves generating multiple prediction twins based on the source-load deterministic prediction model. Each prediction twin performs a prediction task for a future time and outputs its own future deterministic prediction result. Then, the source-load uncertain prediction model is used to perform a prediction task for a future time and outputs the future uncertain prediction interval. The prediction accuracy of different prediction twins is judged based on the deviation between the future deterministic prediction result and the future uncertain prediction interval.

[0112] The specific steps are as follows:

[0113] 11. Select and collect data for deterministic prediction of influence sources and loads, and preprocess the data;

[0114] 12. Select the input features for source-load deterministic prediction from the data preprocessed in step 11, and construct multiple source-load deterministic prediction models using various methods (such as long-short-term memory network, LSTM, and convolutional neural network, CNN) to obtain multiple source-load deterministic prediction twins corresponding to the multiple source-load deterministic prediction models.

[0115] 13. Select the best input features for source-load uncertainty prediction from the data preprocessed in step 11, and construct a source-load uncertainty prediction model (e.g., a source-load uncertainty prediction model can be constructed using natural gradient boosting or NGBoost).

[0116] 14. Using the multiple source-load deterministic prediction twins obtained in step 12 and the source-load uncertainty prediction model constructed in step 13, the source-load output at future times is predicted, and the future deterministic prediction results corresponding to the multiple source-load deterministic prediction twins and the probability distribution function corresponding to the source-load uncertainty prediction model are obtained. The future uncertainty prediction interval is obtained based on the probability distribution function corresponding to the source-load uncertainty prediction model.

[0117] 15. Calculate the prediction accuracy of the future deterministic prediction result relative to the future uncertain prediction interval for each source-load deterministic prediction twin, thereby obtaining the uncertainty prediction accuracy of each source-load deterministic prediction twin.

[0118] In practice, the uncertainty prediction accuracy P of each source-load deterministic prediction twin is calculated using the PICP index. The calculation formula is as follows:

[0119] ;

[0120] Where N is the number of deterministic predictions for the future. This indicates whether the future deterministic prediction result corresponding to each source-load deterministic prediction twin is within the future uncertainty prediction interval. If the future deterministic prediction result corresponding to the source-load deterministic prediction twin is within the future uncertainty prediction interval, then... ,otherwise ;

[0121] Step 2: Source-load prediction using "virtual simulation of real data" involves inputting historical source-load operation data, using multiple source-load deterministic prediction twins to perform prediction tasks on the historical source-load operation data, and judging the prediction accuracy of different source-load deterministic prediction twins based on the deviation between the historical deterministic prediction results corresponding to multiple source-load deterministic prediction twins and the historical actual values.

[0122] The specific steps are as follows:

[0123] 21. Input the historical operation data of the source-load, and use the multiple source-load deterministic prediction twins obtained in step 12 to predict the historical operation data of the source-load, so as to obtain the historical deterministic prediction results corresponding to the multiple source-load deterministic prediction twins;

[0124] 22. Calculate the prediction accuracy of the historical deterministic prediction results relative to the historical true values ​​for each source-load deterministic prediction twin, and obtain the deterministic prediction accuracy of each source-load deterministic prediction twin;

[0125] In practice, the RMSE index is used to calculate the deterministic prediction accuracy R of each source-load deterministic prediction twin. The calculation formula is as follows:

[0126] ;

[0127] in, This represents the historical deterministic prediction results corresponding to each source-load deterministic prediction twin; This represents the historical true value of the source-load; n represents the number of deterministic prediction results based on historical data.

[0128] Step 3: Source-load prediction multi-twin fusion decision, that is, based on the prediction accuracy of the multi-twins according to the two prediction tasks of "predicting reality with virtual" and "performing reality with virtual", the optimal prediction twin is selected after fusion evaluation, and the deterministic prediction values ​​of renewable energy and load are obtained through the optimal prediction twin.

