Low-carbon power grid dispatching method based on hybrid optimization algorithm

Abstract

The present invention provides a low-carbon power grid dispatching method based on a hybrid optimization algorithm, and the dispatching method is specifically as follows: based on power grid operation data and carbon emission targets, an optimization dispatching model is established; a preliminary solution is performed based on a dragonfly algorithm to obtain a preliminary global solution of the optimization dispatching model; the preliminary global solution is used as the starting point of an ant colony algorithm, and local optimization is performed through the ant colony algorithm; the quality of the current iterative solution of the ant colony algorithm is monitored in real time, and when the quality of the current iterative solution meets an interruption condition, the dragonfly algorithm is returned until the ant colony algorithm can iterate to meet the corresponding iteration end condition, and a local optimization result is output to obtain a power grid optimization dispatching plan. The present invention can use a hybrid algorithm composed of a dragonfly algorithm and an ant colony algorithm to process multi-objective optimization power grid dispatching, and ensure the convergence speed of the optimization iteration of the multi-optimization target power grid dispatching plan and the accuracy of the optimization result.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid dispatching, and in particular to a low-carbon power grid dispatching method based on a hybrid optimization algorithm. Background Art

[0002] The grid dispatching plan can efficiently coordinate the allocation and dispatching of power resources, and is the key to ensuring the stable operation of the power system and achieving efficient use of energy. Among them, the formulation and implementation of the grid dispatching plan are directly related to the power supply reliability and economy of the power system. In order to improve the overall performance of the power system, the optimization iterative algorithm is often used to optimize the grid dispatching plan to ensure that the power system can meet the various load requirements while minimizing operating losses and energy waste. And considering the large-scale access of new energy, the environmental protection of grid operation has gradually become one of the goals of the grid dispatching plan formulation, in order to achieve a dual improvement in economic and environmental benefits.

[0003] In the face of multi-objective optimization scenarios of power supply reliability, economy and environmental protection of complex power grids under large-scale access of new energy, the dragonfly algorithm is often used to optimize the formulation of power grid dispatching schemes in this scenario, so as to meet the multiple optimization goals of power grid dispatching optimization while adapting to the fluctuations brought by new energy. Although the dragonfly algorithm has good global search capabilities and fast convergence characteristics, it can adapt to multi-objective power grid dispatching optimization in complex scenarios. However, the dragonfly algorithm is relatively weak in local search capabilities. When faced with problems such as load balancing in power grid dispatching that require fine adjustment, it is difficult to find the optimal solution or local details close to the optimal solution. The accuracy of the obtained power grid dispatching scheme is low, and the optimization effect of the power grid dispatching is not good. It can be seen that when using the dragonfly algorithm to iterate the optimization of the power grid dispatching scheme, it is difficult to take into account both the convergence speed and the accuracy of the optimization results. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcoming in the prior art that it is difficult to balance the convergence speed and the accuracy of the optimization results when using the dragonfly algorithm to iterate the optimization of power grid dispatching schemes, and to provide a low-carbon power grid dispatching method based on a hybrid optimization algorithm. When using the dragonfly algorithm to perform multi-objective optimization of power grid dispatching that includes power grid operation and carbon emissions, the dragonfly algorithm is first used to perform a global search to quickly obtain the corresponding preliminary global solution, and then the ant colony algorithm with the advantages of self-organization, self-adaptation and distribution is used to perform local optimization on the obtained preliminary global solution. While ensuring the convergence speed of the iterative optimization of the power grid dispatching scheme, the accuracy of the optimization results is guaranteed.

[0005] The purpose of the present invention is achieved through the following technical solutions:

[0006] The low-carbon power grid dispatching method based on hybrid optimization algorithm includes:

[0007] Step 1: Collect grid operation data in real time and establish an optimization dispatch model based on grid operation data and carbon emission targets;

[0008] Step 2: Preliminary solution of the optimization scheduling model is performed based on the dragonfly algorithm to obtain a preliminary global solution of the optimization scheduling model;

[0009] Step 3: Using the preliminary global solution as the starting point of the ant colony algorithm, the preliminary global solution of the optimization scheduling model is locally optimized by the ant colony algorithm;

[0010] Step 4: During the iteration process of the ant colony algorithm, monitor the quality of the current iterative solution of the ant colony algorithm. If the quality of the current iterative solution meets the interruption condition, re-solve the optimization scheduling model through the dragonfly algorithm to obtain a new preliminary global solution, and return to step 3; if the quality of the current iterative solution does not meet the interruption condition, execute step 5;

[0011] Step 5: Continue to execute the ant colony algorithm. When the iteration end condition of the ant colony algorithm is met, output the local optimization result and obtain the power grid optimization scheduling plan.

[0012] Carbon emissions are also taken as a goal for power grid dispatch optimization, and the dragonfly algorithm is used to achieve multi-objective optimization of power grid dispatch, thereby achieving a dual improvement in the economic and environmental benefits of power grid operation. Specifically, the dragonfly algorithm is first used to perform multi-objective optimization of power grid dispatch that includes power grid operation and carbon emissions. The dragonfly algorithm's excellent global search capability and fast convergence characteristics are used to perform a global search to quickly obtain the corresponding preliminary global solution. And considering the dragonfly algorithm's lack of local search capabilities, after obtaining the preliminary global solution, the ant colony algorithm with the advantages of self-organization, self-adaptation, and distribution is used to perform local optimization on the preliminary global solution obtained, ensuring the convergence speed of the optimization iteration of the power grid dispatch plan while ensuring the accuracy of the optimization results.

