Safety-constrained unit commitment optimization acceleration method and system based on large language model

Abstract

The present invention belongs to the technical field of power systems and relates to a method and system for accelerating the optimization of safety-constrained unit commitment based on a large language model. The method comprises: obtaining basic data on the economic operation of the power system; constructing and solving an integer-relaxed safety-constrained unit commitment model; generating code corresponding to a seed algorithm; evolving an optimal neighborhood search algorithm based on a neighborhood search algorithm of the large language model; constructing a safety-constrained unit commitment model within a limited neighborhood according to the optimal neighborhood search algorithm; solving the safety-constrained unit commitment model within a limited neighborhood; constructing a safety-constrained unit commitment model; and solving the safety-constrained unit commitment model. The present invention can generate high-quality neighborhood constraints and achieve efficient and feasible solution optimization; improve the efficiency of solving the safety-constrained unit commitment problem, enhance the efficiency of solving the power system optimization solution based on safety-constrained unit commitment, support the efficient optimization operation of a high-proportion power system, and ensure the safe and stable operation of the power system.

Description

Technical Field

[0001] The present invention belongs to the technical field of power systems, and in particular relates to a method and system for accelerating safety-constrained unit commitment optimization based on a large language model. Background Art

[0002] Safety-constrained unit commitment is the mathematical foundation for power system scheduling, market clearing, and system planning. It is typically modeled as a mixed-integer programming problem. With the increasing proportion of renewable energy in power systems and the expansion of interconnected large-scale power grids, power systems will experience an expansion in both temporal and spatial decision space. This will significantly increase the number of constraints and the degree of constraint coupling in unit commitment problems, posing significant challenges to solving them. Research is urgently needed on model dimensionality reduction and constraint reduction methods.

[0003] Data-driven and heuristic methods are commonly used to accelerate unit commitment solutions. The former accelerates solution by building models to learn from historical data for prediction, fixing integer variables or ineffective constraints for prediction. The latter reduces the problem size through specialized design of the problem structure, accelerating the solution of safety-constrained unit commitments.

[0004] Existing research on data-driven and heuristic methods for accelerating unit commitment solutions remains insufficient. Data-driven methods require large amounts of data for training, resulting in high computational costs for offline training. Furthermore, most models trained using data-driven methods are black-box models and lack interpretability. Heuristic methods, on the other hand, rely on historical experience and human expert design, and thus have limitations in various areas. Furthermore, most heuristic methods fail to achieve robust case-generalization acceleration, typically achieving efficient acceleration only for a subset of cases. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention aims to improve the computational efficiency of safety-constrained unit combination from the perspective of large language model-driven heuristic algorithm evolution, and provide a safety-constrained unit combination optimization acceleration method and system based on a large language model.

[0006] In a first aspect, the present invention provides a safety-constrained unit commitment optimization acceleration method based on a large language model, comprising generating a large language model-driven neighborhood search algorithm based on the large language model and solving the safety-constrained unit commitment;

[0007] Generate a large language model-driven neighborhood search algorithm based on the large language model, including: obtaining basic data on the economic operation of the power system; constructing and solving an integer relaxation safety constraint unit commitment model; generating corresponding code for the seed algorithm; and evolving the neighborhood search algorithm based on the large language model to obtain the optimal neighborhood search algorithm;

[0008] Solving the safety-constrained unit combination includes: constructing a safety-constrained unit combination model of a limited neighborhood according to an optimal neighborhood search algorithm; solving the safety-constrained unit combination model of the limited neighborhood; constructing the safety-constrained unit combination model; and solving the safety-constrained unit combination model.

[0009] In a second aspect, the present invention provides a safety-constrained unit commitment optimization acceleration system based on a large language model, comprising an algorithm generation unit and a solution unit;

[0010] An algorithm generation unit, which is used to generate a large language model-driven neighborhood search algorithm based on the large language model; the algorithm generation unit includes an acquisition unit, a model building and solving unit, a code generation unit, and an evolutionary update unit;

[0011] An acquisition unit, used to acquire basic data on economic operation of the power system;

[0012] Model building and solving unit, used to build and solve the integer relaxed safety constraint unit commitment model;

[0013] A code generation unit, used to generate code corresponding to the seed algorithm;

