Micro-grid intra-day multi-stage rolling coordination control method and system

Abstract

The invention discloses a micro-grid intra-day multi-stage rolling coordination control method and system, and the method comprises the steps: operating a rolling coordination control method based on an established multi-stage micro-grid intra-day optimization model, reading a power prediction parameter, and reading particles to judge and select an optimal intra-day scheduling result; on the basis of the read non-dominated solution particle parameters of the optimal intra-day scheduling result, constructing a cycle rolling optimization initial particle, executing a short-term scheduling rolling optimization strategy, and judging whether a preset constraint condition is met or not according to a current rolling optimization result; and performing cyclic rolling optimization until a preset iteration condition is met, updating the optimal position of the individual, and outputting an optimal solution set. According to the method, multi-time-scale collaborative optimization operation is realized, the influence of wind power photovoltaic volatility and uncertainty on the system is reduced, and the utilization rate of renewable energy sources is improved while the operation state of equipment is optimized; and continuous approaching to the day-ahead planned output within the day can be accelerated, and the solving time is effectively shortened.

Description

Technical Field

[0001] This invention relates to the field of optimized economic dispatch technology for new energy power systems, and in particular to a method and system for intraday multi-stage rolling coordinated control of microgrids. Background Technology

[0002] In the field of wind power to hydrogen production, the economic dispatch of power systems is a multi-objective problem with nonlinear large-scale constraints. To obtain high-quality Pareto solutions, it is necessary to develop a more reasonable algorithm framework to improve the convergence and distribution of the frontier and provide dispatchers with scientific and reasonable strategy results. Microgrid-related research mainly focuses on the establishment of optimization models, design of solution algorithms, and power allocation strategies in dispatch rules. It often involves the combination of dispatching at different time scales. It is necessary to coordinate global and local energy management through the coordinated connection of multi-scale dispatching to address the deviation of power prediction at different scales.

[0003] The existing technical solution is based on fitting relevant curves to collected hydropower plant data, setting a rolling calculation cycle and actual calculation interval, reading in real-time hydrological data and the planned daily output of the units, and using the fitted curves to deduce the hydrological conditions for the next period. This process is repeated to obtain the water discharge flow for each period, and then corresponding water discharge warnings are issued based on the water discharge flow. After the warning is issued, the output of the corresponding units is optimized and corrected, and finally, the warning information and optimization correction strategy are output. However, since the mutual influence of the optimization process at different times is not considered, it belongs to a static scheduling mode. Therefore, there is no correlation between variables at different times during the static scheduling optimization process, resulting in poor overall anti-interference capability. When unreasonable scheduling connections occur, it will increase the number of equipment start-ups and shutdowns, reduce energy utilization, interfere with the voltage control effect of the microgrid, and thus affect the reliability of power supply. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a multi-stage rolling coordinated control method and system for microgrids during the day.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a multi-stage rolling coordinated control method for microgrids during the day, comprising: Based on the established multi-stage microgrid intraday optimization model, a rolling coordination control method is used to read power prediction parameters and select the optimal intraday scheduling result by reading particle judgment. Based on the non-dominated solution particle parameters of the optimal intraday scheduling result that has been read, construct the initial particle for cyclic round rolling optimization, execute the short-term scheduling rolling optimization strategy, and determine whether the preset constraints are met based on the current rolling optimization result; Perform cyclical optimization until the preset iteration conditions are met, update the optimal position of each individual and output the optimal solution set.

[0006] In some embodiments, it also includes: A multi-stage microgrid intraday optimization model is established based on the objective function, constraints, and decision variables. The multi-stage microgrid intraday optimization model includes a one-stage optimization model and a two-stage optimization model. The one-stage optimization model is used to transmit the output curves of each device to the two-stage model, and the two-stage optimization model is used to feed back the output process of each device to the one-stage model.

[0007] In some embodiments, the rolling coordination control method based on the established multi-stage microgrid intraday optimization model, reading power prediction parameters, and reading particle judgment to select the optimal intraday scheduling result includes: Based on the established multi-stage microgrid intraday optimization model, the equipment output curves and equipment operation information of the day-ahead scheduling 24 hours are read, and the power prediction parameters are read based on the optimal economic intraday scheduling algorithm. Based on the diversity of results from the non-dominated (Pareto) solution set, the particles are read cyclically, and the optimal intraday scheduling optimization result that satisfies the constraints is selected. If the constraints are met, the optimal intraday scheduling result will be used as the intraday scheduling constraint; otherwise, the next particle will be read and the decision will continue.

[0008] In some embodiments, the intraday scheduling constraints include an intraday scheduling equipment power range equal to the target daytime power multiplied by the relative adjustment amount, an intraday scheduling target output using the daytime scheduling equipment output curve as a reference, and interpolation to interpolate the scheduling results. The penalty cost is a weighted sum of the absolute values ​​of the deviations between the daytime target capacity and the intraday corrected capacity.