[0013] Furthermore, the step 2 of performing a preliminary solution to the optimization scheduling model based on the dragonfly algorithm to obtain a preliminary global solution to the optimization scheduling model also includes:

[0014] According to the best fitness value of each iteration of the dragonfly algorithm, the fitness value change curve of the dragonfly algorithm is obtained;

[0015] The optimization node is selected according to the fitness value change curve, and the optimal solution of each iteration after the optimization node is recorded to construct a set of alternative global solutions.

[0016] The optimal fitness of the dragonfly algorithm is continuously monitored to monitor the quality changes of the global optimal solution. The optimal solution of each iteration is recorded and counted, and the corresponding set of candidate global solutions is constructed. The data in the candidate global solution set are all the better results obtained in the optimization iteration, which can provide a more reliable data basis for data variation in the local optimization process of the ant colony algorithm and improve the optimization accuracy of the ant colony algorithm.

[0017] Furthermore, in step 4, during the iteration process of the ant colony algorithm, the quality of the current solution of the ant colony algorithm is monitored, including:

[0018] Record the optimal fitness value after each iteration and the pheromone concentration of each path;

[0019] Calculate the degree of change of the best fitness of each iteration compared with the best fitness of the previous iteration;

[0020] When the number of consecutive occurrences of the change degree of the best fitness between the previous and next iterations being lower than the preset threshold exceeds the preset threshold, the pheromone concentration of each path in the corresponding iteration is combined to determine whether the quality of the current iteration solution meets the interruption condition;

[0021] When the number of consecutive occurrences of the degree of change of the best fitness between the two iterations is lower than the preset threshold does not exceed the preset threshold, it is determined that the quality of the current iterative solution does not meet the interruption condition, and step five is continued.

[0022] The degree of change of the best fitness of the ant colony algorithm is monitored in real time. If the best fitness has not been significantly improved in several consecutive iterations, it proves that the current ant colony algorithm may be trapped in the local optimum. However, the situation where the best fitness has not been significantly improved does not necessarily only exist when the ant colony algorithm is trapped in the optimum. It may also be that the preliminary global solution provided by the dragonfly algorithm is the result obtained after being trapped in the local optimum. Some local optimums of the ant colony algorithm can be processed by mutation processing. If it is the local optimum of the dragonfly algorithm, it is necessary to return to the dragonfly algorithm and re-obtain the preliminary global solution. Therefore, when the degree of change of the best fitness is monitored and it is found that there may be a situation of being trapped in the local optimum solution, the corresponding interruption condition is set, and then the pheromone concentration of the path in the ant colony algorithm is combined for identification, so as to determine whether it is necessary to return to the dragonfly algorithm and re-search the global solution.

[0023] Furthermore, judging whether the interruption condition is met in combination with the pheromone concentration of each path in the corresponding iteration includes:

[0024] Retrieve the pheromone concentration increase of each path in each iteration during the iteration in which the change degree of the corresponding best fitness is lower than the preset threshold;

[0025] Obtain the number of paths and corresponding paths in which the pheromone concentration increase exceeds a preset increase threshold in each iteration;

[0026] If the number of selected paths is lower than the number of the first paths, and the number of the same paths whose pheromone concentration exceeds the preset increase threshold in each iteration exceeds the number of the second paths, one of the alternative global solutions is randomly selected from the alternative global solutions, and the pheromones of some paths in the ant colony algorithm are randomly replaced;

[0027] According to the path after pheromone replacement, the ant colony algorithm performs a test iteration to obtain the degree of change between the best fitness of the test iteration and the best fitness value of the current iteration;

[0028] When the degree of change of the best fitness value of the test iteration compared with the best fitness value of the current iteration is lower than a preset threshold, it is judged that the quality of the current iteration solution meets the interruption condition;

[0029] When the degree of change of the best fitness value of the test iteration compared with the best fitness value of the current iteration is not less than a preset threshold, it is determined that the quality of the current iteration solution does not meet the interruption condition;

[0030] In other cases, the quality of the current iterative solution is judged to meet the interruption condition.

[0031] The path pheromone of the ant colony algorithm is the basis of its optimization iteration. The path pheromone is the data basis for obtaining the best fitness of the ant colony algorithm. In addition to the feature that the best fitness value has not significantly improved, the ant colony algorithm will also show that the population preferentially selects some of the paths, and only the pheromone of these paths has improved, and the pheromone of the remaining paths is basically unchanged, that is, the pheromone distribution is unbalanced. Therefore, when there is a feature that the best fitness has not significantly improved, the imbalance of pheromone distribution is identified, and when the imbalance of pheromone distribution is identified, one of the alternative global solutions is selected, and the ant colony algorithm is mutated according to the alternative global solution to deal with the problem of the ant colony algorithm falling into the local optimum. Considering that the situation that some ant colony algorithms fall into the local optimum cannot be handled by mutation processing, a test iteration is performed on the ant colony algorithm after mutation to determine whether the mutation processing can handle the current situation of falling into the local optimum. If the result of the test iteration shows that the best fitness has not changed much compared with before the mutation, it proves that the mutation processing cannot handle the current situation of falling into the local optimum, that is, the interruption condition is met. At this time, it is necessary to return to the dragonfly algorithm to re-search the global search and re-determine the preliminary global solution.

[0032] Furthermore, the step of continuing to execute the ant colony algorithm in step 5 further includes:

[0033] When the degree of change of the best fitness value of the test iteration is not less than the preset threshold value compared with the best fitness value of the current iteration, the iteration result of the test iteration replaces the current iteration solution and the ant colony algorithm continues to be executed.

[0034] For the local optimal situation that can be handled by mutation processing, the ant colony algorithm is continued on the basis of test iteration to follow the iteration end condition of the ant colony algorithm, while ensuring the search efficiency and accuracy of the ant colony algorithm.