[0014] An evolutionary update unit, used to evolve the neighborhood search algorithm based on the large language model to obtain the optimal neighborhood search algorithm;

[0015] A solving unit, used for solving the safety constraint unit combination; the solving unit includes a first combination model building unit, a combination model solving unit, a second combination model building unit and a solving sub-unit;

[0016] The first combination model building unit is used for constructing a safety constraint unit combination model of a limited neighborhood using an optimal neighborhood search algorithm;

[0017] A combination model solving unit, used to solve the safety constraint unit combination model of a limited neighborhood;

[0018] The second combination model building unit is used to build a safety constraint unit combination model;

[0019] The solving subunit is used to solve the safety constrained unit combination model.

[0020] On the basis of the above technical solution, the present invention can also be improved as follows.

[0021] Furthermore, the basic data of the economic operation of the power system include the upper and lower limits of unit output, the upper limit of the unit increase and decrease ramp rate, the minimum continuous start and shutdown time of the unit, the unit operation cost function, the unit start and shutdown cost function, the power flow transfer distribution factor, the line transmission capacity and the system positive and negative reserve rate.

[0022] Furthermore, an integer relaxed safety constraint unit commitment model is constructed and solved, including:

[0023] Determine the decision variables of the integer relaxation safety constraint unit commitment, the decision variables including the output variables of the generator sets in a set period and the start and stop relaxation variables in the set period;

[0024] Determine the objective function corresponding to the integer relaxed safety constraint unit commitment model; the objective function is to minimize the total operating cost and start-up and shutdown cost in each period;

[0025] Establish constraints based on safety-constrained unit combinations; these constraints include system power balance constraints, system positive and negative standby constraints, system positive and negative ramp constraints, minimum continuous start-up and shutdown time constraints for generator sets, and line safety constraints;

[0026] A mixed integer linear programming solver is used to solve the objective function corresponding to the defined integer relaxed safety constraint unit commitment model, and the integer relaxed safety constraint unit commitment result is obtained.

[0027] Furthermore, the code corresponding to the seed algorithm is generated, including determining the seed neighborhood search algorithm and generating code text based on the seed neighborhood search algorithm:

[0028] A seed neighborhood search algorithm is determined, including: defining a neighborhood distance between the start-stop variables of the safety-constrained unit combination and the optimal solution of the integer relaxed safety-constrained unit combination, and obtaining a neighborhood constraint by limiting the neighborhood distance of the optimal solution to be no greater than the system reserve rate.

[0029] Furthermore, the algorithm evolution based on the large language model includes:

[0030] An evaluation function for evaluating the neighborhood search algorithm is constructed. The evaluation function is used to evaluate the performance of the neighborhood search algorithm. A set of safety-constrained unit commitment examples are set for evaluation. The results and time consumption of the safety-constrained unit commitment examples are solved using the neighborhood search algorithm and a mixed-integer linear programming solver. The results and time consumption of the solutions are compared with those of directly solving the safety-constrained unit commitment using the mixed-integer linear programming solver. The evaluation function is used to calculate the evaluation indicators of the neighborhood search algorithm.

[0031] Determine a code generation model and a prompt word generation model; the code generation model is used to generate code for the neighborhood search algorithm; the prompt word generation model is used to summarize the characteristics of the generated code corresponding to the neighborhood search algorithm and input them into the code generation model;

[0032] Generate neighborhood search algorithm based on evolutionary algorithm;

[0033] Based on the pre-set prompt words and seed algorithm codes, a code generation model is used to generate a batch of codes corresponding to the neighborhood search algorithms; the performance of the neighborhood search algorithms is evaluated based on the evaluation function of the neighborhood search algorithms, and all neighborhood search algorithms are ranked;

[0034] Based on the sorted neighborhood search algorithms, algorithm pairs are randomly extracted from them. Based on each algorithm pair and the pre-set prompt words, a large model is generated using the prompt words to process each algorithm pair. Based on the algorithm performance, algorithm improvement suggestions are generated to update the neighborhood search algorithm.

[0035] The performance of the generated neighborhood search algorithm is evaluated based on the evaluation function of the neighborhood search algorithm, and the neighborhood search algorithm is ranked;

[0036] The process of generating neighborhood search algorithms is repeated until the evaluation score of the neighborhood search algorithm remains unchanged after evaluating a set number of neighborhood search algorithms or the maximum number of generated algorithms is reached, and the optimal neighborhood search algorithm is obtained.

