Method and device for determining unit scheduling scheme in power grid system

Abstract

The invention provides a method and device for determining a unit scheduling scheme in a power grid system, and belongs to the technical field of power system scheduling. The determination method comprises the steps that key variable data of the power grid system are acquired, and the key variable data comprise power output data, load demand data and power transmission line state data of a generator set; establishing a mixed integer nonlinear programming problem model according to the power output data, the load demand data and the power transmission line state data of the generator set; performing hierarchical processing on the mixed integer nonlinear programming problem model based on a constraint type to obtain an intermediate processing result; and calculating the intermediate processing result according to a self-adaptive optimization algorithm, and determining a scheduling scheme of the unit. According to the scheme, unit scheduling optimization of a large-scale regional power grid can be efficiently carried out under complex constraint conditions, and the scheduling scheme can be dynamically adjusted according to real-time data.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system dispatching, and specifically relates to a method, device, electronic equipment, computer-readable storage medium and computer program product for determining a unit dispatching plan in a power grid system. Background Art

[0002] With the continuous development of power systems and the expansion of regional power grids, the structure and operation characteristics of power grids have become increasingly complex. Power grids contain a large number of generators, transmission lines and substations, which face a variety of complex constraints and uncertainties in their operation. In order to ensure the safe, stable and efficient operation of the power grid, the grid dispatch decision-making process has become crucial, especially in large-scale regional power grids. How to reasonably dispatch the unit combination has become a very challenging problem. Summary of the invention

[0003] The present invention aims to solve at least one of the technical problems existing in the prior art and to provide a method, device, computer-readable storage medium and computer program product that can efficiently perform unit combination optimization of a large-scale regional power grid under complex constraints.

[0004] In order to achieve the above objectives, in a first aspect, an embodiment of the present disclosure provides a method for determining a unit dispatching scheme in a power grid system, the method comprising:

[0005] Acquiring key variable data of the power grid system, wherein the key variable data includes power output data of the generator set, load demand data and transmission line status data;

[0006] Establishing a mixed integer nonlinear programming problem model according to the power output data of the generator set, the load demand data and the transmission line status data;

[0007] Performing hierarchical processing on the mixed integer nonlinear programming problem model based on constraint types to obtain an intermediate processing result;

[0008] The intermediate processing results are calculated according to an adaptive optimization algorithm to determine a scheduling plan for the generator set.

[0009] In some embodiments, the mixed integer nonlinear programming problem model is hierarchically processed based on the constraint type to obtain an intermediate processing result, including:

[0010] Processing the mixed integer nonlinear programming problem model using convex hull constraint screening to obtain a first processing result;

[0011] Processing the first processing result by Lagrangian relaxation screening to obtain a second processing result;

[0012] The second processing result is processed by using a branch and bound and pruning method to obtain the intermediate processing result.

[0013] In some embodiments, the power grid system includes a power grid management system, a generator set control system, and a transmission line monitoring system, and the step of obtaining key variable data of the power grid system includes:

[0014] Acquire load demand data of the generator set from the power grid management system in real time, wherein the load demand data includes the load demand, fluctuation range and predicted load curve of each generator set;

[0015] Acquire power output data of the generator set in real time from the generator set control system, the power output data including current output power, upper and lower limits of rated power, start and stop status, start and stop time, and operating efficiency of each generator set;

[0016] The transmission line status data of the transmission line is acquired in real time from the transmission line monitoring system, wherein the transmission line status data includes the capacity limit, current load rate, voltage level and line loss of each transmission line.

[0017] In some embodiments, the mixed integer nonlinear programming problem model is established according to the power output data, load demand data and transmission line status data of the generator set; comprising:

[0018] A variable data set is established according to the power output data, load demand data and transmission line status data of the generator set, wherein the variable data set is expressed as:

[0019] X={P i,t ,U i,t ,D t ,L j,t ∣i∈[1,N],j∈[1,M],t∈[1,T]}

[0020] Among them, P i,t represents the power output of the ith generator set at time t, U i,t represents the start and stop status of the generator set at time t, D t represents the load demand of the power grid system at time t, L j,t represents the load rate of transmission line j at time t, N is the total number of generator sets in the power grid system, M is the total number of transmission lines in the power grid system, and T represents the number of moments in the dispatching cycle when collecting key variable data;

[0021] Preprocessing the variable data set, and generating a variable matrix according to the preprocessed variable data set, wherein the variable matrix includes comprehensive information of the generator set status, load demand and transmission line status;

[0022] A mixed integer nonlinear programming problem model is established according to the variable matrix.

[0023] In some embodiments, the mixed integer nonlinear programming problem model is established according to the power output data, load demand data and transmission line status data of the generator set; comprising:

[0024] Determining objective functions and constraints based on power output data, load demand data, and transmission line status data of the generator set;

[0025] According to the objective function and constraints, a mixed integer nonlinear programming problem model is established; the objective function is the total operating cost of the power grid system in the scheduling period; the constraints include system load demand constraints, upper and lower limit constraints of generator set power output, start and stop time constraints of generator sets and capacity constraints of the transmission line; wherein,

[0026] The total operating cost is expressed as:

[0027]

[0028] Among them, P i,t is the output power of generator set i at time t, U i,t is the start and stop status of the unit at time t, C i,t (P i,t ) is the power generation cost function of generator set i at time t, S i,t (U i,t ,U i,t-1 ) is the start-stop cost function of generator set i at time t; N is the total number of generator sets in the power grid system, and T represents the number of moments in the dispatching period when key variable data are collected;

[0029] The system load demand constraint is:

[0030]

[0031] The upper and lower limits of the power output of the generator set are:

[0032]

[0033] Among them, P i,min and P i,max are the minimum power and maximum power limits of generator set i respectively;

[0034] The start and stop time constraints of the generator set are:

[0035] U i,t -U i,t-1 ≥Ti,start ×(1-U i,t-1 )-T i,stop ×U i,t-1 ;

[0036] Among them, T i,start and T i,stop are the minimum start time and minimum stop time of generator set i respectively;

[0037] The capacity constraint of the transmission line is:

[0038]

[0039] Among them, L j,t represents the load rate of transmission line j at time t, L j,max is the maximum capacity of line j.

[0040] In some embodiments, calculating the intermediate processing result according to the adaptive optimization algorithm to determine the scheduling scheme of the generator set includes:

[0041] Inputting the intermediate processing results into an initial hybrid deep reinforcement learning model, so that the initial hybrid deep reinforcement learning model establishes a state space and an action space according to the key variable data of the power grid system, wherein the state space includes the current output power of the generator set, load demand, and transmission line status, and the action space includes the start and stop operation and power adjustment of the generator set;

[0042] Training the initial hybrid deep reinforcement learning model based on a cumulative reward function to optimize the strategy of the state space and the action space;

[0043] According to the state space and action space after strategy optimization, the scheduling plan of the generator set is determined; wherein,

[0044] The cumulative reward function is:

[0045]

[0046] Among them, R t is the cumulative reward at time t, γ is the discount factor, r(s k ,a k ) is in state s k Next, perform action a k Instant rewards received.

[0047] In some embodiments, calculating the intermediate processing result according to the adaptive optimization algorithm to determine the scheduling scheme of the generator set further includes:

[0048] In the process of training the initial hybrid deep reinforcement learning model based on the cumulative reward function to optimize the strategy of the state space and the action space, real-time operation data from the power grid system is received in real time, wherein the real-time operation data includes load demand changes, generator unit state changes, and transmission line state updates;

[0049] The distribution of the state space and the action space is dynamically adjusted according to the real-time operation data to adapt to the current state of the power grid system.