[0129] The specific steps are as follows:

[0130] 31. Construct a source-load deterministic prediction evaluation index system and a source-load uncertainty prediction evaluation index system based on the PICP index and RMSE index, respectively;

[0131] 32. Calculate the index values ​​of the multiple source-load deterministic prediction twins in the different evaluation index systems constructed in step 31;

[0132] 33. Normalize the index values ​​of each source-load deterministic prediction twin in the different evaluation index systems constructed in step 31 to obtain the basic probability assignment of different evaluation index evidence;

[0133] 34. The basic probability assignments of different evaluation index evidence for each source-load deterministic prediction twin are fused using the classical evidence combination formula. The final fused probability assignment for each source-load deterministic prediction twin is output, and the prediction twin with the highest fused probability assignment is selected as the prediction master twin. The classical combination formula is as follows:

[0134] ;

[0135] In the formula, This represents the final fusion probability assignment of the source-load deterministic prediction twins; Represents the source-load deterministic prediction twin of the i-th source-load. The basic probability assignment of the corresponding indicator evidence; This represents the j-th source-load deterministic prediction twin. The basic probability assignment of the corresponding indicator evidence; K is the conflict coefficient; where express and The corresponding twins when they are the same; φ represents and Different simultaneous situations;

[0136] Step 4, "Optimization and control based on virtual reality", that is, generating multiple control twins according to different risk strategy combinations, each of the multiple control twins performs control tasks for the future and outputs its own optimized control results. Then, based on the source-load uncertainty prediction information, a source-load virtual scenario is generated, and the control results of the multiple control twins are used to perform virtual control on the virtual scenario and output the virtual control results.

[0137] The specific steps are as follows:

[0138] 41. For multiple microgrid systems in the highway green energy system, optimization objective functions, equipment mechanism models, and operational constraints are constructed respectively, resulting in the following optimization problem for each microgrid system:

[0139]

[0140] In the formula, and Let X1 be the deterministic prediction of the i-th renewable energy source and load obtained from the master twin, i = 1 - N, and N be the number of deterministic predictions of renewable energy source and load. Let h and g be the equality and inequality constraints, respectively. The objective function is to optimize the system based on deterministic predictions of the future.

[0141] 42. The system optimization objective function based on deterministic prediction values ​​in step 41 is optimized using an optimization algorithm to obtain the historical optimization objective value f for each microgrid system. 10 ;

[0142] 43. Based on the future optimization target value f obtained in step 42 10 We set the target deviation coefficient and the weights of uncertainty variables, and constructed IGDT optimization models under different risk strategies.

[0143] Among them, the constructed IGDT optimization models under different risk strategies are the IGDT models under risk aversion strategy and risk preference strategy, and their expressions are as follows:

[0144]

[0145]

[0146] In the formula, L is the deviation factor; and These are the step sizes for renewable energy and load, respectively; and All are target deviation coefficients, among which The avoidance coefficient; This is the preference coefficient; and These are the future real values ​​of the i-th renewable energy source and the load, respectively; Optimize the objective function for a system based on future true values; and These are the target thresholds in the IGDT model for risk aversion strategy and the IGDT model for risk preference strategy, respectively. The optimization objective of the IGDT model for risk aversion strategies is to maximize the bias factor. The optimization objective of the risk preference strategy IGDT model is to minimize the bias factor;

[0147] 44. The IGDT model constructed in step 43 is solved using a two-layer structure optimization. The lower layer uses an optimization algorithm (such as particle swarm optimization) to solve the system optimization objective function given the deviation factor L. The optimization objective function is consistent with the optimization objective function in step 41. The upper layer uses a bisection iteration method. By judging the relationship between the target value solved in the lower layer and the target value of IGDT, the possible interval of the deviation factor is continuously bisected. The error precision of the deviation factor is set. The iteration ends when the interval size of the deviation factor meets the error precision. The optimized control results and deviation factor of a single microgrid under independent operation are obtained, and the fluctuation range of the uncertainty variable is determined.

[0148] 45. For each microgrid system, different risk strategies are formulated. In step 43, multiple target deviation coefficients are formulated. Each target deviation coefficient represents a risk strategy. Then, different IGDT optimization models are constructed and solved according to the different target deviation coefficients to obtain the risk decision scheme for each microgrid system under different risk strategies.

[0149] 46. ​​Based on the risk decision schemes of each microgrid system under different risk strategies in step 45, construct a risk strategy set for the overall highway green energy system. In the risk strategy set, each risk strategy corresponds to a risk decision scheme for the overall microgrid system. Finally, generate multiple control twins based on the different risk decision schemes.