[0035] Furthermore, the optimization scheduling model is re-solved by the dragonfly algorithm described in step 4 to obtain a new preliminary global solution, including:

[0036] Starting from the optimization node, the initial parameters of the dragonfly algorithm are set according to the first optimal solution recorded, and the optimization scheduling model is re-solved to obtain a new preliminary global solution.

[0037] The optimized node is the node with the largest change selected according to the change of the best fitness value of the dragonfly algorithm. This node has the greatest improvement potential and the fastest progress speed in the optimization process. Therefore, using this node as the starting node for re-solving the preliminary global solution can effectively improve the efficiency of the dragonfly algorithm in global search and reduce the amount of computational iteration of the dragonfly algorithm.

[0038] Furthermore, in step 1, based on the grid operation data and carbon emission targets, an optimization dispatch model is established, including:

[0039] Based on the grid operation data and carbon emission targets, an optimization scheduling model is established with minimizing carbon emissions, minimizing power generation costs and maximizing renewable energy utilization as objective functions and power balance constraints as constraints.

[0040] In the objective function, the two optimization objectives of power generation cost and carbon emissions are taken into consideration at the same time, achieving a good balance between low-carbon scheduling of power grid operation and economic benefits.

[0041] Furthermore, the objective function is expressed as:

[0042] ;

[0043] Where C is the objective function value, is the total carbon emissions of power generation equipment in the power grid, is the total power generation cost of the power grid power generation equipment, is the utilization rate of renewable energy, , and are the weight coefficients of total carbon emissions, power generation costs and renewable energy utilization rate, For the The power generation capacity of each power generation device, For the The carbon emission coefficient of each power generation equipment, is the number of power generation equipment, For the The unit power generation cost of a power generation device is is the power generated by renewable energy, is the total power generation of the grid.

[0044] Furthermore, the constraint condition is expressed as:

[0045] ;

[0046] in, is the total power generation of the grid, is the power of the energy storage device, is the load demand of the power grid, is the transmission loss of the power grid.

[0047] Furthermore, it also includes:

[0048] When executing the grid optimization dispatching plan, the grid operation data is updated in real time, and the updated grid operation data is periodically input into the hybrid optimization algorithm composed of the dragonfly algorithm and the ant colony algorithm to adjust the grid optimization dispatching plan.

[0049] A real-time feedback mechanism for the algorithm has been set up to adapt to the dynamic changes in load demand and renewable energy power generation in the power grid, so as to ensure that in the event of emergencies such as changes in power grid load and equipment failures, it can respond quickly, adjust the power grid dispatch plan, and ensure the normal operation of the power grid.

[0050] The beneficial effects of the present invention are:

[0051] (1) Carbon emissions are also taken as the goal of power grid dispatch optimization, and the dragonfly algorithm is used to achieve multi-objective optimization of power grid dispatch, so as to achieve a dual improvement in the economic and environmental benefits of power grid operation. Specifically, the dragonfly algorithm is first used to perform multi-objective optimization of power grid dispatch that includes power grid operation and carbon emissions. The dragonfly algorithm's excellent global search ability and fast convergence characteristics are used to perform a global search to quickly obtain the corresponding preliminary global solution. In addition, considering the dragonfly algorithm's lack of local search ability, after obtaining the preliminary global solution, the ant colony algorithm with the advantages of self-organization, self-adaptation and distribution is used to perform local optimization on the preliminary global solution obtained, ensuring the convergence speed of the optimization iteration of the power grid dispatch plan while ensuring the accuracy of the optimization result.

[0052] (2) In the process of local optimization using the ant colony algorithm, the quality of the optimal solution obtained is monitored in real time, so that when the algorithm falls into a local optimal situation, it can make timely adjustments to escape from the local optimal situation and ensure the accuracy of the optimization results.

[0053] (3) In the case of being trapped in a local optimum, the specific situation of being trapped in the local optimum is identified according to the operation of the ant colony algorithm, so as to choose to jump out of the layout optimum by mutation processing or returning to the dragonfly algorithm to re-search the global search. And only when the ant colony algorithm cannot handle the situation of being trapped in the local optimum, will it return to the dragonfly algorithm to re-search the global search, so as to reduce repeated calculations and improve the optimization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a schematic diagram of a process of the present invention;

[0055] Figure 2 It is a schematic diagram of the basic structure of a large-scale power grid system in a city according to an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0057] Example:

[0058] Low-carbon power grid dispatching methods based on hybrid optimization algorithms, such as Figure 1 As shown, including:

[0059] Step 1: Collect grid operation data in real time and establish an optimization dispatch model based on grid operation data and carbon emission targets;

[0060] Step 2: Preliminary solution of the optimization scheduling model is performed based on the dragonfly algorithm to obtain a preliminary global solution of the optimization scheduling model;

[0061] Step 3: Using the preliminary global solution as the starting point of the ant colony algorithm, the preliminary global solution of the optimization scheduling model is locally optimized by the ant colony algorithm;

[0062] Step 4: During the iteration process of the ant colony algorithm, monitor the quality of the current iterative solution of the ant colony algorithm. If the quality of the current iterative solution meets the interruption condition, re-solve the optimization scheduling model through the dragonfly algorithm to obtain a new preliminary global solution, and return to step 3; if the quality of the current iterative solution does not meet the interruption condition, execute step 5;

[0063] Step 5: Continue to execute the ant colony algorithm. When the iteration end condition of the ant colony algorithm is met, output the local optimization result and obtain the power grid optimization scheduling plan.