[0037] Furthermore, when constructing a safety-constrained unit commitment model with a limited neighborhood, the unit start-stop decision variables of the integer relaxation safety-constrained unit commitment model are adjusted to discrete variables, and the unit output decision variables remain continuous variables; neighborhood constraints are generated based on the optimal neighborhood search algorithm to obtain a safety-constrained unit commitment model with a limited neighborhood.

[0038] Furthermore, solving the safety constraint unit commitment model of the limited neighborhood includes:

[0039] The mixed integer linear programming solver is used to solve the safety-constrained unit commitment model in a limited neighborhood, and the results of the safety-constrained unit commitment model in a limited neighborhood are obtained.

[0040] Furthermore, a safety constraint unit combination model is constructed, including:

[0041] According to the safety-constrained unit commitment model with limited neighborhood, the neighborhood constraints are deleted to obtain the safety-constrained unit commitment model.

[0042] Furthermore, solving the safety-constrained unit commitment model includes: using the obtained result of the safety-constrained unit commitment model in a limited neighborhood as a hot start solution of a mixed integer linear programming solver, using the mixed integer linear programming solver to solve the safety-constrained unit commitment model, and obtaining a result of the safety-constrained unit commitment model.

[0043] The beneficial effects of the present invention are as follows: based on a large language model, the present invention utilizes a neighborhood search algorithm to realize directed evolution under the framework of an evolutionary algorithm, and generates a neighborhood search algorithm driven by a large language model, which can generate high-quality neighborhood constraints and realize efficient feasible solution optimization; without losing the accuracy of the solution, based on the high-quality feasible solution found, the efficiency of solving the safety-constrained unit combination problem is improved, the efficiency of optimizing the power system based on the safety-constrained unit combination is improved, the efficient optimization operation of the high-proportion power system is supported, and the safe and stable operation of the power system is guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 Schematic diagram of the safety-constrained unit commitment optimization acceleration method based on a large language model provided in Example 1 of the present invention;

[0045] Figure 2 This is a principle block diagram of the safety-constrained unit combination optimization acceleration system based on a large language model provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0046] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0047] Example 1

[0048] As an example, Figure 1 As shown, to solve the above technical problems, this embodiment provides a safety-constrained unit commitment optimization acceleration method based on a large language model, including generating a large language model-driven neighborhood search algorithm based on the large language model and solving the safety-constrained unit commitment;

[0049] Generate a large language model-driven neighborhood search algorithm based on the large language model, including: obtaining basic data on the economic operation of the power system; constructing and solving an integer relaxation safety constraint unit commitment model; generating corresponding code for the seed algorithm; and evolving the neighborhood search algorithm based on the large language model to obtain the optimal neighborhood search algorithm;

[0050] Solving the safety-constrained unit combination includes: constructing a safety-constrained unit combination model of a limited neighborhood according to an optimal neighborhood search algorithm; solving the safety-constrained unit combination model of the limited neighborhood; constructing the safety-constrained unit combination model; and solving the safety-constrained unit combination model.

[0051] Based on a large language model, this invention uses a neighborhood search algorithm within the framework of an evolutionary algorithm to implement directed evolution. This generates a large language model-driven neighborhood search algorithm that can generate high-quality neighborhood constraints and achieve efficient optimization of feasible solutions. Without sacrificing solution accuracy, the high-quality feasible solutions found improve the efficiency of solving safety-constrained unit commitment problems, enhance the efficiency of power system optimization solutions based on safety-constrained unit commitment, support the efficient optimization operation of high-proportion power systems, and ensure the safe and stable operation of power systems.

[0052] Optionally, the basic data for economic operation of the power system include upper and lower limits of unit output, upper limit of unit ramp-up and ramp-down rate, minimum continuous start-up and shutdown time of the unit, unit operation cost function, unit start-up and shutdown cost function, power flow transfer distribution factor, line transmission capacity and system positive and negative reserve rate.