[0009] In some embodiments, the step of constructing initial particles for cyclic round rolling optimization based on the non-dominated solution particle parameters of the read optimal intraday scheduling results, executing a short-term scheduling rolling optimization strategy, and determining whether preset constraints are met based on the current rolling optimization results includes: Set the rolling calculation parameters, read the non-dominated solution particle parameters of the previous rolling optimization result, and construct the initial particles for the next rolling optimization. Execute a short-term scheduling rolling optimization strategy and determine whether the preset constraints are met based on the current rolling optimization result; If the current rolling optimization result meets the preset constraints, the equipment capacity value is updated, and the equipment output control command for the first moment of the current rolling cycle is issued; otherwise, the next rolling optimization cycle is executed.

[0010] In some embodiments, before each execution of the next round of rolling optimization, it is necessary to reread the equipment output curve and equipment operation information of the day-ahead scheduling 4 hours after the current time. If the current iteration count of the rolling cycle has not reached the preset number, continue to execute the next round of rolling optimization.

[0011] In some embodiments, the rolling coordination control method includes: Determine whether the parameters in the particle satisfy the power balance equation. If not, perform constraint processing on the power balance equation. Calculate the dominance relationships of all particles within the particle swarm and update the non-dominated solutions to the repository; Based on the crowding index, select the first leader within different feasible solution sets in the non-dominated solution set, and update the inertia factor and variation factor; and Select a second leader from the remaining particles in different feasible solution sets within the non-dominated solution set; In the remaining iterations, the random particles learn the second leader according to the same learning mechanism.

[0012] In some embodiments, the rolling coordination control method further includes: The initial iteration is performed based on the first leader, updating the particle position, velocity, and feasibility results, and performing constraint repair on the power balance equation and the capacity balance equation. Based on the feasibility and fitness values ​​of the current iterative population, valid solutions are collected, identified, and updated in the first repository. Based on the currently determined effective solutions, an elite population is formed and clonal proliferation and local mutation strategies are implemented. Pruning is performed on particles in the non-dominated solution set based on the crowding index, and non-dominated solutions with high crowding and similarity are deleted. The second leader is selected based on the current non-dominated solutions and the principle of minimizing the weighted sum of economic and penalty costs. Valid solutions are collected and updated in the second repository to combine and generate the final repository.

[0013] In some embodiments, the short-term scheduling rolling optimization strategy includes: Read the number of tanks started and stopped at the current time in the day-ahead scheduling solution, and based on the change value of the number of tanks started and stopped before and after the time, find the position of the change in the number of tanks started and stopped, and determine the adjustment range of the tanks. Set different position constraint ranges, randomly select positions based on the constraint ranges and discretize the deviations, determine the changes in the number of tank groups at different times based on the determined start and stop position offsets, and determine the power constraint ranges of different tank groups, so as to determine and modify the variable constraint ranges.

[0014] In a second aspect, the present invention also provides a microgrid intraday multi-stage rolling coordination control system for operating the microgrid intraday multi-stage rolling coordination control method as described in the first aspect, the control system comprising: The intraday optimization model building module is used to establish a multi-stage intraday optimization model for microgrids based on the objective function, constraints, and decision variables. The intraday optimization model solution module is used to run the rolling coordination control method, read the power prediction parameters, read the particles to determine and select the optimal intraday scheduling result, construct the initial particles for cyclic rolling optimization, execute the short-term scheduling rolling optimization strategy, determine whether the current rolling optimization result meets the preset constraints, perform cyclic rolling optimization until the preset iteration conditions are met, update the optimal position of the individual and output the optimal solution set.

[0015] The present invention has the following beneficial effects: 1. This invention addresses the problem of electrolyzer state combination linkage across different time scales. By analyzing the mutual influence of coupling variables at different times, it dynamically adjusts and optimizes the start-up and shutdown times of the electrolyzer, thereby achieving multi-time scale collaborative optimization operation. This reduces the impact of the volatility and uncertainty of wind and solar power on the system, optimizes equipment operation, and improves the utilization rate of renewable energy. 2. This invention adds a repair strategy with different equality and inequality constraints, and improves the two-stage particle swarm algorithm by adding an elite group cloning mutation mechanism and an initial solution generation strategy. This effectively enhances the global search capability of the algorithm. With the goal of minimizing the total output adjustment of the day-ahead and intraday scheduling equipment, it realizes multi-source output coordination and power allocation among electrolytic cell groups based on linear programming, thereby accelerating the continuous approach of intraday to the day-ahead planned output and effectively shortening the solution time. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the rolling coordination control method proposed in this invention. Figure 1 ; Figure 2 This is a flowchart illustrating the rolling coordination control method proposed in this invention. Figure 2 ; Figure 3 This is a flowchart illustrating the rolling coordination control method proposed in this invention. Figure 3 ; Figure 4 This is a flowchart illustrating the rolling coordination control method proposed in this invention. Figure 4 ; Figure 5 This is a flowchart illustrating the rolling coordination control method proposed in this invention. Figure 5 ; Figure 6 This is a schematic diagram illustrating the principle of the rolling coordination control method proposed in this invention; Figure 7 This is a flowchart illustrating the rolling coordination control method proposed in this invention. Figure 6 ; Figure 8A schematic diagram illustrating the connection between two time scales and the unit combination offset mechanism; Figure 9 This is a schematic diagram of the rolling coordination control system proposed in this invention; Figure 10 A comparative diagram showing the results of the improved PSO algorithm for the elite clone local search mechanism; Figure 11 This is a diagram illustrating the comparison of day-ahead scheduling capacity constraints. Figure 12 This is a schematic diagram comparing the 24-hour power prediction input from the daytime dispatch. Figure 13 This is a schematic diagram comparing the 4-hour power prediction input from the daytime dispatch. Figure 14 This is a diagram showing the comparison of capacity parameters for intraday and day-ahead scheduling results when the rolling number is 4. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] This application provides a multi-stage rolling coordinated control method and system for microgrids within a day. It solves the problem in existing static scheduling optimization processes where variables at different times lack correlation, resulting in poor overall anti-interference capability. Furthermore, unreasonable scheduling connections increase equipment start-up and shutdown frequency, reduce energy utilization, interfere with microgrid voltage control, and ultimately affect power supply reliability. This application enables multi-timescale coordinated optimization, reducing the impact of wind and solar power volatility and uncertainty on the system, optimizing equipment operating status while improving renewable energy utilization; it also accelerates the approach to the planned daily output, effectively shortening the solution time.