[0150] 47. Within the regulatory twin, multiple microgrid systems perform joint optimization scheduling under a risk strategy. The joint optimization scheduling uses the same optimization algorithm as the single microgrid optimization scheduling. Unlike the single microgrid optimization scheduling, the joint optimization scheduling incorporates energy trading between different microgrids. After optimization, the multi-microgrid optimization scheduling results of each regulatory twin are output.

[0151] 48. Based on the optimized control results of a single microgrid operating independently obtained in step 44 and the optimized control results of multiple microgrids obtained in step 47, determine the cooperative mode of the multi-microgrid system within the control twin.

[0152] 49. Based on the cooperative mode of the multi-micro-energy network system in the twin, and using the Shapley value method in cooperative game theory, calculate the proportion of marginal contribution of the members in the alliance to distribute the benefits, and finally output the twin's optimized control scheme.

[0153] 410. Based on the source-load uncertainty prediction model constructed in step 13, obtain the probability distribution function, and randomly generate virtual source-load scenarios using the Monte Carlo method or the Latin hypercube sampling method. Each control twin performs virtual control on the virtual source-load scenario according to the control scheme obtained in step 49, and outputs the virtual control results.

[0154] Step 5: "Virtual simulation of real-world" optimization and control. Input historical system operation data, including historical prediction data and historical actual values. Based on the historical prediction data, use multiple control twins to perform control tasks and output their respective optimization and control schemes. Then, optimize and control the scenario under the historical actual values ​​according to the control schemes of each twin and output the historical control results.

[0155] The specific steps include the following:

[0156] 51. For multiple microgrid systems in the highway green energy system, optimization objective functions, equipment mechanism models, and operational constraints are constructed respectively, resulting in the following optimization problem for each microgrid system:

[0157]

[0158] In the formula, and X1 and X2 are the deterministic predictions of the i-th renewable energy source and load obtained from the prediction master twin, respectively, based on historical data; X3 is the historical variable to be optimized, and h and g are the equality and inequality constraints, respectively. The objective function for system optimization is based on deterministic predictions of historical data.

[0159] 52. Optimize the system optimization objective function based on deterministic prediction values ​​in step 51 using an optimization algorithm (such as particle swarm optimization) to obtain the historical optimization objective value f for each microgrid system. 20 ;

[0160] 53. Based on the historical optimization target value f obtained in step 52 20 We set the target deviation coefficient and the weights of uncertainty variables, and constructed IGDT optimization models under different risk strategies.

[0161] Among them, the constructed IGDT optimization models under different risk strategies are the IGDT models under risk aversion strategy and risk preference strategy, and their expressions are as follows:

[0162] ;

[0163] ;

[0164] In the formula and These are the historical true values ​​of the i-th renewable energy source and the load, respectively; Optimize the objective function for a system based on historical true values; and These are the target thresholds in the IGDT model for risk aversion strategy and the IGDT model for risk preference strategy, respectively.

[0165] 54. The IGDT model constructed in step 53 is solved using a two-layer structure optimization. The lower layer uses an optimization algorithm (such as particle swarm optimization) to solve the system optimization objective function given the deviation factor L. The optimization objective function is consistent with the optimization objective function in step 51. The upper layer uses a bisection iteration method. By judging the relationship between the target value solved in the lower layer and the target value of IGDT, the possible interval of the deviation factor is continuously bisected. The error precision of the deviation factor is set. The iteration ends when the interval size of the deviation factor meets the error precision. The optimized control results and deviation factor of a single microgrid under independent operation are obtained, and the fluctuation range of the uncertainty variable is determined.

[0166] 55. For each microgrid system, different risk strategies are formulated. In step 53, multiple target deviation coefficients are formulated. Each target deviation coefficient represents a risk strategy. Then, different IGDT optimization models are constructed and solved according to the different target deviation coefficients to obtain the risk decision scheme for each microgrid system under different risk strategies.

[0167] 56. Based on the risk decision schemes of each microgrid system under different risk strategies in step 55, construct a risk strategy set for the overall highway green energy system. In the risk strategy set, each risk strategy corresponds to a risk decision scheme for the overall microgrid system. Finally, generate multiple control twins based on the different risk decision schemes.

[0168] 57. Within the regulatory twin, multiple microgrid systems perform joint optimization scheduling under a risk strategy. The joint optimization scheduling uses the same optimization algorithm as the single microgrid optimization scheduling. Unlike the single microgrid optimization scheduling, the joint optimization scheduling incorporates energy trading between different microgrids. After optimization, the multi-microgrid optimization scheduling results of each regulatory twin are output.