[0053] Optionally, construct and solve an integer relaxed safety constraint unit commitment model, including:

[0054] Determine the decision variables of the integer relaxation safety constraint unit commitment, the decision variables including the output variables of the generator sets in a set period and the start and stop relaxation variables in the set period;

[0055] Determine the objective function corresponding to the integer relaxed safety constraint unit commitment model; the objective function is to minimize the total operating cost and start-up and shutdown cost in each period;

[0056] Establish constraints based on safety-constrained unit combinations; these constraints include system power balance constraints, system positive and negative standby constraints, system positive and negative ramp constraints, minimum continuous start-up and shutdown time constraints for generator sets, and line safety constraints;

[0057] A mixed integer linear programming solver is used to solve the objective function corresponding to the defined integer relaxed safety constraint unit commitment model, and the integer relaxed safety constraint unit commitment result is obtained.

[0058] generator sets In the period The output variable is , generator set In the period The start-stop slack variables are , The value is a continuous value between 0 and 1. In the period The start-stop slack variables are , For the crew The start-stop cost function, For the crew The output cost function is Defined as a generator set The initial start-stop state, is the total time, is the total number of units, and the expression of the objective function corresponding to the integer relaxation safety constraint unit commitment model is:

[0059] .

[0060] Establish the constraint conditions based on safety constraint unit combination, set For the period Internal system net load, is the system positive reserve rate, is the system negative reserve rate, For the period The maximum value of the internal system net load, For the period The maximum value of the internal system net load, For generator sets The output limit, For generator sets The lower limit of output, For generator sets Increase the output limit, For generator sets Reduce the output limit, For generator sets Output limit at startup, For generator sets Output limit before shutdown, For generator sets Minimum continuous power-on time, For generator sets Minimum continuous downtime, in hours, is the number of time periods for the unit combination, Indicates the Lines, For the line The biggest trend, For nodes Line The power flow transfer distribution factor, For the crew The node where For the crew The node where:

[0061] The system power balance constraint is expressed as: ;

[0062] The system positive reserve constraint is expressed as: ;

[0063] The system negative reserve constraint is expressed as: ;

[0064] The system positive ramp reserve constraint is expressed as: ;

[0065] The system negative ramp reserve constraint is expressed as: ;

[0066] The output range constraint of the generator set is expressed as: ;

[0067] The positive ramp constraint of the generator set is expressed as:

[0068] ;

[0069] The negative ramp constraint of the generator set is expressed as:

[0070] ;

[0071] The minimum continuous start-stop time constraint of the generator set is expressed as:

[0072] ;

[0073] The minimum continuous start-stop time constraint of the generator set is expressed as:

[0074] ;

[0075] The line safety constraint is expressed as:

[0076] .

[0077] Optionally, generate code corresponding to the seed algorithm, including determining the seed neighborhood search algorithm and generating code text based on the seed neighborhood search algorithm:

[0078] A seed neighborhood search algorithm is determined, including: defining a neighborhood distance between the start-stop variables of the safety-constrained unit combination and the optimal solution of the integer relaxed safety-constrained unit combination, and obtaining a neighborhood constraint by limiting the neighborhood distance of the optimal solution to be no greater than the system reserve rate.

[0079] An example of the algorithm is: define the safety constraint unit combination unit start and stop variables as , the optimal solution of unit commitment with integer relaxation safety constraints is , the neighborhood distance is , and then by limiting the distance to no greater than Get neighborhood constraints, is the weighting coefficient, which is used to describe the strength of the actual solution approaching the relaxed solution, as follows:

[0080] ;

[0081] ;

[0082] .

[0083] Optional, algorithmic evolution based on large language models, including:

[0084] An evaluation function for evaluating the neighborhood search algorithm is constructed. The evaluation function is used to evaluate the performance of the neighborhood search algorithm. A set of safety-constrained unit commitment examples for evaluation are set up, and the solutions are solved based on the neighborhood search algorithm and the mixed integer linear programming solver. The results and time consumption of the solutions are compared with the results and time consumption of directly using the mixed integer linear programming solver to solve the safety-constrained unit commitment. The evaluation index of the neighborhood search algorithm is calculated through the evaluation function. is the algorithm's indicator, is the number of examples used to evaluate the performance of the algorithm, is the weight of efficiency index and accuracy index, Solve the problem based on the neighborhood search algorithm or directly use the solver Time consuming, Solve the problem based on the neighborhood search algorithm or directly use the solver The lower the evaluation index, the better the algorithm performance.