[0050] In some embodiments, calculating the intermediate processing result according to the adaptive optimization algorithm to determine the scheduling scheme of the generator set further includes:

[0051] Identify operating costs, scheduling stability and system security as multiple game parties;

[0052] The utility function of generator set i is calculated by using combinatorial optimization and game theory models. The utility function is expressed as:

[0053] U i (x) = α i C(x)+β i S(x)+γ i T(x)

[0054] Among them, C(x) is the operating cost, S(x) is the scheduling stability, T(x) is the system security, α i , β i and γ i are the weight coefficients of the operating cost, dispatching stability and system security of generator set i respectively;

[0055] Solving the utility function of generator set i through the Pareto optimal solution set of the game theory model to balance the relationship between different game parties in generator set i;

[0056] An immediate reward is determined based on the solution of the utility function for each generator set.

[0057] In some embodiments, the method further comprises:

[0058] The generator sets in the power grid system are dispatched according to the determined dispatching plan for the generator sets.

[0059] In a second aspect, the present disclosure further provides a device for determining a unit dispatching scheme in a power grid system, the device comprising:

[0060] A data receiving module, used to obtain key variable data of the power grid system, wherein the key variable data includes power output data of the generator set, load demand data and transmission line status data;

[0061] A model building module, used to build a mixed integer nonlinear programming problem model according to the power output data, load demand data and transmission line status data of the generator set;

[0062] A constraint screening module, used for performing hierarchical processing on the mixed integer nonlinear programming problem model based on constraint types to obtain an intermediate processing result;

[0063] The adaptive optimization module is used to calculate the intermediate processing results according to the adaptive optimization algorithm to determine the scheduling plan of the generator set.

[0064] In a third aspect, the present disclosure further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the computing method of the first aspect described above is implemented.

[0065] In a fourth aspect, the present disclosure further provides a computer-readable storage medium storing a computer program for executing the computing method of the first aspect.

[0066] In a fifth aspect, the present disclosure further provides a computer program product, including a computer program / instruction, which implements the steps of the computing method of the first aspect when the computer program / instruction is executed by a processor.

[0067] The present disclosure combines a multi-level constraint screening mechanism and an adaptive optimization algorithm to optimize the unit dispatching scheme in a large-scale regional power grid system. Specifically, the present disclosure simplifies the complex mixed integer nonlinear programming problem into an easy-to-solve problem through a series of efficient constraint screening mechanisms (such as convex hull constraint screening, Lagrangian relaxation screening, and branch and bound, etc.), thereby greatly reducing the computational complexity. At the same time, the application of the adaptive optimization algorithm enables the model to be dynamically adjusted according to the real-time data of the power grid system, thereby improving the adaptability and real-time response capability of the dispatching scheme. Compared with traditional optimization methods, the innovation of the present disclosure is that by organically combining these optimization techniques to form an integrated solution framework, not only the solution efficiency is improved, but also the actual changes of the power grid can be dynamically adapted, achieving a more intelligent and efficient unit combination optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0069] Figure 1A flow chart of a method for determining a unit dispatching scheme in a power grid system provided by an embodiment of the present disclosure;

[0070] Figure 2 An implementation provided by the present disclosure Figure 1 The method flow chart of step S3 in FIG.

[0071] Figure 3 An implementation provided by the present disclosure Figure 1 The method flow chart of step S1 in FIG.

[0072] Figure 4 An implementation provided by the present disclosure Figure 1 The method flow chart of step S4 in FIG.

[0073] Figure 5 A structural block diagram of a device for determining a unit dispatching scheme in a power grid system provided by an embodiment of the present disclosure;

[0074] Figure 6 A structural block diagram of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0075] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail and completely in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0076] Most of the unit dispatching schemes in existing power grid systems are determined based on traditional optimization algorithms, such as mixed integer programming, dynamic programming or heuristic algorithms. These methods are effective in dealing with small-scale power grid systems or simple constraints. However, when these methods are applied to large-scale regional power grid systems, the efficiency of solving them is significantly reduced due to the large number of units in the power grid system and the complex constraints. In addition, the real-time requirements of the power grid require that the dispatching decision can respond quickly to changes in the power grid state, while traditional methods often find it difficult to find the optimal solution that satisfies all constraints within a reasonable time. With the increase in the proportion of renewable energy in the power grid system, the uncertainty and volatility in the unit dispatching process are further aggravated. The existing methods are unable to cope with such complexity, resulting in the dispatching scheme being difficult to achieve the expected results in practical applications.

[0077] Based on the above problems existing in the prior art, the present disclosure proposes a solution that combines a multi-level constraint screening mechanism and an adaptive optimization algorithm to solve the unit dispatch optimization problem of large-scale regional power grids. The present disclosure simplifies the complex mixed integer nonlinear programming problem into an easy-to-solve problem, thereby greatly reducing the computational complexity. At the same time, the application of the adaptive optimization algorithm enables the model to be dynamically adjusted according to the real-time data of the power grid, improving the adaptability and real-time response capability of the dispatching scheme.

[0078] In a first aspect, an embodiment of the present disclosure provides a method for determining a unit dispatching plan in a power grid system.

[0079] Figure 1 The present invention provides a flowchart of a method for determining a unit dispatching scheme in a power grid system according to an embodiment of the present invention. Figure 1 As shown, the determination method includes the following steps:

[0080] S1. Acquire key variable data of the power grid system, wherein the key variable data includes power output data of generator sets, load demand data and transmission line status data.

[0081] Specifically, the present disclosure first needs to obtain the necessary key variable data from the large-scale regional power grid system. Accurately obtaining variable data can provide an important basis for the construction and optimization solution of the subsequent mixed integer nonlinear programming problem model, and directly affect the accuracy and effectiveness of the final unit scheduling plan. Among them, the power output data of the generator set is used to determine the availability and operating status of the generator set in each time period, providing a basis for subsequent scheduling decisions; load demand data is an important parameter for load balancing calculations to ensure the balance of supply and demand in the power grid system; transmission line status data is used to ensure that overload or voltage instability will not occur during the unit scheduling process.

[0082] S2. Establishing a mixed integer nonlinear programming problem model according to the power output data, load demand data and transmission line status data of the generator set.

[0083] Specifically, the power grid system faces a variety of complex constraints and uncertainties during operation. The mixed integer nonlinear programming (MINLP) problem model describes the scheduling problem of units in the power grid system by defining the objective function and operation constraints. It can be widely used in optimization problems, especially those with discrete and continuous decision variables, and is used to deal with unit scheduling optimization problems.

[0084] S3. Performing hierarchical processing on the mixed integer nonlinear programming problem model based on constraint types to obtain intermediate processing results.

[0085] Specifically, the constraint types include but are not limited to convex hull constraint screening, Lagrangian relaxation screening, branch and bound screening, etc. Hierarchical processing refers to the reasonable division of complex constraints in the mixed integer nonlinear programming problem model, and processing parts of the mixed integer nonlinear programming problem model in turn based on the constraint type, and each processing is performed on the basis of the results of the previous processing.

[0086] When solving a mixed-integer nonlinear programming problem model, the disclosed embodiment gradually simplifies the complexity of the problem through hierarchical screening based on the constraint type, so that large-scale complex problems that may originally be unsolvable become solvable, thereby reducing the computational complexity.

[0087] S4. Calculate the intermediate processing results according to an adaptive optimization algorithm to determine a scheduling plan for the generator set.