[0169] 58. Based on the optimized control results of a single microgrid operating independently obtained in step 54 and the optimized control results of multiple microgrids obtained in step 57, determine the cooperative mode of the multi-microgrid system within the control twin.

[0170] 59. Based on the cooperative mode of the multi-micro-energy network system in the twin, and using the Shapley value method in cooperative game theory, calculate the proportion of marginal contribution of the members in the alliance to distribute the benefits, and finally output the twin's optimized control scheme.

[0171] 510. Input the historical source-load true values ​​of the system. Each control twin optimizes and controls the historical source-load scenario according to the control scheme obtained in step 59, and outputs the historical control results.

[0172] Step 6: Optimize the fusion decision of multiple twins for regulation. That is, based on the regulation results of multiple twins in the two regulation tasks of "predicting reality with virtual" and "performing reality with virtual", the optimal regulation twin is selected after fusion evaluation, and the twin is used to carry out the actual regulation task of the system.

[0173] The specific steps are as follows:

[0174] 61. Select multiple evaluation indicators (such as operating costs and energy utilization efficiency) to construct a comprehensive evaluation index system for highway green energy systems;

[0175] 62. Calculate the virtual control results of different control twins under different evaluation indicators in the two control tasks of "predicting reality with virtuality" and "performing reality with virtuality".

[0176] 63. Normalize the index values ​​of various regulatory twins under different evaluation indicators calculated in step 62 to obtain the evidence and basic probability assignments corresponding to different evaluation indicators. Each regulatory twin has its corresponding basic probability assignments for different indicator evidence.

[0177] 64. Calculate the similarity between different evaluation index evidences corresponding to each regulatory twin, and then calculate the credibility of each index evidence. The formulas for calculating similarity and credibility are shown below:

[0178]

[0179]

[0180] In the formula, Let m1 and m2 represent the similarity between them; m1 and m2 represent indicator evidence 1 and indicator evidence 2, respectively; n is the number of twins; N is the number of pieces of evidence. Let be the credibility of the i-th piece of evidence; Let be the credibility weight of the i-th piece of evidence;

[0181] 65. A game-theoretic-based combinatorial weighting method is adopted to calculate combinatorial weights, achieving a unification of subjective weights, objective weights, and credibility weights. Specifically, in subjective weighting, the analytic hierarchy process (AHP) is used to assign subjective weights to different evaluation indicator evidence. In objective weighting, credibility entropy is used to assign entropy weights to different evaluation indicator evidence; the lower the credibility entropy, the greater the indicator weight, and vice versa. The formula for calculating credibility entropy Ed is as follows:

[0182] ;

[0183] In the formula, The index values ​​are normalized values ​​for all twins;

[0184] 66. After calculating the combined weights according to step 65, calculate the reward / penalty factor based on the combined weights to distinguish the importance of the basic probability assignments of different twins under different indicators. The calculation formula for the reward / penalty factor is as follows:

[0185]

[0186] In the formula, Let i be the combined weight of the i-th piece of evidence; The average of the combined weights of all evidence; α is the reward / penalty coefficient; Assign a basic probability value to the j-th twin based on the i-th piece of evidence; for Corresponding reward and punishment factors;

[0187] 67. Based on the reward and punishment factors obtained in step 66, perform evidence fusion on the indicator evidence. The main framework of evidence fusion is as follows: fuse every two pieces of evidence once, then calculate the average of all fusion results to obtain the corrected evidence, and finally fuse the corrected evidence using the classic evidence combination formula to output the fusion probability assignment corresponding to each regulatory twin.

[0188] When evidence and evidence During the fusion process, the trust obtained by assigning a basic probability value to each piece of evidence is first calculated using a unified trust allocation formula. Here, only m is used as an example. i (A1) and m j Taking the trust obtained by (A1) as an example, the trust obtained by the basic probability assignment of other twins is similar.

[0189] ;

[0190] ;

[0191] in, and They represent m respectively i (A1) and m j (A1) The trust gained, for Reward and punishment factors; for Reward and punishment factors; for Reward and punishment factors; for Reward and punishment factors.