[0019] Please refer to the following examples for details: Reference Figures 1-14 The present invention provides an embodiment of a multi-stage rolling coordinated control method for microgrids during the day, the specific structure of which includes: S100 uses a rolling coordination control method based on the established multi-stage microgrid intraday optimization model to read power prediction parameters and select the optimal intraday scheduling result by reading particle judgment. S200: Based on the non-dominated solution particle parameters of the read optimal intraday scheduling result, construct the initial particle for cyclic round rolling optimization, execute the short-term scheduling rolling optimization strategy, and determine whether the preset constraints are met based on the current rolling optimization result. S300 performs cyclical rolling optimization until the preset iteration conditions are met, updates the optimal position of each individual and outputs the optimal solution set.

[0020] Please continue reading. Figure 1 Before performing the above steps, the following is also included: S000, a multi-stage microgrid intraday optimization model is established based on the objective function, constraints and decision variables; It needs to be explained in detail that before rolling coordinated control, a microgrid system consisting of wind turbine generators, photovoltaic generators, energy storage batteries, alkaline electrolyzers, proton exchange membrane (PEM) electrolyzers, fuel cells, and hydrogen storage systems needs to be constructed. Based on the objective function consisting of minimizing the overall cost and minimizing the constraint penalty cost, and based on the constraints consisting of power flow equality constraints, ramp rate constraints, system upper and lower reserve capacity constraints, output constraints of each device, capacity constraints of each device, and power balance constraints, as well as decision variables consisting of wind turbine generator output, photovoltaic generator output, energy storage battery output, total output of alkaline electrolyzer group, PEM electrolyzer output, fuel cell output, hydrogen load output, grid connection output, and the offset of the start and stop times of the group, a multi-stage microgrid intraday optimization model is established. For example, in the scheduling system of a wind power-hydrogen production coupling system, the day-ahead scheduling phase has achieved core functions such as peak shaving and valley filling and emergency backup power through dynamic adjustment of battery capacity, thus achieving initial smoothing of the system output process. Based on this, intraday scheduling needs to further achieve stable control of the output process, while ensuring that the overall output curve of intraday scheduling is similar to the shape of the day-ahead scheduling output curve, so as to ensure the consistency and continuity of the scheduling scheme.

[0021] Considering the significant differences in source-load power curves between day-ahead and intraday scheduling, and the need for intraday scheduling to take into account more refined electrolyzer start-up and shutdown constraints, strategies such as optimized load allocation within electrolyzer groups and precise correction of electrolyzer start-up and shutdown times are required to fully exploit the joint scheduling benefits of the wind-power coupled system. To simultaneously consider the economic objectives of the scheduling scheme and the goal of minimizing the total output adjustment for both day-ahead and intraday periods, this application's scheduling algorithm adopts a two-stage optimization strategy: the first stage utilizes a heuristic particle swarm optimization (PSO) algorithm to conduct multi-source-load coordination optimization, constructing a dual-source-load optimization model centered on economic cost and penalty cost. The first stage optimizes the model and repairs the equality and inequality constraints in the model based on preset rules to ensure the effectiveness and adaptability of the constraints. The second stage addresses the limitations of rule-based constraint processing methods, which are difficult to cover the entire scenario and have a large computational load. For particles that still cannot meet the constraints after the first stage optimization, a multi-source output coordination optimization based on linear programming is further performed to simultaneously complete the load distribution between electrolyzer units. Ultimately, the optimization and control aims to minimize the adjustment of key parameters such as battery energy storage capacity, hydrogen storage capacity, and grid-connected power between the day-ahead and intraday scheduling cycles.