[0064] With the construction and development of new power systems, the proportion of renewable energy in power systems continues to increase. Renewable energy has the advantages of being clean, low-carbon and renewable. Therefore, when formulating power grid dispatching plans, in order to promote the consumption and utilization of renewable energy, carbon emission targets are also taken as one of the goals of power grid dispatching, and corresponding optimization dispatching models are established to improve the utilization rate of renewable energy while reducing coal consumption and carbon emissions of conventional thermal power generation, thereby achieving a dual improvement in the economic and environmental benefits of power grid operation.

[0065] Among them, the grid operation data includes power generation equipment data, load data, energy storage data and meteorological data of the environment. Power generation equipment data includes the power generation, operation status, carbon emission coefficient and other data of various power generation equipment such as traditional thermal power generation, natural gas power generation, wind power, photovoltaic power generation, etc. Load data includes load demand data in the power grid, which can reflect the load changes in the power grid at different time periods. Energy storage data includes data such as the capacity, charge and discharge status and power of energy storage equipment in the power grid. Meteorological data includes meteorological data used to predict wind power, photovoltaic and other renewable energy power generation, such as wind speed, solar radiation intensity, etc.

[0066] Based on the collected grid operation data, an optimization dispatch model is established based on the grid operation data and carbon emission targets, including:

[0067] Based on the grid operation data and carbon emission targets, an optimization scheduling model is established with minimizing carbon emissions, minimizing power generation costs and maximizing renewable energy utilization as objective functions and power balance constraints as constraints.

[0068] The objective function includes two major parts: economic objectives and environmental objectives. The economic objectives are reflected by minimizing the cost of power generation, while the environmental objectives are reflected by minimizing carbon emissions and maximizing the utilization of renewable energy.

[0069] In this embodiment, the objective function is constructed with minimizing carbon emissions as the main optimization goal, minimizing power generation costs and maximizing renewable energy utilization as auxiliary optimization goals. In order to distinguish the importance of different optimization goals, a corresponding weight coefficient is set for each optimization goal in the objective function.

[0070] Specifically, the expression of the constructed objective function is:

[0071] ;

[0072] Where C is the objective function value, is the total carbon emissions of power generation equipment in the power grid, is the total power generation cost of the power grid power generation equipment, is the utilization rate of renewable energy, , and are weight coefficients of total carbon emissions, power generation costs, and renewable energy utilization rate. , and are 0.5, 0.25 and 0.25 respectively, For the The power generation capacity of each power generation device, For the The carbon emission coefficient of each power generation equipment, is the number of power generation equipment, For the The unit power generation cost of a power generation device is is the power generated by renewable energy, is the total power generation of the grid.

[0073] The constraint condition is expressed as:

[0074] ;

[0075] in, is the total power generation of the grid, is the power of the energy storage device, is the load demand of the power grid, is the transmission loss of the power grid.

[0076] In addition to the power balance constraints mentioned above, there are also power generation equipment output constraints, load demand constraints, line transmission capacity constraints, etc., which can be set according to the operating conditions of each device in the power grid.

[0077] Taking into account the dual goals of power grid operation economy and environmental protection when formulating power grid dispatching plans, the dragonfly algorithm is used as a solution to the multi-objective optimization dispatching model.

[0078] The dragonfly algorithm simulates the movement of dragonfly groups in space and uses their predation and avoidance behaviors to find the optimal solution. This algorithm not only focuses on individual behavior, but also emphasizes collaboration and information sharing among groups, so that it can effectively perform global search in a larger search space. In multi-objective optimization problems, global search capabilities are crucial, helping the algorithm find multiple potential Pareto optimal solutions, that is, a solution that cannot be improved by any changes without compromising other objectives, rather than being limited to local optimality.

[0079] The specific process of using the dragonfly algorithm to preliminarily solve the optimization scheduling model is as follows:

[0080] Initialize the dragonfly algorithm and set the parameters of the dragonfly algorithm, including the number of dragonflies N and the maximum number of iterations and fitness improvement threshold , and define the boundary and constraint conditions of the solution space. And the boundary and constraint conditions of the solution space are specifically defined by the constraint conditions of the set optimization scheduling model.

[0081] After completing the parameter setting, N dragonfly individuals are randomly generated, represented by (x, y), and each individual represents a potential solution.

[0082] In each iteration, the position of the individual is updated by simulating the dragonfly's predation and avoidance behavior. The new position of the dragonfly individual is calculated by the following formula:

[0083] ;

[0084] in, is the new position of the jth dragonfly individual, is the current position of the jth dragonfly individual, and is the weight factor, is the position of the individual with the highest current fitness, is the position of the individual with the lowest current fitness.

[0085] The weighting factors mentioned above and , weight factor It reflects the dragonfly individual's movement towards the current optimal solution. The trend and degree of approach determine the step size or speed of the dragonfly individual moving towards the optimal solution in the search space, and the weight factor It reflects the behavior of dragonfly individuals to avoid collisions with other individuals or falling into local optimal solutions, and determines the extent to which dragonfly individuals stay away from certain individuals in the search space in order to maintain diversity.

[0086] The values ​​of the above two weight factors jointly determine the flight direction and step length of the dragonfly individual in the search space, thereby affecting the search strategy and performance of the algorithm. This embodiment sets a larger weight coefficient based on the global search function of the dragonfly algorithm. and The specific value can be adjusted according to the actual operation of the algorithm.

[0087] Calculate the fitness value of each dragonfly individual , and record the current optimal solution Its fitness Among them, the fitness value That is the objective function value of the optimization scheduling model.