[0192] get and Then, calculate and The basic probability assignment of twins after the fusion of two pieces of evidence is illustrated here, using twin A1 as an example. Its expression is

[0193]

[0194] Then, all fusion results corresponding to the twin are normalized to obtain the corrected basic probability assignment for twin A1. .

[0195]

[0196] The corrected basic probability assignments for other twins are obtained using the same method, resulting in corrected evidence after fusion of the two pieces of evidence. When the number of pieces of evidence is greater than or equal to 3, evidence is fused once for every two pieces of evidence (the order of fusion does not affect the fusion result). The basic probability assignments of all corrected evidence are added together, and the average value is taken to obtain the final corrected evidence, the expression of which is:

[0197]

[0198] The final revised evidence is fused N-1 times using the classic evidence combination formula in step 34 to obtain the final fusion result, which is the fusion probability assignment for each twin.

[0199] 68. After outputting the fusion probability assignment for each twin in step 67, analyze the twins corresponding to the fusion probability assignment. The twin with the highest fusion probability assignment is the final selected control master twin.

[0200] 69. The optimal control twin selected in step 68 is applied to the actual operation of the system to achieve optimized system operation.

[0201] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the present invention. Although detailed descriptions have been provided with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments, and they should all be covered within the protection scope of the claims.

Claims

1. A method for optimizing the operation of a green energy system for highways, characterized in that, Step 1: "Using virtual to predict real" source-load prediction, which involves generating multiple prediction twins based on the source-load deterministic prediction model. Each prediction twin performs a prediction task for a future time and outputs its own future deterministic prediction result. Then, the source-load uncertain prediction model is used to perform a prediction task for a future time and outputs the future uncertain prediction interval. The prediction accuracy of different prediction twins is judged based on the deviation between the future deterministic prediction result and the future uncertain prediction interval. Step 2, "virtual simulation of real" source-load prediction, that is, inputting historical source-load operation data, using multiple source-load deterministic prediction twins to perform prediction tasks on the historical source-load operation data, and judging the prediction accuracy of different source-load deterministic prediction twins based on the deviation between the historical deterministic prediction results corresponding to multiple source-load deterministic prediction twins and the historical actual values. Step 3: Source-load prediction multi-twin fusion decision, that is, based on the prediction accuracy of the multi-twins according to the two prediction tasks of "predicting reality with virtual" and "performing reality with virtual", the optimal prediction twin is selected after fusion evaluation, and the deterministic prediction values ​​of renewable energy and load are obtained through the optimal prediction twin. Step 4, "Optimization and control based on virtual reality", that is, generating multiple control twins according to different risk strategy combinations, each control twin performs control tasks for the future and outputs its own optimized control results. Then, based on the source-load uncertainty prediction information, a source-load virtual scenario is generated. The control results of the multiple control twins are used to perform virtual control on the virtual scenario and output the virtual control results. Step 5: "Virtual simulation of reality" optimization and control. Input the system's historical operating data, including historical prediction data and historical actual values. Based on the historical prediction data, use multiple control twins to perform control tasks and output their respective optimization and control schemes. Then, optimize and control the scenario under the historical actual values ​​according to the control schemes of each twin and output the historical control results. Step 6: Optimize the decision-making of the multi-twin system for regulation. Based on the regulation results of the multi-twin system in the two regulation tasks of "predicting reality with virtual" and "performing reality with virtual", the optimal regulation twin is selected after fusion evaluation, and the twin is used to carry out the actual regulation task of the system.

2. The method for optimizing the operation of a green energy system for highways according to claim 1, characterized in that, The specific steps for Step One are as follows:

11. Select and collect data for deterministic prediction of influence sources and loads, and preprocess the data; 12. Select the optimal input features for source-load deterministic prediction from the preprocessed data in step 11, and construct multiple source-load deterministic prediction models using various methods to obtain multiple source-load deterministic prediction twins corresponding to the multiple source-load deterministic prediction models; 13. Select the best input features for source-load uncertainty prediction from the preprocessed data in step 11, and construct a source-load uncertainty prediction model; 14. Using the multiple source-load deterministic prediction twins obtained in step 12 and the source-load uncertainty prediction model constructed in step 13, the source-load output at future times is predicted, and the future deterministic prediction results corresponding to the multiple source-load deterministic prediction twins and the probability distribution function corresponding to the source-load uncertainty prediction model are obtained. The future uncertainty prediction interval is obtained based on the probability distribution function corresponding to the source-load uncertainty prediction model.