[0088] Specifically, after simplifying the mixed integer nonlinear programming problem model through a multi-level constraint screening mechanism to obtain an intermediate processing result, the present disclosure further uses an adaptive optimization algorithm to solve the intermediate processing result (simplified model). Among them, the adaptive optimization algorithm dynamically adjusts the optimization strategy by combining real-time data to ensure the adaptability and effectiveness of the scheduling scheme in practical applications. The present disclosure improves the quality and execution efficiency of the scheduling scheme through the flexibility and real-time performance of the optimization algorithm.

[0089] The disclosed embodiment combines a multi-level constraint screening mechanism and an adaptive optimization algorithm for the optimization problem of unit dispatching schemes in large-scale regional power grid systems. Specifically, the disclosed embodiment simplifies complex mixed integer nonlinear programming problems into easy-to-solve problems through a series of efficient constraint screening mechanisms (such as convex hull constraint screening, Lagrangian relaxation screening, and branch and bound, etc.), thereby greatly reducing the computational complexity. At the same time, the application of the adaptive optimization algorithm enables the model to be dynamically adjusted according to the real-time data of the power grid system, thereby improving the adaptability and real-time response capability of the dispatching scheme. Compared with traditional optimization methods, the innovation of the disclosed embodiment is that by organically combining these optimization techniques to form an integrated solution framework, not only the solution efficiency is improved, but also the actual changes in the power grid can be dynamically adapted, achieving a more intelligent and efficient unit combination optimization.

[0090] Figure 2 An implementation provided by the present disclosure Figure 1 Flow chart of the method for step S3 in FIG.

[0091] like Figure 2 As shown, in some embodiments, step S3, hierarchical processing of the mixed integer nonlinear programming problem model based on the constraint type to obtain an intermediate processing result, includes steps S31 to S33, specifically including:

[0092] S31, using convex hull constraint screening to process the mixed integer nonlinear programming problem model to obtain a first processing result.

[0093] Specifically, for the non-convex constraints in the mixed integer nonlinear programming problem model, this embodiment first performs convex hull constraint screening. Specifically, by Transformed into the convex hull area Conv(X), complex non-convex problems can be simplified into relatively easy-to-handle convex problems.

[0094]

[0095] Among them, λ i is the weight coefficient, x i is a point that belongs to the non-convex constraint region. Through this process, most feasible solutions are retained while significantly reducing the complexity of the search space.

[0096] S32: Process the first processing result by using Lagrangian relaxation screening to obtain a second processing result.

[0097] Specifically, for the linear infeasible constraints in the first processing result, this embodiment uses the Lagrangian relaxation method to screen them. The specific steps include: relaxing the original constraints h(x)≤0 and g(x)=0 as part of the objective function, and constructing the Lagrangian function L(x,λ,μ):

[0098] L(x,λ,μ)=f(x)+λ T h(x)+μ T g(x)

[0099] Among them, f(x) is the objective function, h(x) and g(x) represent inequality constraints and equality constraints respectively, and λ and μ are Lagrange multipliers. Through the dual optimization algorithm, the Lagrange function is gradually solved, the Lagrange multiplier is iteratively adjusted, and the solution that satisfies the relaxed constraints is screened out, further narrowing the search space.

[0100] S33: Process the second processing result by using a branch and bound and pruning method to obtain the intermediate processing result.

[0101] Specifically, for the integer constraints and combinatorial problems in the second processing result, this embodiment uses the branch and bound method for processing. First, a branch and bound tree is constructed to decompose the problem into several sub-problems. In each sub-problem, its upper and lower bounds are calculated, and these boundary values ​​are quickly estimated by dynamic programming or linear programming methods. Through pruning operations, branches whose lower bounds exceed the upper bounds of the currently known optimal solutions or obviously do not contain the optimal solutions are removed from the search space. This process significantly reduces the complexity of the solution and retains the search path that is most likely to contain the global optimal solution.

[0102] After the above three levels of screening, the originally complex mixed integer nonlinear programming problem model is constructed into a simplified solution space to obtain an intermediate processing result. The intermediate processing result only retains the solution that meets all the screening conditions, thereby greatly reducing the complexity of the problem and laying the foundation for the subsequent adaptive optimization algorithm processing. Through this multi-level constraint screening mechanism, the present embodiment can effectively simplify the solution process while ensuring the accuracy of the model, thereby improving the overall efficiency of the optimization algorithm.

[0103] The disclosed embodiment organically combines convex hull constraint screening, Lagrangian relaxation screening, and branch and bound with pruning methods to form a multi-level constraint screening mechanism. The convex hull screening, Lagrangian relaxation or branch and bound and pruning methods in the prior art are usually applied to the solution of single constraints or simple problems. In the disclosed embodiment, we use a multi-level constraint screening mechanism, combined with a variety of optimization methods, to hierarchically process large-scale, multi-constrained complex scheduling problems in a modular manner, which is particularly suitable for unit combination optimization problems in large-scale regional power grids. This integrated application method improves the solution efficiency and breaks through the limitations of a single technology in dealing with complex constraints and large-scale problems. In this way, we not only simplify the complexity of the problem through the screening mechanism, but also can dynamically adjust the screening and optimization strategies based on the adaptive optimization algorithm to cope with the uncertainty and volatility in the actual operation of the power grid.

[0104] In the disclosed embodiment, hierarchical screening refers to dividing the problem model into different parts according to the different types of constraints in the mixed integer nonlinear programming problem, each part corresponds to a specific type of constraint (non-convex constraint, linear infeasible constraint, integer constraint), and a corresponding algorithm is used for each part. Finally, the processing results of each part are combined to obtain a simplified solution space to facilitate subsequent optimization and solution. In the disclosed embodiment, the mixed integer nonlinear problem model is not simply split into three independent parts for each solution, but through a hierarchical screening mechanism, it is interactive, integrated and progressive, so that each processing step cooperates with each other and gradually simplifies the complexity of the problem. Specifically:

[0105] Interaction: The convexification of non-convex constraints may affect the feasibility of linear constraints, and the relaxation of linear constraints may change the range of integer variables. Therefore, the processing at each level needs to take into account the impact on the overall problem and cannot be completely independent.

[0106] Integration: After the screening at each level is completed, the processed constraints and variables need to be recombined to form a new, simplified solution space that not only satisfies the processed constraints but also retains the key characteristics of the original problem.

[0107] Progressiveness: Hierarchical screening is a progressive process, and the processing of each layer provides a simpler problem structure for the next layer. For example, after convex hull screening, the non-convexity of the problem is reduced, which is convenient for Lagrangian relaxation; after Lagrangian relaxation, the range of continuous variables is clearer, and the solution of integer variables is more efficient.

[0108] Therefore, the hierarchical screening process in the disclosed embodiment is a systematic, multi-step optimization strategy that gradually simplifies the complexity of the problem by using the most appropriate algorithm for different types of constraints. The ultimate goal is to build an optimization model that not only meets the main constraints of the original problem but is also easy to solve, providing an efficient solution basis for the subsequent adaptive optimization algorithm.

[0109] It should be noted that in the embodiments of the present disclosure, the mixed integer nonlinear programming problem model is hierarchically processed based on the constraint type, which is not a solution that can be easily thought of by those skilled in the art. Specifically:

[0110] The structured decomposition of mixed-integer nonlinear programming problem models is not intuitive: large-scale mixed-integer nonlinear programming problems are usually highly coupled, and there may be complex interactions between constraints. Directly decomposing the problem into three parts: non-convex constraints, linear infeasible constraints, and integer constraints requires a deep understanding of the problem structure and ensuring that the solutions to the decomposed parts do not lose the key characteristics of the original problem. This decomposition is not a simple modularization process, but requires innovative ideas to identify and classify the properties of the constraints.