[0085] ;

[0086] Determine a code generation model and a prompt word generation model; the code generation model is used to generate code for the neighborhood search algorithm; the prompt word generation model is used to summarize the characteristics of the generated code corresponding to the neighborhood search algorithm and input them into the code generation model;

[0087] Generate neighborhood search algorithm based on evolutionary algorithm;

[0088] Based on the pre-set prompt words and seed algorithm codes, a code generation model is used to generate a batch of codes corresponding to the neighborhood search algorithms; the performance of the neighborhood search algorithms is evaluated based on the evaluation function of the neighborhood search algorithms, and all neighborhood search algorithms are ranked;

[0089] Based on the sorted neighborhood search algorithms, algorithm pairs are randomly extracted from them. Based on each algorithm pair and the pre-set prompt words, a large model is generated using the prompt words to process each algorithm pair. The characteristics of the algorithms with better performance are compared and summarized, and algorithm improvement suggestions are generated to update the neighborhood search algorithm.

[0090] The performance of the generated neighborhood search algorithm is evaluated based on the evaluation function of the neighborhood search algorithm, and the neighborhood search algorithm is ranked;

[0091] The process of generating neighborhood search algorithms is repeated until the evaluation score of the neighborhood search algorithm remains unchanged after evaluating a set number of neighborhood search algorithms or the maximum number of generated algorithms is reached, and the optimal neighborhood search algorithm is obtained.

[0092] is the optimal unit output of the unit combination with integer relaxation safety constraints, is the optimal unit relaxed start-stop result of the integer relaxed safety constraint unit commitment, is the basic data of economic operation of power system, then:

[0093] .

[0094] Optionally, when constructing a safety-constrained unit combination model with a limited neighborhood, the unit start-stop decision variables of the integer relaxation safety-constrained unit combination model are adjusted to discrete variables, and the unit output decision variables remain continuous variables; neighborhood constraints are generated based on the optimal neighborhood search algorithm to obtain a safety-constrained unit combination model with a limited neighborhood.

[0095] The neighborhood constraint is expressed as: .

[0096] Optionally, solve the safety-constrained unit commitment model in a restricted neighborhood, including:

[0097] The mixed integer linear programming solver is used to solve the safety-constrained unit commitment model in a limited neighborhood, and the results of the safety-constrained unit commitment model in a limited neighborhood are obtained.

[0098] Optionally, a safety constraint unit combination model is constructed, including:

[0099] According to the safety-constrained unit commitment model with limited neighborhood, the neighborhood constraints are deleted to obtain the safety-constrained unit commitment model.

[0100] Optionally, solving the safety-constrained unit combination model includes: using the obtained result of the safety-constrained unit combination model in a limited neighborhood as a hot start solution of a mixed integer linear programming solver, using the mixed integer linear programming solver to solve the safety-constrained unit combination model, and obtaining the result of the safety-constrained unit combination model.

[0101] Based on a large language model, this invention uses a neighborhood search algorithm within the framework of an evolutionary algorithm to implement directed evolution. This generates a large language model-driven neighborhood search algorithm that can generate high-quality neighborhood constraints and achieve efficient feasible solution optimization. Without sacrificing solution accuracy, this method improves the efficiency of solving safety-constrained unit commitment problems and the efficiency of solving power system optimization solutions based on safety-constrained unit commitments, supporting the efficient optimization of high-ratio power systems and ensuring the safe and stable operation of power systems.

[0102] Example 2

[0103] Based on the same principle as the method shown in Example 1 of the present invention, as shown in the attached Figure 2 As shown, an embodiment of the present invention further provides a safety-constrained unit commitment optimization acceleration system based on a large language model, including an algorithm generation unit and a solution unit;

[0104] An algorithm generation unit, which is used to generate a large language model-driven neighborhood search algorithm based on the large language model; the algorithm generation unit includes an acquisition unit, a model building and solving unit, a code generation unit, and an evolutionary update unit;

[0105] An acquisition unit, used to acquire basic data on economic operation of the power system;

[0106] Model building and solving unit, used to build and solve the integer relaxed safety constraint unit commitment model;

[0107] A code generation unit, used to generate code corresponding to the seed algorithm;