[0022] The multi-stage microgrid intraday optimization model includes a one-stage optimization model and a two-stage optimization model: The one-stage optimization model includes objective functions consisting of minimizing overall cost and minimizing constraint penalty cost; constraints based on power flow equality, ramp rate constraints, system upper and lower reserve capacity constraints, output constraints of each device, capacity constraints of each device, and power balance constraints; and decision variables based on wind turbine output, photovoltaic unit output, energy storage battery output, total output of alkaline electrolyzer group, PEM electrolyzer output, fuel cell output, hydrogen load output, grid connection output, and start-up and shutdown time offsets of the cell group. Correspondingly, the objective function of the two-stage optimization model is to minimize the weighted sum of the battery SOC adjustment, grid connection capacity adjustment, and hydrogen storage capacity adjustment during the intraday scheduling compared to the daytime scheduling results; the decision variables include the number of cell groups starting and stopping and the output process of each electrolyzer; the constraints include start-up and shutdown time constraints, output range constraints, power balance constraints, ramp-up constraints, reserve capacity constraints, and capacity constraints. Understandably, the first-stage optimization model is used to transmit the output curves of each device to the second-stage model, including transmitting parameters such as the original device output curves, the number of tanks to start and stop, and the battery charging and discharging power. Among these parameters, the number of tanks to start and stop directly affects the range of device power values ​​during the linear programming calculation, while the battery charging and discharging power directly affects the direction of battery charging and discharging within the linear programming. Correspondingly, the two-stage optimization model is used to feed back the output process of each device to the one-stage model. This includes generating a completely new output process for each device with the goal of minimizing various capacity adjustments. At the same time, it follows specific parameter coverage rules. If the linear programming solution is successful, the new device output parameters are used to cover the original parameters. If the linear programming solution fails, the original output parameters of the device are kept unchanged.

[0023] Please continue reading. Figure 2 In this embodiment, step S100 includes: S110 reads the equipment output curves and equipment operation information of the day-ahead scheduling 24 hours based on the established multi-stage microgrid intraday optimization model, and reads the power prediction parameters based on the optimal economic intraday scheduling algorithm. S120: Based on the diversity of results from the non-dominated (Pareto) solution set, cyclically read particles and determine and select the optimal intraday scheduling optimization result that satisfies the constraints. S130: If the constraint conditions are met, the optimal intraday scheduling optimization result will be used as the intraday scheduling constraint conditions; otherwise, the next particle will be read and the judgment will continue.

[0024] Please continue reading. Figure 3 In this embodiment, step S200 includes: S210, set the rolling calculation running parameters, read the non-dominated solution particle parameters of the previous rolling optimization result, and construct the initial particles for the next rolling optimization; S220, execute the short-term scheduling rolling optimization strategy, and determine whether the preset constraints are met based on the current rolling optimization result; S230: If the current rolling optimization result meets the preset constraints, update the equipment capacity value and issue the equipment output control command for the first moment of the current rolling cycle; otherwise, continue to execute the next round of rolling optimization.

[0025] For example, the time scale for intraday scheduling is 15 minutes, and the time scale for day-ahead scheduling is 1 hour. To avoid the problem of frequent start-up and shutdown of electrolyzers caused by load forecasting deviations, intraday scheduling must follow the electrolyzer operating status determined by day-ahead scheduling. A targeted method for correcting the offset of electrolyzer start-up and shutdown times is proposed. At the same time, the operating parameters of other equipment in the system are optimized from the two dimensions of constraints and power balance. In addition, a penalty cost method is introduced to reduce the output deviation and capacity trend deviation of various equipment caused by load fluctuations, so as to ensure the stability of system operation.

[0026] Before executing intraday scheduling, basic operational information such as the 24-hour equipment output curve and the number of tank start-ups and shutdowns from the previous day's scheduling output must be read. The intraday scheduling algorithm sets optimal economy and minimum penalty cost as dual optimization objectives, and the input parameters cover the power forecast values ​​of wind power, photovoltaic, and hydrogen loads for 24 hours. Relying on the diversity of results from the non-dominated (Pareto) solution set, the intraday scheduling result with the best economy is selected under the premise that the scheduling scheme meets the preset constraints. If the current scheme does not meet the constraints, the next particle is read to continue the screening and judgment. The entire intraday scheduling process must use the optimization results of the previous day's scheduling as constraints to ensure the consistency of scheduling schemes at different time scales.

[0027] It should be explained in detail that, in the constraints of intraday scheduling, the range of equipment power values ​​for intraday scheduling is equal to the product of the day-ahead target power and the relative adjustment amount, thereby achieving effective connection and smooth transition between different levels of scheduling; given the difference in time scale between intraday and day-ahead scheduling, the target output of intraday scheduling needs to be based on the equipment output curve of day-ahead scheduling, and interpolation is used to interpolate the scheduling results to fill the data gaps between different time granularities; the penalty cost is calculated as the weighted sum of the absolute values ​​of the deviations between the day-ahead target capacity and the intraday corrected capacity, thereby quantifying the economic impact of scheduling deviations.

[0028] A rolling calculation cycle of 4 hours is set. During the scheduling execution, key status parameters such as battery capacity, hydrogen storage capacity, and grid connection capacity need to be read in real time. At the same time, the output data of each device and the number of tank group start-ups and shutdowns in the daily plan are retrieved. The number of times the rolling calculation cycle is executed is set. Each rolling calculation covers a 4-hour period. A scheduling time scale of 15 minutes is adopted. The relevant data and operating status are updated every 15 minutes to ensure that the scheduling scheme can adapt to real-time changes in operating conditions.