[0111] The particularity of power grid dispatching: Power grid dispatching has specific industry characteristics and constraints, such as the start and stop time of generator sets, power change rate, line capacity limit, etc. These constraints may be presented in a complex form in mathematics, and traditional optimization methods may not be able to handle them effectively. Classifying these specific constraints and adopting corresponding optimization algorithms requires a deep understanding of the operation mechanism of the power system and the combination of optimization theory, which goes beyond the scope of general optimization problems.

[0112] Comprehensive application of multiple optimization algorithms: In the existing technology, a single optimization algorithm is usually used to solve a specific type of problem. It is not a conventional idea to organically combine multiple optimization techniques such as convex hull constraint screening, Lagrangian relaxation method and branch and bound method. This comprehensive application requires solving the connection and compatibility problems between the algorithms to ensure that the final integrated result still meets the requirements of the original problem.

[0113] Integration of solution results and global optimality: After processing different types of constraints separately, it is a complex problem to effectively integrate the respective solution results to ensure that the final solution not only satisfies all constraints but is also close to the global optimal solution. This involves coordinating the partial solutions and resolving possible conflicts. Such an integration process is not a common processing method in the prior art, which reflects the innovation of this solution.

[0114] Meeting the challenges of real-time and large-scale operation: Real-time dispatching of large-scale power grids places extremely high demands on the efficiency and response speed of algorithms. Traditional methods are difficult to meet this demand. This solution achieves effective response to real-time and large-scale operation through hierarchical decomposition and application of specialized algorithms, which is a breakthrough in existing technologies.

[0115] It should also be noted that in the embodiment of the present disclosure, the mixed integer nonlinear programming problem model is processed based on the constraint type, and the reasons for sequentially processing non-convex constraints, linear infeasible constraints and integer constraints are:

[0116] (1) Dealing with non-convex constraints first can reduce the non-convexity of the problem

[0117] Non-convex constraints are the most difficult part of optimization problems. Directly solving non-convex problems may lead to local optimal solutions or failure to converge. By screening the convex hull constraints and approximating the non-convex constraint region as a convex set, complex non-convex optimization problems can be transformed into convex optimization problems, improving the feasibility and efficiency of the solution.

[0118] (2) Secondly, dealing with linear infeasible constraints can increase the possibility of feasible solution space.

[0119] Even after convexification, some linear constraints may still make the problem infeasible. Through the Lagrangian relaxation method, these strict linear infeasible constraints are relaxed as part of the objective function, increasing the number of feasible solutions and providing a larger solution space for subsequent integer constraint processing.

[0120] (3) Finally, integer constraints are processed to reduce computational complexity.

[0121] After the continuous variables have been processed in the first two steps and the feasible range has been determined, processing the integer variables can greatly reduce the solution space that needs to be searched. Integer constraints often involve combinatorial explosion, and simplifying the continuous part first can improve the efficiency of integer optimization.

[0122] The disclosed embodiment sequentially uses convex hull constraint screening, Lagrangian relaxation screening and branch and bound and pruning methods to process non-convex constraints, linear infeasible constraints and integer constraints respectively. On the one hand, according to the above-mentioned sequential processing, each step further simplifies the problem on the basis of the previous step, making the subsequent solution more efficient and gradually simplifying the complexity of the problem; on the other hand, this sequential processing method takes into account the mutual influence between constraints. First, the non-convex constraints that have the greatest impact on the problem solution are processed, then the linear constraints are processed, and finally the integer constraints are processed. This can better coordinate the relationship between various constraints, avoid conflicts and the generation of infeasible solutions, and ensure coordination between constraints; thirdly, through this progressive processing, the problem is gradually simplified, and the solution space is effectively reduced, so that the optimization algorithm can find the global optimal solution in a smaller space, improving the solution efficiency and the reliability of the results.

[0123] It should also be noted that the processing order of the above-mentioned hierarchical screening first requires an in-depth understanding of the structure of the mixed integer nonlinear programming problem, knowing which constraints have the greatest impact on the solution of the problem, and how to simplify the problem by processing in sequence. This is not a simple decomposition, but based on a comprehensive consideration of the constraint characteristics and the applicability of the optimization algorithm, it requires a deep understanding of the mixed integer nonlinear programming problem model; secondly, in general, technicians in this field may tend to process various constraints in parallel in order to improve the calculation speed. However, in complex power grid scheduling problems, there are often interdependencies and influences between constraints, and simple parallel processing may lead to feasibility problems of the solution. The method disclosed in the present invention creates better conditions for subsequent steps by first processing key constraints and gradually simplifying them. It is a non-intuitive processing method; furthermore, this sequential processing not only improves the calculation efficiency, but also enhances the stability of the algorithm and the quality of the results. In the actual application of power grid scheduling, this processing method can better meet the requirements of real-time and accuracy, and can improve the overall optimization performance.

[0124] Therefore, in the present disclosure, the processing of the three constraints is carried out in the order of first non-convex constraints, then linear infeasible constraints, and finally integer constraints. This processing order is based on a deep understanding of the structure of the optimization problem and the characteristics of the constraints, and aims to gradually simplify the problem and improve the efficiency of solving and the quality of the results. The innovation of this sequential processing method is that it is not a conventional method that is easy for technicians in this field to think of, but requires a deep understanding of the problem and innovative algorithm design, which reflects the technical progress of this solution in dealing with complex optimization problems.

[0125] The disclosed embodiment decomposes the mixed integer nonlinear programming (MINLP) problem into non-convex constraints, linear infeasible constraints and integer constraints, and processes each part separately using a corresponding optimization algorithm (such as convex hull constraint screening, Lagrangian relaxation screening, branch and bound and pruning), and finally integrates the results to obtain a simplified solution space, which has the following technical effects and benefits:

[0126] Improve solution efficiency: For different types of constraints, the most suitable algorithm is used to process them, which can give full play to the advantages of each algorithm and significantly reduce the computational complexity. For example, convex hull constraint screening can effectively process non-convex constraints and transform them into convex optimization problems, which are easy to solve; Lagrange relaxation method can relax linear constraints that are difficult to meet as part of the objective function, thereby avoiding infeasible solutions; branch and bound method is very efficient for integer constraints.

[0127] Enhance the solvability of problems: Through hierarchical screening, the complexity of the problem is gradually simplified, making large-scale complex problems that may have been unsolvable become solvable. This is especially important for large-scale systems with high dimensions and multiple constraints such as power grids.

[0128] Improve the robustness and stability of the algorithm: Handling different types of constraints separately can reduce the numerical instability and convergence problems that the algorithm may encounter when dealing with global complex problems, and improve the robustness of the overall solution.

[0129] Facilitates real-time dynamic adjustment: This decomposition method allows the algorithm of each part to be dynamically adjusted according to real-time data, better adapting to changes in the operating status of the power grid and meeting the needs of real-time optimization.

[0130] In some embodiments, the power grid system includes a power grid management system, a generator set control system, and a transmission line monitoring system.

[0131] Figure 3 An implementation provided by the present disclosure Figure 1 Flow chart of the method for step S1 in FIG.

[0132] In some embodiments, step S1, obtaining key variable data of the power grid system, includes steps S11 to S13, such as Figure 3 As shown, S1 specifically includes:

[0133] S11. Acquire the load demand data of the generator set from the power grid management system in real time, wherein the load demand data includes the load demand, fluctuation range and predicted load curve of each generator set.

[0134] S12. Acquire power output data of the generator set from the generator set control system in real time, wherein the power output data includes current output power, upper and lower limits of rated power, start / stop status, start / stop time, and operating efficiency of each generator set.

[0135] S13. Acquire the transmission line status data of the transmission line in real time from the transmission line monitoring system, wherein the transmission line status data includes the capacity limit, current load rate, voltage level and line loss of each transmission line.