15. Calculate the prediction accuracy of the future deterministic prediction result relative to the future uncertain prediction interval for each source-load deterministic prediction twin, thereby obtaining the uncertainty prediction accuracy of each source-load deterministic prediction twin. The uncertainty prediction accuracy P of each source-load deterministic prediction twin is calculated using the PICP index, and the calculation formula is as follows: ； Where N is the number of deterministic predictions for the future. This indicates whether the future deterministic prediction result corresponding to each source-load deterministic prediction twin is within the future uncertainty prediction interval. If the future deterministic prediction result corresponding to the source-load deterministic prediction twin is within the future uncertainty prediction interval, then... ,otherwise .

3. The method for optimizing the operation of a green energy system for highways according to claim 2, characterized in that, The specific steps for step two are as follows:

21. Input the historical operation data of the source-load, and use the multiple source-load deterministic prediction twins obtained in step 12 to predict the historical operation data of the source-load, so as to obtain the historical deterministic prediction results corresponding to the multiple source-load deterministic prediction twins; 22. Calculate the prediction accuracy of the historical deterministic prediction results relative to the historical true values ​​for each source-load deterministic prediction twin, and obtain the deterministic prediction accuracy of each source-load deterministic prediction twin.

4. The method for optimizing the operation of a green energy system for highways according to claim 3, characterized in that, The specific steps for step three are as follows:

31. Construct a source-load deterministic prediction evaluation index system and a source-load uncertainty prediction evaluation index system based on the PICP index and RMSE index, respectively; 32. Calculate the index values ​​of the multiple source-load deterministic prediction twins in the different evaluation index systems constructed in step 31; 33. Normalize the index values ​​of each source-load deterministic prediction twin in the different evaluation index systems constructed in step 31 to obtain the basic probability assignment of different evaluation index evidence; 34. The basic probability assignments of different evaluation index evidence for each source-load deterministic prediction twin are fused using the classical evidence combination formula. The final fused probability assignment for each source-load deterministic prediction twin is output, and the prediction twin with the highest fused probability assignment is selected as the prediction master twin. The classical combination formula is as follows: ； In the formula, This represents the final fusion probability assignment of the source-load deterministic prediction twins; Represents the source-load deterministic prediction twin of the i-th source-load. The basic probability assignment of the corresponding indicator evidence; This represents the j-th source-load deterministic prediction twin. The basic probability assignment of the corresponding indicator evidence; K is the conflict coefficient; where express and The corresponding twins when they are the same; φ represents and Different situations at the same time.

5. The method for optimizing the operation of a green energy system for highways according to claim 4, characterized in that, The specific steps for step four are as follows:

41. For multiple microgrid systems in the highway green energy system, optimization objective functions, equipment mechanism models, and operational constraints are constructed respectively, resulting in the following optimization problem for each microgrid system: ； In the formula, and Let X1 be the deterministic prediction of the i-th renewable energy source and load obtained from the master twin, i = 1 - N, and N be the number of deterministic predictions of renewable energy source and load, respectively. Let h and g be the equality and inequality constraints, respectively. The objective function is to optimize the system based on deterministic predictions of the future.

42. The system optimization objective function based on deterministic prediction values ​​in step 41 is optimized using an optimization algorithm to obtain the future optimization objective value f for each microgrid system. 10 ; 43. Based on the future optimization target value f obtained in step 42 10 We set the target deviation coefficient and the weights of uncertainty variables, and constructed IGDT optimization models under different risk strategies. Among them, the constructed IGDT optimization models under different risk strategies are the IGDT models under risk aversion strategy and risk preference strategy, and their expressions are as follows: ； ； In the formula, L is the deviation factor; and These are the step sizes for renewable energy and load, respectively; and All are target deviation coefficients, among which The avoidance coefficient; This is the preference coefficient; and These are the future real values ​​of the i-th renewable energy source and the load, respectively; Optimize the objective function for a system based on future true values; and These are the target thresholds in the IGDT model for risk aversion strategy and the IGDT model for risk preference strategy, respectively. The optimization objective of the IGDT model for risk aversion strategies is to maximize the bias factor. The optimization objective of the risk preference strategy IGDT model is to minimize the deviation factor.