[0029] It should be explained in detail that before each round of rolling optimization calculation, the corresponding equipment output curves and equipment operation information for the wind power forecast, photovoltaic forecast, and hydrogen load forecast values ​​at the day-ahead scheduling level within the 4-hour period after the current moment must be reread. At the same time, the particle parameters of the non-dominated solutions in the previous rolling optimization results are retrieved to construct the initial particle swarm for the next rolling optimization. Then, the short-term scheduling rolling optimization calculation is executed according to the algorithm flow, and the optimization results are read to determine the constraint satisfaction. If the results do not meet the constraint requirements, the current round of rolling optimization continues to be executed. When the rolling optimization results meet the preset requirements, the core parameters such as battery capacity, hydrogen storage capacity, and grid connection capacity are immediately updated, and the equipment output control command for the first moment of the current rolling cycle is issued. Furthermore, if the number of iterations in the current rolling round has not yet reached the preset maximum value (Nmax), the next round of rolling optimization (N=N+1) must continue to be executed until the iteration termination condition is met.

[0030] Through the above technical solution, this application precisely focuses on the collaborative correlation problem of electrolyzer state combinations at different time scales. By deeply analyzing the mutual influence mechanism of coupling variables at different times, the start-up and shutdown times of the electrolyzers are dynamically adjusted and their working states are optimized, ultimately achieving collaborative optimization of multi-time scale scheduling. This solution can effectively reduce the negative impact of the volatility and uncertainty of renewable energy sources such as wind power and photovoltaics on the system. On the basis of optimizing equipment operating status and reducing losses from frequent start-up and shutdown of equipment, it significantly improves the utilization rate of renewable energy.

[0031] Please continue reading. Figures 1-6 In this embodiment, the rolling coordination control method includes: Determine whether the parameters in the particle satisfy the power balance equation. If not, perform constraint processing on the power balance equation. Calculate the dominance relationships of all particles within the particle swarm and update the non-dominated solutions to the repository; Based on the crowding index, select the first leader within different feasible solution sets in the non-dominated solution set, and update the inertia factor and variation factor; and Select a second leader from the remaining particles in different feasible solution sets within the non-dominated solution set; In the remaining iterations, the random particles learn the second leader according to the same learning mechanism; Please continue reading. Figures 1-6 In this embodiment, the rolling coordination control method further includes: The initial iteration is performed based on the first leader, updating the particle position, velocity, and feasibility results, and performing constraint repair on the power balance equation and the capacity balance equation. Based on the feasibility and fitness values ​​of the current iterative population, valid solutions are collected, identified, and updated in the first repository. Based on the currently determined effective solutions, an elite population is formed and clonal proliferation and local mutation strategies are implemented. Pruning is performed on particles in the non-dominated solution set based on the crowding index, and non-dominated solutions with high crowding and similarity are deleted. The second leader is selected based on the current non-dominated solutions and the principle of minimizing the weighted sum of economic and penalty costs. Valid solutions are collected and updated in the second repository to combine and generate the final repository.

[0032] For example, since this scheduling involves a large number of spatiotemporally coupled constraints, the strong correlation and complexity of such constraints greatly increase the difficulty of solving the system and the computational load. Therefore, it is necessary to make targeted improvements to the traditional particle swarm optimization (PSO) algorithm. This application adapts the initial solution generation strategy and complex constraint handling strategy to this scheduling scenario, so as to balance the computational efficiency and solution accuracy of the algorithm. First, the non-dominated solution set (corresponding to the time range [T:T+16]) obtained from the previous rolling optimization is filtered, and the parameters corresponding to the first time point T are removed, while the output data of each device in the time period [T+1;T+16] are retained. At the same time, the new time period parameter (T+17) corresponding to the current rolling optimization time is randomly generated. The retained historical output data is integrated with the randomly generated new time period parameter to construct the initial solution of the population particles for the current rolling optimization time range [T+1:T+17]. To address the issue that the number of particles in the non-dominated solution set obtained from the previous rolling optimization is lower than the preset population size, a hybrid supplementation strategy of partial particle cloning and partial randomization is adopted to ensure that the population size meets the algorithm iteration requirements. Considering that this scheduling is a complex real-time dynamic optimization problem, and that the parameter adjustment range of intraday scheduling is determined based on the day-ahead scheduling output curve, the adjustment range of each parameter is relatively limited. Therefore, the algorithm design needs to clearly prioritize both global optimization capability and convergence speed. To overcome the limitation of traditional particle swarm optimization algorithms being prone to getting stuck in the optimal solution during the local search phase, an elite population cloning and mutation mechanism is added to the improved algorithm: cloning operations preserve high-quality gene fragments in the population, ensuring the algorithm's ability to explore the globally optimal region; mutation operations introduce new search directions, preventing population evolution from stagnation, thereby effectively enhancing the global search performance of the improved particle swarm optimization algorithm and improving the optimization efficiency and accuracy within the limited parameter range.

[0033] The improved particle swarm optimization algorithm iterative process includes: first, completing the population initialization operation, verifying the parameters of each particle in the particle swarm one by one to determine whether they satisfy the power balance equation constraint; if the particle parameters do not satisfy the core equation constraint, then immediately performing targeted constraint repair processing; then calculating the dominance relationship between all particles in the particle swarm, and updating the selected non-dominated solutions to a dedicated repository; based on the crowding index, selecting the first leader particle (A) from different feasible solution subsets of the non-dominated solution set, and dynamically updating the inertia factor and mutation factor according to the algorithm iteration progress; simultaneously selecting the second leader particle (B) from the remaining feasible solution subsets of the non-dominated solution set, and randomly selecting some particles during the iteration process to learn parameters from the second leader particle (B) according to a preset learning mechanism, thereby enriching the diversity of the population.