[0136] After being acquired, the above data are integrated into the key variable data set X, which is expressed as:

[0137] X={P i,t ,U i,t ,D t ,L j,t ∣i∈[1,N],j∈[1,M],t∈[1,T]}

[0138] Among them, P i,t represents the power output of the ith generator set at time t, U i,t represents the start and stop status of the i-th generator set at time t, D t represents the system load demand at time t, L j,t represents the load rate of transmission line j at time t.

[0139] In some embodiments, step S3, establishing a mixed integer nonlinear programming problem model according to the power output data, load demand data and transmission line status data of the generator set, specifically comprises the following steps:

[0140] A variable data set is established according to the power output data, load demand data and transmission line status data of the generator set; the variable data set is preprocessed, and a variable matrix is ​​generated according to the preprocessed variable data set, wherein the variable matrix contains comprehensive information of the generator set status, load demand and transmission line status; and a mixed integer nonlinear programming problem model is established according to the variable matrix.

[0141] Wherein, the variable data set is represented as:

[0142] X={P i,t ,U i,t ,D t ,L j,t ∣i∈[1,N],j∈[1,M],t∈[1,T]}

[0143] Among them, P i,t represents the power output of the ith generator set at time t, U i,trepresents the start and stop status of the generator set at time t, D t represents the load demand of the power grid system at time t, L j,t It represents the load rate of transmission line j at time t, N is the total number of generator sets in the power grid system, M is the total number of transmission lines in the power grid system, and T represents the number of moments in the dispatching cycle to collect key variable data.

[0144] In this embodiment, in order to ensure the accuracy and continuity of the data, the variable data set X is also preprocessed. The preprocessing step includes removing noise data and abnormal data, and performing necessary time series interpolation and smoothing to ensure the consistency and reliability of the data in time. These preprocessing operations lay the foundation for the subsequent establishment of an accurate mixed integer nonlinear programming problem model.

[0145] In some embodiments, step S3, establishing a mixed integer nonlinear programming problem model based on the power output data, load demand data and transmission line status data of the generator set; specifically includes: determining the objective function and constraints based on the power output data, load demand data and transmission line status data of the generator set; and establishing a mixed integer nonlinear programming problem model based on the objective function and constraints.

[0146] Among them, the objective function is the total operating cost of the power grid system during the scheduling period; the constraints include system load demand constraints, upper and lower limit constraints on generator set power output, start and stop time constraints on generator sets, and capacity constraints on the transmission lines.

[0147] In some embodiments, the objective function f(x) represents the minimization of the total operating cost in the entire scheduling period, including the power generation cost and the start-stop cost. The total operating cost is expressed as:

[0148]

[0149] Among them, P i,t is the output power of generator set i at time t, U i,t is the start and stop status of the unit at time t, C i,t (P i,t ) is the power generation cost function of generator set i at time t, S i,t (U i,t ,U i,t-1 ) is the start-stop cost function of generator set i at time t; N is the total number of generator sets in the power grid system, and T represents the number of moments in the dispatching cycle to collect key variable data.

[0150] To ensure the supply and demand balance of the power grid system, the mixed integer nonlinear programming problem model requires that in each time period t, the sum of the output power of all generator sets must be equal to the load demand of the system. In some embodiments, the system load demand constraint is:

[0151]

[0152] In addition, the output power of each generator set must be within its rated power range, that is, the upper and lower limits of the generator set power output are:

[0153]

[0154] Among them, P i,min and P i,max are the minimum power and maximum power limits of generator set i respectively.

[0155] In order to ensure that the start and stop operation of the unit meets the actual operation requirements, the start and stop time constraints of the generator set are:

[0156] U i,t -U i,t-1 ≥T i,start ×(1-U i,t-1 )-T i,stop ×U i,t-1 ;

[0157] Among them, T i,start and T i,stop are the minimum start time and minimum stop time of generator set i respectively.

[0158] To ensure that the load of the transmission line is within its design capacity, the capacity constraint of the transmission line is:

[0159]

[0160] Among them, L j,t represents the load rate of transmission line j at time t, L j,max is the maximum capacity of line j.

[0161] Through the definition of the above objective function and constraint conditions, this embodiment establishes a complete mixed integer nonlinear programming problem model. This model effectively describes the complexity of the unit combination scheduling problem in the power grid and provides a mathematical basis for the subsequent optimization solution process.

[0162] In the embodiment of the present disclosure, in steps S1 to S3, after the mixed integer nonlinear programming problem model is established, in order to cope with the complexity and difficulty of solving the model, the present invention adopts a multi-level constraint screening mechanism to process the model. Through this screening mechanism, the solution space of the problem can be effectively simplified and the efficiency of the optimization solution can be improved. This process is to reduce the computational complexity while ensuring the accuracy of the model, and lay the foundation for the adaptive optimization algorithm in the subsequent step S4.

[0163] Figure 4 An implementation provided by the present disclosure Figure 1 Flow chart of the method for step S4 in FIG.

[0164] In some embodiments, step S4, calculating the intermediate processing result according to the adaptive optimization algorithm to determine the scheduling plan of the unit, includes S41-S43, such as Figure 4 As shown, specifically including:

[0165] S41. Input the intermediate processing results into the initial hybrid deep reinforcement learning model, so that the initial hybrid deep reinforcement learning model establishes a state space and an action space according to the key variable data of the power grid system, wherein the state space includes the current generator set output power, load demand, and transmission line status, and the action space includes the start and stop operations and power adjustment of the generator set.

[0166] S42. Training the initial hybrid deep reinforcement learning model based on a cumulative reward function to perform strategy optimization on the state space and action space.

[0167] S43. Determine a scheduling plan for the unit according to the state space and action space after strategy optimization.

[0168] Wherein, the cumulative reward function is:

[0169]

[0170] Among them, R t is the cumulative reward at time t, γ is the discount factor, r(s k ,a k ) is in state s k Next, perform action a k Instant rewards received.

[0171] Specifically, in this embodiment, the hybrid deep reinforcement learning algorithm (HDRL) is first initialized, and the intermediate processing results (simplified problem model) processed by the multi-level constraint screening mechanism are used as input. The HDRL model establishes a state space S and an action space A based on the key variable data of the power grid (including historical operation data and current status). Among them, the state space S includes the output power, load demand, transmission line status, etc. of the current generator set, while the action space A includes the start and stop operations of the unit and the scheduling operations such as power adjustment.

[0172] In the state space S and action space A, the HDRL model learns and makes decisions through a policy optimization algorithm. The immediate reward function r(s,a) is the key to optimization and is defined as a measure of the pros and cons of the scheduling scheme based on operating cost, scheduling stability, and system security:

[0173]

[0174] Among them, R t is the cumulative reward, and γ is the discount factor, which indicates the degree of consideration of future rewards. By maximizing the cumulative reward R t ,The model continuously optimizes the scheduling strategy to make it tend to the global optimal solution.

[0175] In some embodiments, S4, calculating the intermediate processing results according to the adaptive optimization algorithm to determine the scheduling plan of the unit, includes not only S41 to S43, but also S44 (not shown in the figure), specifically including:

[0176] S44. In the process of training the initial hybrid deep reinforcement learning model based on the cumulative reward function to perform strategy optimization on the state space and the action space, real-time operation data from the power grid system is received in real time, and the real-time operation data includes changes in load demand, changes in the state of generator units, and updates on the status of transmission lines; the distribution of the state space and the action space is dynamically adjusted according to the real-time operation data to adapt to the current state of the power grid system.