[0088] If the improvement of the fitness value of the dragonfly algorithm is less than the fitness improvement threshold in consecutive K iterations , then stop the dragonfly algorithm and return the current optimal solution The record is the preliminary global solution. The number of iterations K can be used to improve the threshold according to computing resources, time requirements and fitness. It can be set based on factors such as size.

[0089] In determining the optimal solution of the Dragonfly algorithm Then, the optimal solution , that is, the initial global solution is used as the starting point of the ant colony algorithm for local optimization.

[0090] At the same time, during the iteration process of the dragonfly algorithm, the iteration results of the dragonfly algorithm are also monitored in real time, and the corresponding candidate global solution set is constructed to provide data integration for the subsequent optimization processing of the ant colony algorithm. The specific steps include:

[0091] According to the best fitness value of each iteration of the dragonfly algorithm, the fitness value change curve of the dragonfly algorithm is obtained;

[0092] The optimization node is selected according to the fitness value change curve, and the optimal solution of each iteration after the optimization node is recorded to construct a set of alternative global solutions.

[0093] When selecting the optimization node according to the fitness value change curve, the slope of each point on the fitness change curve is calculated, and the point with the largest slope is determined according to the slope calculation result, which is the optimization node. This optimization node is the node with the largest change in the best fitness value of the Dragonfly algorithm. This node has the greatest improvement potential and the fastest progress speed during the optimization process.

[0094] The optimal solution after optimizing the nodes is the better solution found by the dragonfly algorithm. This part is used as the data basis for the subsequent ant colony algorithm to perform data mutation and other processing in the local optimization process, which can further improve the optimization accuracy of the ant colony algorithm.

[0095] Taking the preliminary global solution as the starting point of the ant colony algorithm, the specific process of local optimization is as follows:

[0096] Initialize the parameters of the ant colony algorithm, including the number of ants M, the pheromone volatility rate , heuristic information coefficient and pheromone importance coefficient .

[0097] And each ant from At the beginning, the next path is selected based on the pheromone concentration and heuristic information to construct a new solution.

[0098] The probability of selecting a path is calculated as follows:

[0099] ;

[0100] Among them, n, m and k are the nodes set in the ant colony algorithm, and N is the node set set in the ant colony algorithm; is the probability of an ant transferring from node n to node m; is the pheromone concentration on the path from node n to node m, is the pheromone concentration on the path from node n to node k; and They are all heuristic factors, which can be set to the reciprocal of the distance from node n to node m and the reciprocal of the distance from node n to node k.

[0101] After each ant constructs a solution, its fitness value is calculated , and update the pheromone concentration on the path according to the fitness:

[0102] ;

[0103] in, is the pheromone concentration on the path from node n to node m after update; is the increase in pheromone concentration determined by the fitness value.

[0104] The increase in pheromone concentration represents the total amount of pheromones released by all ants passing through the path in this iteration. This increase is related to the quality of the path, and the fitness value is an indicator directly used to evaluate the quality of the solution. The higher the fitness value, the better the quality of the solution, that is, the higher the path quality. Therefore, the increase in pheromone concentration can be determined based on the change in the fitness value.

[0105] In this embodiment, the ant colony algorithm also uses the objective function value of the optimization scheduling model as the fitness value.

[0106] During the iteration process of the ant algorithm, the quality of the current solution is also monitored in real time to promptly identify situations where the algorithm falls into the local optimal state, thus ensuring the optimization accuracy of the algorithm.

[0107] Specifically, during the iteration process of the ant colony algorithm, monitoring the quality of the current solution of the ant colony algorithm includes:

[0108] Record the optimal fitness value after each iteration and the pheromone concentration of each path;

[0109] Calculate the degree of change of the best fitness of each iteration compared with the best fitness of the previous iteration;

[0110] When the number of consecutive occurrences of the change degree of the best fitness between the previous and next iterations being lower than the preset threshold exceeds the preset threshold, the pheromone concentration of each path in the corresponding iteration is combined to determine whether the quality of the current iteration solution meets the interruption condition;

[0111] When the number of consecutive occurrences of the degree of change of the best fitness between the two iterations is lower than the preset threshold does not exceed the preset threshold, it is determined that the quality of the current iterative solution does not meet the interruption condition, and step five is continued.

[0112] When the best fitness value does not change significantly for consecutive times, it proves that there may be a situation of falling into a local optimal solution. However, there are two situations of falling into a local optimal solution. One is that the ant colony algorithm falls into a local optimal solution during the iteration process, and the other is that the dragonfly algorithm has fallen into a local optimal solution when providing a preliminary global solution. When the ant colony algorithm is locally optimized, the fitness value will not be significantly improved.

[0113] There are two situations when the ant colony algorithm falls into a local optimum. One is that it can jump out of the local optimum through data processing such as mutation. The other is that the data processing strategy is not sufficient to support jumping out of the local optimum and needs to return to re-optimize and iterate.

[0114] Therefore, for the various situations mentioned above, when it is identified that the change degree of the best fitness between the two iterations is lower than the preset threshold and the number of consecutive occurrences exceeds the preset threshold, the current ant colony algorithm iteration result is not directly processed by data processing or returned to re-optimize the iteration to jump out of the local optimal process, but first combine the path pheromone situation of the ant colony algorithm to determine whether the interruption condition is currently met, and then select the processing method according to the satisfaction of the interruption condition.

[0115] When the number of consecutive occurrences does not exceed the threshold, it can be determined that the local optimum has not occurred. It can be directly concluded that the interruption condition is not met. No data processing is required, and the ant colony algorithm can continue to iterate.