[0034] The algorithm uses the first leader (A) particle as the core to carry out initial iterative calculations. During the iteration process, the position and velocity parameters of each particle are updated in real time, and the feasibility of the particles is judged. Simultaneously, constraint repair processing is performed on the power balance equation and the capacity balance equation. In this process, a load optimization allocation algorithm based on linear programming is embedded. Through the accurate solution capability of linear programming, the optimal load allocation of equipment such as electrolytic cells is achieved, further improving the satisfaction of constraints and the rationality of the scheduling scheme.

[0035] In addition, the algorithm selects and collects effective solutions based on the feasibility and fitness values ​​of each particle in the current iterative population, and updates them to the first storage repository (A1). Based on the effective solutions in the first storage repository (A1), it combines and constructs an elite particle swarm (F), performs a cloning and proliferation operation on the elite swarm to make its particle number reach the preset target size, and then performs a local mutation operation on all the proliferated elite particles. After the mutation is completed, the dominance relationship between particles is recalculated.

[0036] In addition, after the elite group undergoes cloning and mutation operations, some particles may exhibit a situation where they have no dominant relationship but are close in Euclidean distance. Such similar particles will reduce the diversity and distribution of the solution set. Therefore, based on the crowding index, pruning is performed on the particles in the non-dominated solution set to remove similar non-dominated solutions with high crowding and ensure the uniform distribution of the solution set. The pruned and optimized non-dominated solutions are updated to the second repository (B1). The second leader (B) is selected based on the principle of minimizing the weighted sum of economic cost and penalty cost. Valid solutions from the iteration process are collected again and updated to the second repository (B1). Finally, the valid solutions from the first repository (A1) and the second repository (B1) are integrated to form the final repository (A1+B1).

[0037] Please continue reading. Figures 1-7 In this embodiment, the short-term scheduling rolling optimization strategy includes: Read the number of tanks started and stopped at the current time in the day-ahead scheduling solution, and based on the change value of the number of tanks started and stopped before and after the time, find the position of the change in the number of tanks started and stopped, and determine the adjustment range of the tanks. Set different position constraint ranges, randomly select positions based on the constraint ranges and discretize the deviations, determine the changes in the number of tank groups at different times based on the determined start and stop position offsets, and determine the power constraint ranges of different tank groups, so as to determine and modify the variable constraint ranges.

[0038] For example, in the execution flow of the optimization algorithm, the coordinated optimization of day-ahead and intraday scheduling is achieved by precisely optimizing the variable of the start-up time of the electrolytic cell group: First, the number of electrolytic cells started and stopped at the current time is read from the daily scheduling solution. Then, the changes in the number of cells started and stopped at adjacent times are tracked and analyzed to locate the key time node T where the number of cells started and stopped changes (e.g., when T=8, corresponding to the 2-hour time in the daily scheduling). Based on this key node, a reasonable adjustment range for the start and stop times is defined. In this application, the adjustment range is set as [max(T [2,1):min(T+2,16)], thereby limiting the boundary of the start-stop time offset; based on this, during the particle learning iteration process or population initialization stage, a constraint interval matching the above adjustment range is set for the start-stop time parameters of each particle. The particle randomly selects the start-stop position within the constraint interval, and the selected position parameters are discretized to generate the deviation; after the start-stop position offset of the tank group is determined, the change law of the number of tank groups with the time dimension can be determined, and then the power constraint range of the electrolytic cell can be determined according to the change characteristics of the number of tank groups, and the dynamic correction of the constraint range of relevant variables in the algorithm can be completed accordingly, ensuring the rationality and effectiveness of parameter optimization.

[0039] It should be noted in detail that this application verifies the superiority of the improved PSO algorithm with the addition of an elite clone local search mechanism through comparative experiments. The relevant comparative results can be found in the following references. Figure 10 In the traditional improved PSO algorithm without introducing the elite clonal local search mechanism, when the scheduling scheme satisfies all constraints and the cost reaches the minimum optimal case, its optimal cost value is higher than -180, and the corresponding solution set is distributed in region A (in the cost accounting system, the revenue from electricity and hydrogen sales is negative and included in the cost statistics, while the consumption of other items is positive and included in the cost statistics); while the improved PSO algorithm with the introduction of the elite clonal local search mechanism has an optimal cost value lower than -190, and the corresponding solution set is distributed in region B (region A1 and region A are equivalent cost regions).

[0040] From the perspective of solution diversity, under the condition that the penalty cost threshold is set below 600, the improved PSO algorithm with the introduction of the elite clonal local search mechanism outputs 40 non-dominated solution sets, which is significantly higher than the 18 non-dominated solution sets of the improved PSO algorithm without the introduction of this mechanism. At the same time, the former can provide a significantly wider range of minimum cost choices, thus effectively improving the efficiency of algorithm optimization performance and solution set diversity.