[0177] In the solution process, real-time operation data from the power grid is received and processed in real time, such as changes in load demand, generator status updates, and transmission line status changes. Based on these real-time data, the distribution and strategy of the state space S and action space A are dynamically adjusted to ensure that the optimization process can respond to changes in the power grid in a timely manner. The dynamic adjustment strategy includes adjusting the start and stop sequence of the units, output power distribution, and load optimization of the transmission lines, so as to maximize the actual application effect of the scheduling scheme.

[0178] In some embodiments, S4, calculating the intermediate processing results according to the adaptive optimization algorithm to determine the scheduling plan of the unit, includes not only S41 to S44, but also S45, the process of determining the immediate reward (S45 is not shown in the figure).

[0179] Among them, S45 specifically includes: determining operating costs, scheduling stability and system safety as multiple game parties; using combinatorial optimization and game theory models to calculate the utility function of generator set i, and solving the utility function of generator set i through the Pareto optimal solution set of the game theory model to balance the relationship between different game parties in generator set i; determining instant rewards based on the solution results of the utility function of each generator set.

[0180] Wherein, the utility function is expressed as:

[0181] U i (x) = α i C(x)+β i S(x)+γ i T(x)

[0182] Among them, C(x) is the operating cost, S(x) is the scheduling stability, T(x) is the system security, α i , β i and γ i are the weight coefficients of operating cost, dispatching stability and system security of generator set i respectively.

[0183] Specifically, in this embodiment, the adaptive optimization algorithm also combines combinatorial optimization with game theory models. By converting the multi-objective optimization problem into a cooperative game of multiple game players, defining the utility function, and calculating the Pareto optimal solution set, the model achieves a dynamic balance between multiple objectives.

[0184] In order to further improve the solution efficiency, this embodiment uses a distributed parallel solution algorithm (DPPA) to decompose the adaptive optimization problem into multiple subtasks and execute them in parallel. A multi-core processor or GPU is used to process each subtask in parallel, and the solution results of each subtask are integrated in real time under a coordination mechanism. Through this distributed parallel processing method, the solution speed is greatly accelerated, while ensuring the stability and accuracy of the solution.

[0185] Through the implementation of the above-mentioned adaptive optimization algorithm, the disclosed embodiment can optimize the unit dispatching plan in real time in a complex and changeable power grid environment, improve the flexibility and response speed of power grid dispatching, and ensure the effectiveness of the plan in actual operation.

[0186] In some embodiments, the method not only includes S1 to S4, but also includes S5: scheduling the units in the power grid system according to the determined scheduling plan of the units.

[0187] After solving and dynamically adjusting the optimization strategy through the adaptive optimization algorithm, the ultimate goal of this disclosure is to generate and output an optimized unit dispatching plan for dispatching units in a power grid system. This plan aims to balance the safety, stability and economy of the power grid system and can be actually applied to the dispatching operation of large-scale regional power grids. The output dispatching plan not only needs to meet all the operating constraints, but also needs to be feasible and effective in actual operation.

[0188] In this embodiment, the specific process of finally determining the optimized unit commitment scheduling plan includes the following key steps:

[0189] 1. Summarize the optimization solution results:

[0190] After completing the adaptive optimization solution, this embodiment first summarizes the optimization solution results in each time period. The results include the start and stop status U of each generator set. i,t , output power P i,t and the load status L of each transmission line j,t The summarized results must satisfy all constraints defined in step S2, such as upper and lower limits of power output, load balance, start and stop time, and capacity constraints of transmission lines, to ensure that the generated solution is a feasible solution.

[0191] 2. Formation of optimization plan:

[0192] After summarization and verification, this embodiment forms a complete unit commitment scheduling plan based on the optimization solution results. The plan includes the following contents:

[0193] Unit start and stop status: the start and stop decision U of each generator unit in each time period i,t , ensuring that the number of starts and stops of the units is minimized while meeting the load demand of the power grid, so as to reduce the start and stop costs.

[0194] Power allocation: Grid load demand D for each time period t The optimized solution provides detailed generator power distribution P i,t , ensuring that each unit operates within its rated power range and that the total output power meets the load requirements.

[0195] Transmission line load: The solution also includes the load condition L of each transmission line. j,t , ensure that the line load factor L j,t Not exceeding its maximum capacity L j,max , to prevent line overload or voltage instability.

[0196] 3. Evaluation and verification of the scheme:

[0197] In this embodiment, after the plan is formed, the scheduling plan is evaluated and verified through a preset simulation platform. Through simulation, the performance of the plan in different operating scenarios can be tested, such as load peaks, renewable energy fluctuations, and equipment failures.

[0198] The main indicators of simulation verification include: total operating cost, system stability, response time, and scheduling effectiveness in emergencies, ensuring that the solution can maintain good performance in various scenarios.

[0199] 4. Output of the scheduling plan:

[0200] After the evaluation and verification are passed, the present embodiment finally outputs the optimized unit combination dispatching scheme, which can be in the form of digital files, dispatching instructions or other formats suitable for reading and executing by the power grid dispatching system.

[0201] The solution is directly used to guide grid dispatching operations, helping grid managers achieve efficient resource allocation and reliable power supply.

[0202] Through the above steps, the unit combination dispatching scheme finally determined by the present disclosure is not only optimal in theory, but also verified by simulation to ensure its feasibility and effectiveness in practical applications. This scheme provides strong support for the stable operation and cost control of the power grid.

[0203] The present disclosure greatly simplifies the complexity of solving the problem of unit combination of large-scale regional power grids by combining a multi-level constraint screening mechanism with an adaptive optimization algorithm, and improves the calculation speed and efficiency. At the same time, through the dynamic adjustment of real-time data, the adaptability and real-time performance of the dispatching scheme are guaranteed, so that the power grid can maintain stable operation under load changes, equipment failures, etc. This method can effectively reduce operating costs, improve the flexibility and safety of power grid dispatching, and has significant practical application value.

[0204] The disclosed embodiment uses a multi-level constraint screening mechanism, combined with a variety of optimization methods, to hierarchically process large-scale, multi-constrained complex scheduling problems in a modular manner, thereby improving the solution efficiency, breaking through the limitations of a single technology in dealing with complex constraints and large-scale problems, and is particularly suitable for the unit combination optimization problem of a large-scale regional power grid. In this way, the disclosure not only simplifies the complexity of the problem through the screening mechanism, but also dynamically adjusts the screening and optimization strategies to cope with the uncertainty and volatility in the actual operation of the power grid. In addition, the disclosure closely combines the adaptive optimization algorithm with these screening mechanisms, especially using real-time data to dynamically adjust the strategy during the scheduling process, making the solution process more intelligent and efficient. At the same time, the disclosure introduces a real-time data feedback mechanism in the scheduling process, and uses the real-time operation data of the power grid to continuously adjust the optimization strategy. This dynamic adjustment mechanism enhances the adaptability of the solution, so that the scheduling solution can be updated according to the real-time changes in the operating status of the power grid, which is a problem that has not been fully solved in the prior art.

[0205] In a second aspect, based on the same inventive concept, an embodiment of the present disclosure also provides a device for determining a unit dispatching plan in a power grid system.

[0206] Figure 5 FIG. 1 is a structural block diagram of a device 10 for determining a unit dispatching scheme in a power grid system provided by an embodiment of the present disclosure. Figure 5 As shown, the determination device 10 includes a data receiving module 100 , a model building module 200 , a constraint screening module 300 and an adaptive optimization module 400 .

[0207] The data receiving module 100 is used to obtain key variable data of the power grid system, and the key variable data includes power output data of the generator set, load demand data and transmission line status data.