[0116] The step of judging whether the interruption condition is satisfied by combining the pheromone concentration of each path in the corresponding iteration includes:

[0117] Retrieve the pheromone concentration increase of each path in each iteration during the iteration in which the change degree of the corresponding best fitness is lower than the preset threshold;

[0118] Obtain the number of paths and corresponding paths in which the pheromone concentration increase exceeds a preset increase threshold in each iteration;

[0119] If the number of selected paths is lower than the number of the first paths, and the number of the same paths whose pheromone concentration exceeds the preset increase threshold in each iteration exceeds the number of the second paths, one of the alternative global solutions is randomly selected from the alternative global solutions, and the pheromones of some paths in the ant colony algorithm are randomly replaced;

[0120] According to the path after pheromone replacement, the ant colony algorithm performs a test iteration to obtain the degree of change between the best fitness of the test iteration and the best fitness value of the current iteration;

[0121] When the degree of change of the best fitness value of the test iteration compared with the best fitness value of the current iteration is lower than a preset threshold, it is judged that the quality of the current iteration solution meets the interruption condition;

[0122] When the degree of change of the best fitness value of the test iteration compared with the best fitness value of the current iteration is not less than a preset threshold, it is determined that the quality of the current iteration solution does not meet the interruption condition;

[0123] In other cases, the quality of the current iterative solution is judged to meet the interruption condition.

[0124] The characteristics of the ant colony algorithm being stuck in a local optimal solution mainly include stagnation of solution quality and imbalance in pheromone distribution. The number of consecutive occurrences in which the change in the best fitness between two iterations is lower than the preset threshold can identify the characteristic of stagnation of solution quality. However, this characteristic of being stuck in a local optimal solution is not limited to the ant colony algorithm. The dragonfly algorithm also exhibits this characteristic when it is stuck in a local optimal solution.

[0125] The pheromone distribution imbalance is based on the fact that the ant colony algorithm accumulates a large amount of pheromones on the path of the better solution, while the pheromones on other paths are relatively less, which causes ants to be more inclined to choose the path they have already traveled. Specifically, the pheromone distribution imbalance can be identified through the distribution of the ant population. If the ant population gradually concentrates on a few solutions during the iteration process, it can be determined that there is a pheromone distribution imbalance.

[0126] Therefore, when a situation that may have fallen into a local optimum is identified, the increase in pheromone concentration of each path in several iterations where the optimal fitness has not been significantly improved is further extracted to identify whether there is an imbalance in the pheromone distribution.

[0127] Among them, the number of paths whose pheromone concentration increase exceeds the preset increase threshold in each iteration reflects the number of candidate solutions of the ant colony algorithm in this iteration. Then, the overlap of paths whose pheromone concentration increase exceeds the preset increase threshold is identified. If the pheromone concentration increases only on the same path almost every time, it can be determined that there is an imbalance in the pheromone distribution.

[0128] The preset increase threshold of the pheromone concentration increase, the number of first paths and the number of second paths can be set according to the initialized ant colony algorithm parameters, the actual operation of the algorithm and the requirements.

[0129] When the imbalanced distribution of pheromones is identified, it is determined that the ant colony algorithm is currently trapped in a local optimum.

[0130] When the ant colony algorithm falls into the local optimum, we try to deal with it through mutation processing.

[0131] However, since it is impossible to confirm the effect of mutation processing on being trapped in the local optimum, in order to avoid unnecessary calculations, a test iteration is performed on the mutation processing. If the mutation processing is effective for the situation of being trapped in the local optimum, the optimal fitness obtained in the test iteration should be better than the optimal fitness before the mutation processing, and the degree of improvement in the optimal fitness should be greater than the preset threshold. If the improvement effect is not obvious, that is, the degree of improvement in the optimal fitness is less than or equal to the preset threshold, or the optimal fitness is lower than before the mutation processing, it proves that the effect of the mutation processing is poor. At this time, it can be judged that the mutation processing cannot handle the current situation of being trapped in the local optimum, and the interruption condition is met. It is necessary to return to the dragonfly algorithm and re-search the global search.

[0132] In addition, if the number of screened paths is lower than the number of the first paths, and the number of the same paths whose pheromone concentration exceeds the preset increase threshold in each iteration exceeds the number of the second paths, that is, it is identified that there is no pheromone distribution imbalance, then it proves that the dragonfly algorithm is currently trapped in a local optimal situation and the interruption condition is met, and it is necessary to return to the dragonfly algorithm and re-perform a global search.

[0133] When the degree of change of the best fitness value of the test iteration is not less than the preset threshold compared with the best fitness value of the current iteration, it is proved that the mutation processing can effectively deal with the current situation of being trapped in the local optimum. The mutation processing can jump out of the local optimum and continue to execute the optimization iteration of the ant colony algorithm.

[0134] If it is proved that the ant colony algorithm can achieve the effect of jumping out of the local optimum, considering that the test iteration is also an iteration of the ant colony algorithm, in order not to affect the determination of the iteration end condition of the ant colony algorithm, the iteration result that has a trend of jumping out of the local optimum after the test iteration will directly replace the current solution, and then continue to execute the ant colony algorithm.

[0135] For those that meet the interruption conditions, it is necessary to return to the dragonfly algorithm and re-perform global search to obtain the preliminary global solution again.

[0136] In order to optimize the efficiency of re-global search and reduce the amount of calculation, the optimization scheduling model is re-solved by the Dragonfly algorithm to obtain a new preliminary global solution. Starting from the optimization node, the initial parameters of the Dragonfly algorithm are set according to the first optimal solution recorded, and the optimization scheduling model is re-solved to obtain a new preliminary global solution.