[0041] To address the issue of electrolytic cell group start-up time offset that may be caused by random load fluctuations, this application proposes a multi-timescale scheduling coordination strategy and a unit combination offset mechanism: The daytime scheduling plan specifies that the number of electrolytic cell groups will be reduced from 4 to 2 at 1 hour, and will be increased from 2 to 4 at 2 hours. Based on the mechanism proposed in this application, intraday scheduling can select the optimal cell group start-up time within a 30-minute time window before and after the start-up and shutdown times set by the daytime scheduling (corresponding to 4 sampling points on a 15-minute intraday time scale). The daytime scheduling only allows for the translational adjustment of the cell group start-up and shutdown times, without changing the baseline value of the number of cell groups to be started or shut down. This achieves an effective correlation between the optimization results of the daytime scheduling and intraday scheduling at the current moment, thereby avoiding the problem of frequent cell group start-up and shutdown due to load fluctuations.

[0042] Through the synergistic application of the above technical solutions, this application constructs a two-stage particle swarm optimization algorithm that integrates multiple improvement strategies. Specifically, it includes a repair strategy that adds different types of equality and inequality constraints, introduces an elite group cloning mutation mechanism, and designs an initial solution generation strategy adapted to the scheduling scenario. These improvements effectively enhance the algorithm's global search capability and local exploration accuracy. With minimizing the total output adjustment of equipment within the day-ahead and intraday scheduling cycles as the core optimization objective, a solution module based on linear programming is embedded to achieve coordinated output of multiple energy sources and optimal power allocation among electrolyzer groups. Ultimately, this achieves rapid approximation of the intraday scheduling output curve to the day-ahead scheduling planned output curve, while effectively shortening the algorithm's solution time.

[0043] Refer to Figure 9 The present invention also provides an embodiment of a microgrid intraday multi-stage rolling coordinated control system for operating the microgrid intraday multi-stage rolling coordinated control method described in the above embodiments. The control system includes: The intraday optimization model building module is used to establish a multi-stage intraday optimization model for microgrids based on the objective function, constraints, and decision variables. The intraday optimization model solution module is used to run the rolling coordination control method, read the power prediction parameters, read the particles to determine and select the optimal intraday scheduling result, construct the initial particles for cyclic rolling optimization, execute the short-term scheduling rolling optimization strategy, determine whether the current rolling optimization result meets the preset constraints, perform cyclic rolling optimization until the preset iteration conditions are met, update the optimal position of the individual and output the optimal solution set.

[0044] For example, the effect analysis of the intraday two-stage rolling coordinated control of microgrids guided by optimal economy in this application is as follows, and the relevant verification results are as follows: Figures 11-14 : Figure 11The results of the verification of the core input information of the day-ahead scheduling scheme are shown. The data shows that the maximum deviation of the power balance is less than 3% of the power capacity, which fully meets the basic requirements of the scheduling scheme for the feasibility of the input data. Key parameters such as battery state of charge (SOC), grid-connected power, and hydrogen storage system capacity are all within the preset constraint range, and the number of start-ups and shutdowns of the electrolyzer throughout the day strictly conforms to the equipment life design standard. Figure 12 The figure shows the 24-hour power forecast curve used by the day-ahead dispatching system. The curve covers three core power data: wind power output, photovoltaic power output, and hydrogen load. As can be seen from the curve characteristics, the forecast deviation shows a trend of gradually increasing with the increase of time offset, with the maximum relative deviation reaching 40%. This deviation characteristic also reflects the uncertainty of long-term power forecasting. Figure 13 The figure shows the 24-hour power forecast curve on which intraday dispatch is based. The power types it covers are consistent with those of day-ahead dispatch, including wind power output, photovoltaic power output, and hydrogen load. Compared with the forecast data of day-ahead dispatch, the power forecast accuracy of intraday dispatch is significantly improved, the forecast deviation is significantly reduced, and the maximum relative deviation is only 10%, which effectively reduces the risk of dispatch mismatch caused by forecast error. Figure 14 The figure shows a comparison of capacity parameters between intraday and day-ahead scheduling results after 1 hour of rolling (4 rolling cycles). The algorithm used in this test had a population size of 100 and 50 iterations. The comparison results show that key indicators such as battery state of charge (SOC), hydrogen storage capacity, and grid connection capacity are all stably within the preset constraints, fully ensuring the safety and compliance of the system operation. At the same time, the correction amount of capacity parameters by intraday scheduling and the target amount of day-ahead scheduling are highly similar in the overall shape of the output process, thus verifying the effectiveness of the two-stage rolling coordinated control strategy proposed in this application in achieving smooth connection of scheduling at different time scales.

[0045] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-stage rolling coordinated control method for microgrids during the day, characterized in that, include: Based on the established multi-stage microgrid intraday optimization model, a rolling coordination control method is used to read power prediction parameters and select the optimal intraday scheduling result by reading particle judgment. Based on the non-dominated solution particle parameters of the optimal intraday scheduling result that has been read, construct the initial particle for cyclic round rolling optimization, execute the short-term scheduling rolling optimization strategy, and determine whether the preset constraints are met based on the current rolling optimization result; Perform cyclical optimization until the preset iteration conditions are met, update the optimal position of each individual and output the optimal solution set.