[0108] An evolutionary update unit, used to evolve the neighborhood search algorithm based on the large language model to obtain the optimal neighborhood search algorithm;

[0109] A solving unit, used for solving the safety constraint unit combination; the solving unit includes a first combination model building unit, a combination model solving unit, a second combination model building unit and a solving sub-unit;

[0110] The first combination model building unit is used for constructing a safety constraint unit combination model of a limited neighborhood using an optimal neighborhood search algorithm;

[0111] A combination model solving unit, used to solve the safety constraint unit combination model of a limited neighborhood;

[0112] The second combination model building unit is used to build a safety constraint unit combination model;

[0113] The solving subunit is used to solve the safety constrained unit combination model.

[0114] Optionally, the basic data for economic operation of the power system include upper and lower limits of unit output, upper limit of unit ramp-up and ramp-down rate, minimum continuous start-up and shutdown time of the unit, unit operation cost function, unit start-up and shutdown cost function, power flow transfer distribution factor, line transmission capacity and system positive and negative reserve rate.

[0115] Optionally, construct and solve an integer relaxed safety constraint unit commitment model, including:

[0116] Determine the decision variables of the integer relaxation safety constraint unit commitment, the decision variables including the output variables of the generator sets in a set period and the start and stop relaxation variables in the set period;

[0117] Determine the objective function corresponding to the integer relaxed safety constraint unit commitment model; the objective function is to minimize the total operating cost and start-up and shutdown cost in each period;

[0118] Establish constraints based on safety-constrained unit combinations; these constraints include system power balance constraints, system positive and negative standby constraints, system positive and negative ramp constraints, minimum continuous start-up and shutdown time constraints for generator sets, and line safety constraints;

[0119] A mixed integer linear programming solver is used to solve the objective function corresponding to the defined integer relaxed safety constraint unit commitment model, and the integer relaxed safety constraint unit commitment result is obtained.

[0120] Optionally, generate code corresponding to the seed algorithm, including determining the seed neighborhood search algorithm and generating code text based on the seed neighborhood search algorithm:

[0121] A seed neighborhood search algorithm is determined, including: defining a neighborhood distance between the start-stop variables of the safety-constrained unit combination and the optimal solution of the integer relaxed safety-constrained unit combination, and obtaining a neighborhood constraint by limiting the neighborhood distance of the optimal solution to be no greater than the system reserve rate.

[0122] Optional, algorithmic evolution based on large language models, including:

[0123] An evaluation function for evaluating the neighborhood search algorithm is constructed. The evaluation function is used to evaluate the performance of the neighborhood search algorithm. A set of safety-constrained unit commitment examples are set for evaluation. The results and time consumption of the safety-constrained unit commitment examples are solved using the neighborhood search algorithm and a mixed-integer linear programming solver. The results and time consumption of the solutions are compared with those of directly solving the safety-constrained unit commitment using the mixed-integer linear programming solver. The evaluation function is used to calculate the evaluation indicators of the neighborhood search algorithm.

[0124] Determine a code generation model and a prompt word generation model; the code generation model is used to generate code for the neighborhood search algorithm; the prompt word generation model is used to summarize the characteristics of the generated code corresponding to the neighborhood search algorithm and input them into the code generation model;

[0125] Generate neighborhood search algorithm based on evolutionary algorithm;

[0126] Based on the pre-set prompt words and seed algorithm codes, a code generation model is used to generate a batch of codes corresponding to the neighborhood search algorithms; the performance of the neighborhood search algorithms is evaluated based on the evaluation function of the neighborhood search algorithms, and all neighborhood search algorithms are ranked;

[0127] Based on the sorted neighborhood search algorithms, algorithm pairs are randomly extracted from them. Based on each algorithm pair and the pre-set prompt words, a large model is generated using the prompt words to process each algorithm pair. The characteristics of the algorithms with better performance are compared and summarized, and algorithm improvement suggestions are generated to update the neighborhood search algorithm.

[0128] The performance of the generated neighborhood search algorithm is evaluated based on the evaluation function of the neighborhood search algorithm, and the neighborhood search algorithm is ranked;

[0129] The process of generating neighborhood search algorithms is repeated until the evaluation score of the neighborhood search algorithm remains unchanged after evaluating a set number of neighborhood search algorithms or the maximum number of generated algorithms is reached, and the optimal neighborhood search algorithm is obtained.