[0208] The model building module 200 is used to build a mixed integer nonlinear programming problem model according to the power output data, load demand data and transmission line status data of the generator set;

[0209] The constraint screening module 300 is used to perform hierarchical processing on the mixed integer nonlinear programming problem model based on the constraint type to obtain an intermediate processing result;

[0210] The adaptive optimization module 400 is used to calculate the intermediate processing results according to the adaptive optimization algorithm to determine the scheduling plan of the unit.

[0211] In some embodiments, the constraint screening module 300 includes a first constraint module 301 , a second constraint module 302 , and a third constraint module 303 .

[0212] Among them, the first constraint module 301 is used to process the mixed integer nonlinear programming problem model using convex hull constraint screening to obtain a first processing result; the second constraint module 302 is used to process the first processing result using Lagrangian relaxation screening to obtain a second processing result; the third constraint module 303 is used to process the second processing result using branch and bound and pruning method to obtain the intermediate processing result.

[0213] In some embodiments, the power grid system includes a power grid management system, a generator set control system and a transmission line monitoring system, and the data receiving module 100 is specifically used to obtain the load demand data of the generator set in real time from the power grid management system, and the load demand data includes the load demand, fluctuation range and predicted load curve of each generator set; obtain the power output data of the generator set in real time from the generator set control system, and the power output data includes the current output power, upper and lower limits of rated power, start and stop status, start and stop time and operating efficiency of each generator set; obtain the transmission line status data of the transmission line in real time from the transmission line monitoring system, and the transmission line status data includes the capacity limit, current load rate, voltage level and line loss of each transmission line.

[0214] In some embodiments, the model building module 200 includes a data set acquisition module 201 , a preprocessing module 202 and a building module 203 .

[0215] The data set acquisition module 201 is used to establish a variable data set according to the power output data, load demand data and transmission line status data of the generator set, wherein the variable data set is represented as:

[0216] X={P i,t ,U i,t ,D t ,L j,t ∣i∈[1,N],j∈[1,M],t∈[1,T]}

[0217] Among them, P i,t represents the power output of the ith generator set at time t, U i,t represents the start and stop status of the generator set at time t, D t represents the load demand of the power grid system at time t, L j,t represents the load rate of transmission line j at time t, N is the total number of generator sets in the power grid system, M is the total number of transmission lines in the power grid system, and T represents the number of moments in the dispatching cycle when collecting key variable data;

[0218] The preprocessing module 202 is used to preprocess the variable data set and generate a variable matrix according to the preprocessed variable data set, wherein the variable matrix contains comprehensive information of the generator set status, load demand and transmission line status;

[0219] The building module 203 is used to build a mixed integer nonlinear programming problem model according to the variable matrix.

[0220] In some embodiments, the model building module 200 includes not only a data set acquisition module 201, a preprocessing module 202 and a building module 203, but also a determination module 204. The determination module 204 is used to determine the objective function and the constraint conditions according to the power output data, load demand data and transmission line status data of the generator set. The building module 203 is also used to establish a mixed integer nonlinear programming problem model according to the objective function and the constraint conditions.

[0221] Among them, the objective function is the total operating cost of the power grid system during the scheduling period; the constraints include system load demand constraints, upper and lower limit constraints on generator set power output, start and stop time constraints on generator sets, and capacity constraints on the transmission lines.

[0222] The total operating cost is expressed as:

[0223]

[0224] Among them, P i,t is the output power of generator set i at time t, U i,t is the start and stop status of the unit at time t, C i,t (P i,t ) is the power generation cost function of generator set i at time t, S i,t (U i,t ,U i,t-1 ) is the start-stop cost function of generator set i at time t; N is the total number of generator sets in the power grid system, and T represents the number of moments in the dispatching period when key variable data are collected;

[0225] The system load demand constraint is:

[0226]

[0227] The upper and lower limits of the power output of the generator set are:

[0228]

[0229] Among them, P i,min and P i,max are the minimum power and maximum power limits of generator set i respectively;

[0230] The start and stop time constraints of the generator set are:

[0231] U i,t -U i,t-1 ≥T i,start ×(1-U i,t-1 )-T i,stop ×U i,t-1 ;

[0232] Among them, T i,start and T i,stop are the minimum start time and minimum stop time of generator set i respectively;

[0233] The capacity constraint of the transmission line is:

[0234]

[0235] Among them, L j,t represents the load rate of transmission line j at time t, L j,max is the maximum capacity of line j.

[0236] In some embodiments, the adaptive optimization module 400 includes a space establishment module 401 , a training module 402 , and an optimization module 403 .

[0237] Among them, the space establishment module 401 is specifically used to input the intermediate processing results into the initial hybrid deep reinforcement learning model, so that the initial hybrid deep reinforcement learning model establishes a state space and an action space according to the key variable data of the power grid system, wherein the state space includes the current generator set output power, load demand, and transmission line status, and the action space includes the start and stop operations and power adjustment of the generator set.

[0238] The training module 402 is used to train the initial hybrid deep reinforcement learning model based on the cumulative reward function to optimize the strategy of the state space and action space.

[0239] The optimization module 403 is used to determine the scheduling plan of the unit according to the state space and action space after the strategy optimization.

[0240] Wherein, the cumulative reward function is:

[0241]

[0242] Among them, R t is the cumulative reward at time t, γ is the discount factor, r(s k ,a k ) is in state s k Next, perform action a k Instant rewards received.

[0243] In some embodiments, the training module 402 is also specifically used for: in the process of training the initial hybrid deep reinforcement learning model based on the cumulative reward function to perform strategy optimization on the state space and the action space, receiving real-time operation data from the power grid system in real time, the real-time operation data including changes in load demand, changes in generator unit status, and updates to transmission line status; dynamically adjusting the distribution of the state space and the action space according to the real-time operation data to adapt to the current state of the power grid system.

[0244] In some embodiments, the adaptive optimization module 400 includes not only a space establishment module 401, a training module 402 and an optimization module 403, but also a calculation module 404. The calculation module 404 is specifically used to: determine the operating cost, scheduling stability and system security as multiple game parties; use the combinatorial optimization and game theory model to calculate the utility function of the generator set i, and the utility function is expressed as:

[0245] U i (x) = α i C(x)+β i S(x)+γ i T(x)

[0246] Among them, C(x) is the operating cost, S(x) is the scheduling stability, T(x) is the system security, α i , β i and γ i are the weight coefficients of the operating cost, dispatching stability and system safety of generator set i respectively; the utility function of generator set i is solved by the Pareto optimal solution set of the game theory model to balance the relationship between different game parties in generator set i; the instant reward is determined according to the solution result of the utility function of each generator set.

[0247] In some embodiments, the determination device includes not only a data receiving module 100, a model building module 200, a constraint screening module 300 and an adaptive optimization module 400, but also a scheduling module 500. The scheduling module 500 is used to schedule the units in the power grid system according to the determined scheduling scheme of the units.

[0248] It can be understood that the specific details of the determination device provided in the embodiment of the present disclosure can refer to the specific details of the determination method embodiment of the first aspect mentioned above, and will not be repeated here.

[0249] In a third aspect, an embodiment of the present disclosure also provides an electronic device.

[0250] Figure 6 FIG. 2 is a structural block diagram of an electronic device 20 provided in an embodiment of the present disclosure. Figure 6 As shown, the electronic device 20 includes a memory 62, a processor 61, and a computer program stored in the memory 62 and executable on the processor 61. When the processor 61 executes the computer program, any method described in the first aspect is implemented.

[0251] In some embodiments, the electronic device 20 may further include a computer storage device 63 for storing a computer program of any of the methods described in the first aspect.

[0252] In a fourth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program of the method described in any one of the first aspects above.