[0137] The optimized node is selected based on the change in the best fitness in the process of obtaining the preliminary global solution by the dragonfly algorithm. It is the node with the largest change in the best fitness value of the dragonfly algorithm. The best fitness obtained in each iteration after the optimization node is the best solution. Therefore, the optimized node is used as the starting point for the dragonfly algorithm to re-search the global solution, and the initial parameters of the dragonfly algorithm are reset, including the maximum number of iterations and the position of each dragonfly individual. Other parameters including the number of dragonflies and the fitness improvement threshold can be directly used as the previous parameters, or they can be adjusted according to needs. After completing the parameter setting, the dragonfly algorithm re-searches the global solution to obtain a new preliminary global solution, and then uses the new preliminary global solution as the starting point to perform local optimization through the ant colony algorithm until the iteration end condition of the ant colony algorithm is reached, and the local optimization result is output to obtain the power grid optimization scheduling plan.

[0138] The grid optimization scheduling scheme includes power allocation of various power generation equipment and charging and discharging strategies of energy storage equipment. Through the grid optimization scheduling scheme, carbon emissions can be minimized while meeting the grid load demand.

[0139] When applying the grid optimization and dispatching plan to the actual operation of the grid, the grid operation status is monitored in real time and dynamic adjustments are made. If there are changes in the grid status such as load changes or equipment failures during operation, or changes in load demand, the grid operation data can be updated to adjust the grid optimization and dispatching plan.

[0140] And considering that in the actual operation process, the collection of power grid operation data often has certain errors and delays, even if accurate data can be collected in real time, it takes a certain amount of time to analyze and process the collected data. If the power grid optimization scheduling plan is adjusted in real time, it will not be able to respond quickly to emergencies in the power grid.

[0141] At the same time, considering that frequent changes in the power grid optimization scheduling plan will not only increase the workload of the dispatchers, but may also affect the stability and reliability of the power grid, therefore, in the case where the power grid optimization scheduling plan needs to be adjusted, the updated power grid operation data is periodically input into the hybrid optimization algorithm composed of the dragonfly algorithm and the ant colony algorithm to adjust the power grid optimization scheduling plan.

[0142] On the other hand, this embodiment also provides a low-carbon power grid dispatching system based on a hybrid optimization algorithm. The dispatching system mainly includes three modules: data collection, data analysis and dispatch execution. The data collection module is connected to the power information collection system, remote data collection system and meteorological system of the power grid, and can retrieve power data and load data of each power station and energy storage equipment in the power grid recorded in the power information collection system, relevant data of the energy storage equipment recorded in the remote data collection system, and meteorological data given in the meteorological system.

[0143] After retrieving the corresponding data, the data collection module can pre-process the retrieved data, perform data cleaning and data transformation, etc., to provide basic data support for subsequent power grid dispatching.

[0144] The data analysis module has a built-in hybrid optimization algorithm consisting of the dragonfly algorithm and the ant colony algorithm, which can formulate a power grid optimization scheduling plan based on the relevant data processed by the data collection module.

[0145] The dispatching execution module can formulate relevant instructions according to the grid optimization dispatching plan, and issue the instructions to the corresponding equipment to realize grid dispatching.

[0146] Take a large-scale power grid system in a certain city as an example. The system has multiple power stations, energy storage systems and load centers. The power generation equipment in the power station includes traditional power generation such as thermal power generation and various renewable energy power generation such as hydropower and photovoltaic power generation. The load center is equipped with a corresponding low-carbon power grid dispatching system. The infrastructure of this large-scale power grid system is as follows: Figure 2 shown.

[0147] In this large-scale power grid environment, the power data and carbon emission coefficient of each power station and energy storage equipment are collected or predicted in real time, and an optimized scheduling model is established based on the complex power grid structure.

[0148] The optimization scheduling model is solved using a hybrid optimization algorithm consisting of the dragonfly algorithm and the ant colony algorithm. The dragonfly algorithm is first used for global search to obtain a preliminary scheduling plan, and then the ant algorithm is used for local optimization to fine-tune the power distribution among the power stations to obtain the final grid optimization scheduling plan. The load center then formulates the corresponding designated quality according to the grid optimization scheduling plan and sends it to the corresponding power station and energy storage system to ensure the dynamic balance between the output power of each power station in the large-scale power grid and the load demand.

[0149] When the city's electricity demand fluctuates, a hybrid optimization algorithm consisting of the dragonfly algorithm and the ant colony algorithm can be input to adjust the grid optimization scheduling plan to adjust the charging and discharging status of the energy storage system, prioritize reducing the load of high-carbon power stations, and ensure that carbon emissions in the scheduling process are minimized.

[0150] The above-described embodiment is only a preferred solution of the present invention and does not limit the present invention in any form. There are other variations and modifications without exceeding the technical solution described in the claims.