[0130] Optionally, when constructing a safety-constrained unit commitment model with a limited neighborhood, the unit start / stop decision variables of the integer relaxation safety-constrained unit commitment model are adjusted to discrete variables, while the unit output decision variables remain continuous variables; neighborhood constraints are generated based on an optimal neighborhood search algorithm to obtain a safety-constrained unit commitment model with a limited neighborhood. Optionally, solving the safety-constrained unit commitment model with a limited neighborhood includes:

[0131] The mixed integer linear programming solver is used to solve the safety-constrained unit commitment model in a limited neighborhood, and the results of the safety-constrained unit commitment model in a limited neighborhood are obtained.

[0132] Optionally, a safety constraint unit combination model is constructed, including:

[0133] According to the safety-constrained unit commitment model with limited neighborhood, the neighborhood constraints are deleted to obtain the safety-constrained unit commitment model.

[0134] Optionally, solving the safety-constrained unit combination model includes: using the obtained result of the safety-constrained unit combination model in a limited neighborhood as a hot start solution of a mixed integer linear programming solver, using the mixed integer linear programming solver to solve the safety-constrained unit combination model, and obtaining the result of the safety-constrained unit combination model.

[0135] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A safety-constrained unit commitment optimization acceleration method based on a large language model, characterized by: Including generating a large language model-driven neighborhood search algorithm based on a large language model and solving safety-constrained unit combinations; Generate a large language model-driven neighborhood search algorithm based on a large language model, including: obtaining basic data on economic operation of the power system; constructing and solving an integer relaxation safety constraint unit combination model; generating the corresponding code of the seed algorithm; evolving the neighborhood search algorithm based on the large language model to obtain the optimal neighborhood search algorithm; constructing and solving an integer relaxation safety constraint unit combination model, including: determining the decision variables of the integer relaxation safety constraint unit combination, the decision variables include the output variables of the generator set in a set time period and the start and stop relaxation variables in the set time period; determining the objective function corresponding to the integer relaxation safety constraint unit combination model; the objective function is to minimize the total operating cost and start and stop cost of each time period; establishing constraint conditions based on the safety constraint unit combination; the constraint conditions include system power balance constraint, system positive and negative standby constraint, system positive and negative ramp constraint, minimum continuous start and stop time constraint of the generator set and line safety constraint; using a mixed integer linear programming solver to solve the objective function corresponding to the defined integer relaxation safety constraint unit combination model to obtain the integer relaxation safety constraint unit combination result; algorithm evolution based on the large language model, including: constructing an evaluation function for evaluating the neighborhood search algorithm, The evaluation function is used to evaluate the performance of the neighborhood search algorithm; a set of safety-constrained unit combination examples for evaluation are set, which are solved based on the neighborhood search algorithm and the mixed integer linear programming solver, and the results and time of the solutions are compared with the results and time of directly using the mixed integer linear programming solver to solve the safety-constrained unit combination, and the evaluation indicators of the neighborhood search algorithm are evaluated by the evaluation function calculation; the code generation model and the prompt word generation model are determined; the code generation model is used to generate the code of the neighborhood search algorithm; the prompt word generation model is used to summarize the characteristics of the code corresponding to the generated neighborhood search algorithm and input it into the code generation model; the neighborhood search algorithm is generated based on the evolutionary algorithm; based on the pre-set prompt word and seed algorithm code, a batch of The code corresponding to the neighborhood search algorithm; the performance of the neighborhood search algorithm is evaluated based on the evaluation function of the neighborhood search algorithm, and all neighborhood search algorithms are ranked; based on the ranked neighborhood search algorithm, algorithm pairs are randomly extracted from it, and based on each pair of algorithms in the algorithm pair and the pre-set prompt words, a large model is generated using the prompt words to process each pair of algorithms, and algorithm improvement suggestions are generated according to the algorithm performance to update the neighborhood search algorithm; the performance of the generated neighborhood search algorithm is evaluated based on the evaluation function of the neighborhood search algorithm, and the neighborhood search algorithms are ranked; the process of generating the neighborhood search algorithm is repeated until the evaluation score of the neighborhood search algorithm remains unchanged after evaluating a set number of neighborhood search algorithms or the maximum number of generated algorithms is reached, and the optimal neighborhood search algorithm is obtained; Solving the safety-constrained unit combination includes: constructing a safety-constrained unit combination model of a limited neighborhood according to an optimal neighborhood search algorithm; solving the safety-constrained unit combination model of the limited neighborhood; constructing the safety-constrained unit combination model; and solving the safety-constrained unit combination model.