44. A two-layer structure is used to optimize and solve the IGDT model constructed in step 43. The lower layer uses an optimization algorithm (e.g., particle swarm optimization) to solve the system optimization objective function given the deviation factor L. The optimization objective function is consistent with the optimization objective function in step 41. The upper layer uses a bisection iteration method, which continuously bisections the possible intervals of the deviation factor by judging the relationship between the target value solved in the lower layer and the IGDT target value. An error precision is set for the deviation factor. The iteration ends when the interval size of the deviation factor meets the error precision, thus obtaining the optimized control results and deviation factor for a single microgrid operating independently, and determining the fluctuation range of the uncertainty variable.

45. For each microgrid system, different risk strategies are formulated. In step 43, multiple target deviation coefficients are formulated. Each target deviation coefficient represents a risk strategy. Then, different IGDT optimization models are constructed and solved according to the different target deviation coefficients to obtain the risk decision scheme for each microgrid system under different risk strategies.

46. ​​Based on the risk decision schemes of each microgrid system under different risk strategies in step 45, construct a risk strategy set for the overall highway green energy system. In the risk strategy set, each risk strategy corresponds to a risk decision scheme for the overall microgrid system. Finally, generate multiple control twins based on the different risk decision schemes.

47. Within the regulatory twin, multiple microgrid systems perform joint optimization scheduling under a risk strategy. The joint optimization scheduling uses the same optimization algorithm as the single microgrid optimization scheduling. Unlike the single microgrid optimization scheduling, the joint optimization scheduling incorporates energy trading between different microgrids. After optimization, the multi-microgrid optimization scheduling results of each regulatory twin are output.

48. Based on the optimized control results of a single microgrid operating independently obtained in step 44 and the optimized control results of multiple microgrids obtained in step 47, determine the cooperative mode of the multi-microgrid system within the control twin.

49. Based on the cooperative mode of the multi-micro-energy network system in the twin, and using the Shapley value method in cooperative game theory, calculate the proportion of marginal contribution of the members in the alliance to distribute the benefits, and finally output the twin's optimized control scheme.

410. Based on the source-load uncertainty prediction model constructed in step 13, obtain the probability distribution function, and randomly generate virtual source-load scenarios using the Monte Carlo method or the Latin hypercube sampling method. Each control twin performs virtual control on the virtual source-load scenario according to the control scheme obtained in step 49, and outputs the virtual control results.

6. The method for optimizing the operation of a green energy system for highways according to claim 5, characterized in that, The specific steps for step five are as follows:

51. For multiple microgrid systems in the highway green energy system, optimization objective functions, equipment mechanism models, and operational constraints are constructed respectively, resulting in the following optimization problem for each microgrid system: ； In the formula, and X1 represents the deterministic predictions of the i-th renewable energy source and load obtained from the prediction master twin, respectively, based on historical data; X2 represents the historical variables to be optimized, and h and g represent the equality and inequality constraints, respectively. The objective function for system optimization is based on deterministic predictions of historical data.

52. The system optimization objective function based on deterministic prediction values ​​in step 51 is optimized using an optimization algorithm to obtain the historical optimization objective value f for each microgrid system. 20 ; 53. Based on the historical optimization target value f obtained in step 52 20 We set the target deviation coefficient and the weights of uncertainty variables, and constructed IGDT optimization models under different risk strategies. Among them, the constructed IGDT optimization models under different risk strategies are the IGDT models under risk aversion strategy and risk preference strategy, and their expressions are as follows: ； ； In the formula and These are the historical true values ​​of the i-th renewable energy source and the load, respectively; Optimize the objective function for a system based on historical true values; and These are the target thresholds in the IGDT model for risk aversion strategy and the IGDT model for risk preference strategy, respectively.

54. The IGDT model constructed in step 53 is solved by using a two-layer structure optimization. The lower layer is to solve the system optimization objective function by using an optimization algorithm under the condition of determining the deviation factor L. The optimization objective function is consistent with the optimization objective function in step 51.

55. For each microgrid system, different risk strategies are formulated. In step 53, multiple target deviation coefficients are formulated. Each target deviation coefficient represents a risk strategy. Then, different IGDT optimization models are constructed and solved according to the different target deviation coefficients to obtain the risk decision scheme for each microgrid system under different risk strategies.