[0253] Among them, the computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable red-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, disk or optical disk.

[0254] In a fifth aspect, an embodiment of the present disclosure further provides a computer program product, comprising a computer program / instructions, which, when executed by a processor, implement the steps of any one of the methods described in the first aspect above.

[0255] Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0256] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0257] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0258] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0259] The present disclosure uses specific embodiments to illustrate the principles and implementation methods of the present disclosure. The description of the above embodiments is only used to help understand the method of the present disclosure and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present disclosure, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present disclosure.

Claims

1. A method for determining a unit dispatching plan in a power grid system, characterized in that: The method comprises: Acquiring key variable data of the power grid system, wherein the key variable data includes power output data of the generator set, load demand data and transmission line status data; Establishing a mixed integer nonlinear programming problem model according to the power output data of the generator set, the load demand data and the transmission line status data; Performing hierarchical processing on the mixed integer nonlinear programming problem model based on constraint types to obtain an intermediate processing result; The intermediate processing results are calculated according to an adaptive optimization algorithm to determine a scheduling plan for the generator set.

2. The method according to claim 1, characterized in that The mixed integer nonlinear programming problem model is subjected to hierarchical processing based on the constraint type to obtain an intermediate processing result, including: Processing the mixed integer nonlinear programming problem model using convex hull constraint screening to obtain a first processing result; Processing the first processing result by Lagrangian relaxation screening to obtain a second processing result; The second processing result is processed by using a branch and bound and pruning method to obtain the intermediate processing result.

3. The method according to claim 1, characterized in that The power grid system includes a power grid management system, a generator set control system and a transmission line monitoring system. The key variable data of the power grid system is obtained, including: Acquire load demand data of the generator set from the power grid management system in real time, wherein the load demand data includes the load demand, fluctuation range and predicted load curve of each generator set; Acquire power output data of the generator set from the generator set control system in real time, the power output data including the current output power, upper and lower limits of rated power, start and stop status, start and stop time, and operating efficiency of each generator set; The transmission line status data of the transmission line is acquired in real time from the transmission line monitoring system, wherein the transmission line status data includes the capacity limit, current load rate, voltage level and line loss of each transmission line.

4. The method according to claim 3, characterized in that The method of establishing a mixed integer nonlinear programming problem model according to the power output data, load demand data and transmission line status data of the generator set comprises: A variable data set is established according to the power output data, load demand data and transmission line status data of the generator set, wherein the variable data set is expressed as: X={P i,t ,U i,t ,D t ,L j,t ∣i∈[1,N],j∈[1,M],t∈[1,T]} Among them, P i,t represents the power output of the ith generator set at time t, U i,t represents the start and stop status of the generator set at time t, D t represents the load demand of the power grid system at time t, L j,t represents the load rate of transmission line j at time t, N is the total number of generator sets in the power grid system, M is the total number of transmission lines in the power grid system, and T represents the number of moments in the dispatching cycle when collecting key variable data; Preprocessing the variable data set, and generating a variable matrix according to the preprocessed variable data set, wherein the variable matrix includes comprehensive information of the generator set status, load demand and transmission line status; A mixed integer nonlinear programming problem model is established according to the variable matrix.

5. The method according to claim 3, characterized in that: The method of establishing a mixed integer nonlinear programming problem model according to the power output data, load demand data and transmission line status data of the generator set comprises: Determining objective functions and constraints based on power output data, load demand data, and transmission line status data of the generator set; According to the objective function and constraints, a mixed integer nonlinear programming problem model is established; the objective function is the total operating cost of the power grid system during the dispatching period; the constraints include system load demand constraints, upper and lower limit constraints of generator set power output, start and stop time constraints of generator sets and capacity constraints of the transmission line; wherein, The total operating cost is expressed as: Among them, P i,t is the output power of generator set i at time t, U i,t is the start and stop status of the unit at time t, C i,t (P i,t ) is the power generation cost function of generator set i at time t, S i,t (U i,t ,U i,t-1 ) is the start-stop cost function of generator set i at time t; N is the total number of generator sets in the power grid system, and T represents the number of moments in the dispatching period when key variable data are collected; The system load demand constraint is: The upper and lower limits of the power output of the generator set are: Among them, P i,min and P i,max are the minimum power and maximum power limits of generator set i respectively; The start and stop time constraints of the generator set are: U i,t -U i,t-1 ≥T i,start ×(1-U i,t-1 )-T i,stop ×U i,t-1 ; Among them, T i,start and T i,stop are the minimum start time and minimum stop time of generator set i respectively; The capacity constraint of the transmission line is: Among them, L j,t represents the load rate of transmission line j at time t, L j,max is the maximum capacity of line j.

6. The method according to claim 1, characterized in that The step of calculating the intermediate processing result according to the adaptive optimization algorithm to determine the dispatching scheme of the generator set includes: Inputting the intermediate processing results into an initial hybrid deep reinforcement learning model, so that the initial hybrid deep reinforcement learning model establishes a state space and an action space according to the key variable data of the power grid system, wherein the state space includes the current output power of the generator set, load demand, and transmission line status, and the action space includes the start and stop operation and power adjustment of the generator set; Training the initial hybrid deep reinforcement learning model based on a cumulative reward function to optimize the strategy of the state space and the action space; According to the state space and action space after strategy optimization, the scheduling plan of the generator set is determined; wherein, The cumulative reward function is: Among them, R t is the cumulative reward at time t, γ is the discount factor, r(s k ,a k ) is in state s k Next, perform action a k Instant rewards received.

7. The method according to claim 6, characterized in that The calculating the intermediate processing result according to the adaptive optimization algorithm to determine the dispatching scheme of the generator set also includes: In the process of training the initial hybrid deep reinforcement learning model based on the cumulative reward function to optimize the strategy of the state space and the action space, real-time operation data from the power grid system is received in real time, wherein the real-time operation data includes load demand changes, generator unit state changes, and transmission line state updates; The distribution of the state space and the action space is dynamically adjusted according to the real-time operation data to adapt to the current state of the power grid system.

8. The method according to claim 6, characterized in that The calculating the intermediate processing result according to the adaptive optimization algorithm to determine the dispatching scheme of the generator set also includes: Identify operating costs, scheduling stability and system security as multiple game parties; The utility function of generator set i is calculated by using combinatorial optimization and game theory models. The utility function is expressed as: U i (x)=a i C(x)+β i S(x)+γ i T(x) Among them, C(x) is the operating cost, S(x) is the scheduling stability, T(x) is the system security, α i , β i and γ i are the weight coefficients of the operating cost, dispatching stability and system security of generator set i respectively; Solving the utility function of generator set i through the Pareto optimal solution set of the game theory model to balance the relationship between different game parties in generator set i; An immediate reward is determined based on the solution of the utility function for each generator set.

9. The method according to any one of claims 1 to 8, characterized in that: The method further comprises: The generator sets in the power grid system are dispatched according to the determined dispatching plan for the generator sets.

10. A device for determining a unit dispatching plan in a power grid system, characterized in that: The device comprises: A data receiving module, used to obtain key variable data of the power grid system, wherein the key variable data includes power output data of the generator set, load demand data and transmission line status data; A model building module, used to build a mixed integer nonlinear programming problem model according to the power output data, load demand data and transmission line status data of the generator set; A constraint screening module, used for performing hierarchical processing on the mixed integer nonlinear programming problem model based on constraint types to obtain an intermediate processing result; The adaptive optimization module is used to calculate the intermediate processing results according to the adaptive optimization algorithm to determine the scheduling plan of the generator set.

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 9 is implemented.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program for executing the method according to any one of claims 1 to 9.

13. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 9 are implemented.

Images