2. The safety-constrained unit commitment optimization acceleration method based on a large language model according to claim 1 is characterized in that: The basic data for the economic operation of the power system include the upper and lower limits of unit output, the upper limit of the unit ramp-up and reduction rate, the minimum continuous start-up and shutdown time of the unit, the unit operation cost function, the unit start-up and shutdown cost function, the power flow transfer distribution factor, the line transmission capacity and the system positive and negative reserve rate.

3. The safety-constrained unit commitment optimization acceleration method based on a large language model according to claim 1 is characterized in that: Generate the code corresponding to the seed algorithm, including determining the seed neighborhood search algorithm and generating code text based on the seed neighborhood search algorithm: A seed neighborhood search algorithm is determined, including: defining a neighborhood distance between the start-stop variables of the safety-constrained unit combination and the optimal solution of the integer relaxed safety-constrained unit combination, and obtaining a neighborhood constraint by limiting the neighborhood distance of the optimal solution to be no greater than the system reserve rate.

4. The safety-constrained unit commitment optimization acceleration method based on a large language model according to claim 1 is characterized in that: When constructing a safety-constrained unit commitment model with a limited neighborhood, the unit start-stop decision variables of the integer relaxation safety-constrained unit commitment model are adjusted to discrete variables, while the unit output decision variables remain continuous variables; neighborhood constraints are generated based on the optimal neighborhood search algorithm to obtain a safety-constrained unit commitment model with a limited neighborhood.

5. The safety-constrained unit commitment optimization acceleration method based on a large language model according to claim 1 is characterized in that: Solve the safety-constrained unit commitment model within a limited neighborhood, including: The mixed integer linear programming solver is used to solve the safety-constrained unit commitment model in a limited neighborhood, and the results of the safety-constrained unit commitment model in a limited neighborhood are obtained.

6. The safety-constrained unit commitment optimization acceleration method based on a large language model according to claim 1 is characterized in that: Construct a safety-constrained unit combination model, including: According to the safety-constrained unit commitment model with limited neighborhood, the neighborhood constraints are deleted to obtain the safety-constrained unit commitment model.

7. The safety-constrained unit commitment optimization acceleration method based on a large language model according to claim 1 is characterized in that: Solving the safety-constrained unit commitment model includes: using the obtained safety-constrained unit commitment model result of the limited neighborhood as a hot start solution of a mixed integer linear programming solver, using the mixed integer linear programming solver to solve the safety-constrained unit commitment model, and obtaining a result of the safety-constrained unit commitment model.

8. A safety-constrained unit commitment optimization acceleration system based on a large language model, which implements the safety-constrained unit commitment optimization acceleration method based on a large language model according to any one of claims 1 to 7, characterized in that: It includes an algorithm generating unit and a solving unit; An algorithm generation unit, which is used to generate a large language model-driven neighborhood search algorithm based on the large language model; the algorithm generation unit includes an acquisition unit, a model building and solving unit, a code generation unit, and an evolutionary update unit; An acquisition unit, used to acquire basic data on economic operation of the power system; Model building and solving unit, used to build and solve the integer relaxed safety constraint unit commitment model; A code generation unit, used to generate code corresponding to the seed algorithm; An evolutionary update unit, used to evolve the neighborhood search algorithm based on the large language model to obtain the optimal neighborhood search algorithm; A solving unit, used for solving the safety constraint unit combination; the solving unit includes a first combination model building unit, a combination model solving unit, a second combination model building unit and a solving sub-unit; The first combination model building unit is used for constructing a safety constraint unit combination model of a limited neighborhood using an optimal neighborhood search algorithm; A combination model solving unit, used to solve the safety constraint unit combination model of a limited neighborhood; The second combination model building unit is used to build a safety constraint unit combination model; The solving subunit is used to solve the safety constrained unit combination model.

Images