56. Based on the risk decision schemes of each microgrid system under different risk strategies in step 55, construct a risk strategy set for the overall highway green energy system. In the risk strategy set, each risk strategy corresponds to a risk decision scheme for the overall microgrid system. Finally, generate multiple control twins based on the different risk decision schemes.

57. Within the regulatory twin, multiple microgrid systems perform joint optimization scheduling under a risk strategy. The joint optimization scheduling uses the same optimization algorithm as the single microgrid optimization scheduling. Unlike the single microgrid optimization scheduling, the joint optimization scheduling incorporates energy trading between different microgrids. After optimization, the multi-microgrid optimization scheduling results of each regulatory twin are output.

58. Based on the optimized control results of a single microgrid operating independently obtained in step 54 and the optimized control results of multiple microgrids obtained in step 57, determine the cooperative mode of the multi-microgrid system within the control twin.

59. Based on the cooperative mode of the multi-micro-energy network system in the twin, and using the Shapley value method in cooperative game theory, calculate the proportion of marginal contribution of the members in the alliance to distribute the benefits, and finally output the twin's optimized control scheme.

510. Input the historical source-load true values ​​of the system. Each control twin optimizes and controls the historical source-load scenario according to the control scheme obtained in step 59, and outputs the historical control results.

7. The method for optimizing the operation of a green energy system for highways according to claim 6, characterized in that, The specific steps for step six are as follows:

61. Select multiple evaluation indicators to construct a comprehensive evaluation index system for green energy systems on highways; 62. Calculate the virtual control results of different control twins under different evaluation indicators in the two control tasks of "predicting reality with virtuality" and "performing reality with virtuality".

63. Normalize the index values ​​of various regulatory twins under different evaluation indicators calculated in step 62 to obtain the evidence and basic probability assignments corresponding to different evaluation indicators. Each regulatory twin has its corresponding basic probability assignments for different indicator evidence.

64. Calculate the similarity between different evaluation index evidences corresponding to each regulatory twin, and then calculate the credibility of each index evidence. The formulas for calculating similarity and credibility are shown below: ； ； In the formula, Let m1 and m2 represent the similarity between them; m1 and m2 represent indicator evidence 1 and indicator evidence 2, respectively; n is the number of twins; N is the number of pieces of evidence. Let be the credibility of the i-th piece of evidence; Let be the credibility weight of the i-th piece of evidence; 65. A game-theoretic-based combinatorial weighting method is adopted to calculate combinatorial weights, achieving a unification of subjective weights, objective weights, and credibility weights. Specifically, in subjective weighting, the analytic hierarchy process (AHP) is used to assign subjective weights to different evaluation indicator evidence. In objective weighting, credibility entropy is used to assign entropy weights to different evaluation indicator evidence; the lower the credibility entropy, the greater the indicator weight, and vice versa. The formula for calculating credibility entropy Ed is as follows: ； In the formula, The index values ​​are normalized values ​​for all twins; 66. After calculating the combined weights according to step 65, calculate the reward / penalty factor based on the combined weights to distinguish the importance of the basic probability assignments of different twins under different indicators. The calculation formula for the reward / penalty factor is as follows: ； In the formula, Let i be the combined weight of the i-th piece of evidence; The average of the combined weights of all evidence; α is the reward / penalty coefficient; Assign a basic probability value to the j-th twin based on the i-th piece of evidence; for Corresponding reward and punishment factors; 67. Based on the reward and punishment factors obtained in step 66, perform evidence fusion on the indicator evidence; the main framework of evidence fusion is as follows: fuse every two pieces of evidence once, then calculate the average of all fusion results to obtain the corrected evidence, and finally fuse the corrected evidence using the classic evidence combination formula, and output the fusion probability assignment corresponding to each regulatory twin.

68. After outputting the fusion probability assignment for each twin in step 67, analyze the twins corresponding to the fusion probability assignment. The twin with the highest fusion probability assignment is the final selected control master twin.

69. The optimal control twin selected in step 68 is applied to the actual operation of the system to achieve optimized system operation.

8. The method for optimizing the operation of a green energy system for highways according to claim 7, characterized in that, In step 22, the deterministic prediction accuracy R of each source-load deterministic prediction twin is calculated using the RMSE index. The calculation formula is as follows: ； in, This represents the historical deterministic prediction results corresponding to each source-load deterministic prediction twin; represents the historical true value of the source-load; n represents the number of deterministic prediction results based on historical data.

Images