Claims

1. A low-carbon power grid dispatching method based on a hybrid optimization algorithm, characterized in that: include: Step 1: Collect grid operation data in real time and establish an optimization dispatch model based on grid operation data and carbon emission targets; Step 2: Preliminary solution of the optimization scheduling model is performed based on the dragonfly algorithm to obtain a preliminary global solution of the optimization scheduling model; Step 3: Using the preliminary global solution as the starting point of the ant colony algorithm, the preliminary global solution of the optimization scheduling model is locally optimized by the ant colony algorithm; Step 4: During the iteration process of the ant colony algorithm, monitor the quality of the current iterative solution of the ant colony algorithm. If the quality of the current iterative solution meets the interruption condition, re-solve the optimization scheduling model through the dragonfly algorithm to obtain a new preliminary global solution, and return to step 3; if the quality of the current iterative solution does not meet the interruption condition, execute step 5; Step 5: Continue to execute the ant colony algorithm, and when the iteration end condition of the ant colony algorithm is met, output the local optimization result and obtain the power grid optimization dispatching plan; In step 4, during the iteration process of the ant colony algorithm, the quality of the current iterative solution of the ant colony algorithm is monitored, including: Record the best fitness value after each iteration and the pheromone concentration of each path; Calculate the degree of change between the best fitness value of each iteration and the best fitness value of the previous iteration; When the number of consecutive occurrences of the change degree of the best fitness value between the two iterations being lower than the preset degree threshold exceeds the preset number threshold, the quality of the current iteration solution is judged whether it meets the interruption condition in combination with the pheromone concentration of each path in the corresponding iteration; Retrieving the pheromone concentration increase of each path in each iteration during the iteration process in which the change degree of the corresponding best fitness value is lower than the preset degree threshold; Obtain the number of paths and corresponding paths in which the pheromone concentration increase exceeds a preset increase threshold in each iteration; If the number of selected paths is lower than the number of the first paths, and the number of the same paths whose pheromone concentration increases by more than the preset increase threshold in each iteration exceeds the number of the second paths, one of the alternative global solutions is randomly selected from the set of alternative global solutions, and the pheromones of some paths in the ant colony algorithm are randomly replaced; The fitness value of the ant colony algorithm is the objective function value of the optimization scheduling model.

2. The low-carbon power grid dispatching method based on hybrid optimization algorithm according to claim 1 is characterized in that: The step 2 is to perform a preliminary solution to the optimization scheduling model based on the dragonfly algorithm to obtain a preliminary global solution to the optimization scheduling model, and further includes: According to the best fitness value of each iteration of the dragonfly algorithm, the fitness value change curve of the dragonfly algorithm is obtained; Select the optimization node according to the fitness value change curve, record the optimal solution of each iteration after the optimization node, and build a set of candidate global solutions; The fitness value of the dragonfly algorithm is the objective function value of the optimization scheduling model.

3. The low-carbon power grid dispatching method based on hybrid optimization algorithm according to claim 2 is characterized in that: In step 4, during the iteration process of the ant colony algorithm, the quality of the current iterative solution of the ant colony algorithm is monitored, and the method further includes: When the number of consecutive occurrences of the degree of change of the best fitness value between the two iterations being lower than the preset degree threshold does not exceed the preset number threshold, it is determined that the quality of the current iterative solution does not meet the interruption condition, and step five is continued.

4. The low-carbon power grid dispatching method based on hybrid optimization algorithm according to claim 1 is characterized in that: The step of judging whether the quality of the current iterative solution meets the interruption condition in combination with the pheromone concentration of each path in the corresponding iteration also includes: According to the path after pheromone replacement, the ant colony algorithm performs a test iteration to obtain the degree of change between the best fitness value of the test iteration and the best fitness value of the current iteration; When the degree of change of the best fitness value of the test iteration compared with the best fitness value of the current iteration is lower than the preset degree threshold, it is judged that the quality of the current iteration solution meets the interruption condition; When the degree of change of the best fitness value of the test iteration compared with the best fitness value of the current iteration is not less than a preset degree threshold, it is determined that the quality of the current iteration solution does not meet the interruption condition; In other cases, the quality of the current iterative solution is judged to meet the interruption condition.

5. The low-carbon power grid dispatching method based on hybrid optimization algorithm according to claim 4 is characterized in that: Continuing to execute the ant colony algorithm as described in step 5 also includes: When the degree of change of the best fitness value of the test iteration is not less than the preset degree threshold compared with the best fitness value of the current iteration, the iteration result of the test iteration replaces the current iteration solution and the ant colony algorithm continues to be executed.

6. The low-carbon power grid dispatching method based on hybrid optimization algorithm according to claim 2 is characterized in that: In step 4, the optimization scheduling model is re-solved by the dragonfly algorithm to obtain a new preliminary global solution, including: Starting from the optimization node, the initial parameters of the dragonfly algorithm are set according to the first optimal solution recorded, and the optimization scheduling model is re-solved to obtain a new preliminary global solution.

7. The low-carbon power grid dispatching method based on hybrid optimization algorithm according to claim 1 is characterized in that: In step 1, based on the grid operation data and carbon emission targets, an optimization dispatch model is established, including: Based on the grid operation data and carbon emission targets, an optimization scheduling model is established with minimizing the total carbon emissions, minimizing the total cost of power generation and maximizing the utilization of renewable energy as the objective functions and the power balance constraint as the constraint condition.

8. The low-carbon power grid dispatching method based on hybrid optimization algorithm according to claim 7 is characterized in that: The expression of the objective function is: ； Where C is the objective function value, is the total carbon emissions of power generation equipment in the power grid, is the total power generation cost of the power grid power generation equipment, is the utilization rate of renewable energy, , and are the weight coefficients of total carbon emissions, total power generation cost and renewable energy utilization rate, For the The power generation capacity of each power generation device, For the The carbon emission coefficient of each power generation equipment, is the number of power generation equipment, For the The unit power generation cost of a power generation device is is the power generated by renewable energy, is the total power generation of the grid.

9. The low-carbon power grid dispatching method based on hybrid optimization algorithm according to claim 7 is characterized in that: The constraint condition is expressed as: ； in, is the total power generation of the grid, is the power of the energy storage device, is the load demand of the power grid, is the transmission loss of the power grid.

10. The low-carbon power grid dispatching method based on hybrid optimization algorithm according to claim 1, characterized in that: Also includes: When executing the grid optimization dispatching plan, the grid operation data is updated in real time, and the updated grid operation data is periodically input into the hybrid optimization algorithm composed of the dragonfly algorithm and the ant colony algorithm to adjust the grid optimization dispatching plan.

Images