2. The microgrid intraday multi-stage rolling coordinated control method according to claim 1, characterized in that, Also includes: A multi-stage microgrid intraday optimization model is established based on the objective function, constraints, and decision variables. The multi-stage microgrid intraday optimization model includes a one-stage optimization model and a two-stage optimization model. The one-stage optimization model is used to transmit the output curves of each device to the two-stage model, and the two-stage optimization model is used to feed back the output process of each device to the one-stage model.

3. The intraday multi-stage rolling coordinated control method for microgrids according to claim 1, characterized in that, The rolling coordination control method based on the established multi-stage microgrid intraday optimization model reads power prediction parameters and selects the optimal intraday scheduling result from particle judgment, including: Based on the established multi-stage microgrid intraday optimization model, the equipment output curves and equipment operation information of the day-ahead scheduling 24 hours are read, and the power prediction parameters are read based on the optimal economic intraday scheduling algorithm. Based on the diversity of results from the non-dominated (Pareto) solution set, the particles are read cyclically, and the optimal intraday scheduling optimization result that satisfies the constraints is selected. If the constraints are met, the optimal intraday scheduling result will be used as the intraday scheduling constraint; otherwise, the next particle will be read and the decision will continue.

4. The intraday multi-stage rolling coordinated control method for microgrids according to claim 1, characterized in that, In the intraday scheduling constraints, the equipment power range of intraday scheduling is equal to the target daytime power * relative adjustment amount. The target output of intraday scheduling uses the daytime scheduling equipment output curve as a benchmark, and interpolation is used to interpolate the scheduling results. The penalty cost is the weighted sum of the absolute values ​​of the deviations between the daytime target capacity and the intraday corrected capacity.

5. The intraday multi-stage rolling coordinated control method for microgrids according to claim 1, characterized in that, The non-dominated solution particle parameters are based on the read optimal intraday scheduling results, and initial particles for cyclic round rolling optimization are constructed. A short-term scheduling rolling optimization strategy is executed, and the current rolling optimization result is used to determine whether the preset constraints are met, including: Set the rolling calculation parameters, read the non-dominated solution particle parameters of the previous rolling optimization result, and construct the initial particles for the next rolling optimization. Execute a short-term scheduling rolling optimization strategy and determine whether the preset constraints are met based on the current rolling optimization result; If the current rolling optimization result meets the preset constraints, the equipment capacity value is updated, and the equipment output control command for the first moment of the current rolling cycle is issued; otherwise, the next rolling optimization cycle is executed.

6. The intraday multi-stage rolling coordinated control method for microgrids according to claim 5, characterized in that, Before each round of rolling optimization, it is necessary to reread the equipment output curve and equipment operation information of the day-ahead scheduling 4 hours after the current time. If the current iteration count of the rolling cycle has not reached the preset number, continue to execute the next round of rolling optimization.

7. The intraday multi-stage rolling coordinated control method for microgrids according to claim 1, characterized in that, The rolling coordination control method includes: Determine whether the parameters in the particle satisfy the power balance equation. If not, perform constraint processing on the power balance equation. Calculate the dominance relationships of all particles within the particle swarm and update the non-dominated solutions to the repository; Based on the crowding index, select the first leader within different feasible solution sets in the non-dominated solution set, and update the inertia factor and variation factor; and Select a second leader from the remaining particles in different feasible solution sets within the non-dominated solution set; In the remaining iterations, the random particles learn the second leader according to the same learning mechanism.

8. The intraday multi-stage rolling coordinated control method for microgrids according to claim 7, characterized in that, The rolling coordination control method further includes: The initial iteration is performed based on the first leader, updating the particle position, velocity, and feasibility results, and performing constraint repair on the power balance equation and the capacity balance equation. Based on the feasibility and fitness values ​​of the current iterative population, valid solutions are collected, identified, and updated in the first repository. Based on the currently determined effective solutions, an elite population is formed and clonal proliferation and local mutation strategies are implemented. Pruning is performed on particles in the non-dominated solution set based on the crowding index, and non-dominated solutions with high crowding and similarity are deleted. The second leader is selected based on the current non-dominated solutions and the principle of minimizing the weighted sum of economic and penalty costs. Valid solutions are collected and updated in the second repository to combine and generate the final repository.

9. The intraday multi-stage rolling coordinated control method for microgrids according to claim 1, characterized in that, Short-term scheduling rolling optimization strategies include: Read the number of tanks started and stopped at the current time in the day-ahead scheduling solution, and based on the change value of the number of tanks started and stopped before and after the time, find the position of the change in the number of tanks started and stopped, and determine the adjustment range of the tanks. Set different position constraint ranges, randomly select positions based on the constraint ranges and discretize the deviations, determine the changes in the number of tank groups at different times based on the determined start and stop position offsets, and determine the power constraint ranges of different tank groups, so as to determine and modify the variable constraint ranges.

10. A multi-stage rolling coordinated control system for microgrids during the day, characterized in that, The control method is used to operate the intraday multi-stage rolling coordinated control method for microgrids as described in any one of claims 1 to 9, and the control system includes: The intraday optimization model building module is used to establish a multi-stage intraday optimization model for microgrids based on the objective function, constraints, and decision variables. The intraday optimization model solution module is used to run the rolling coordination control method, read the power prediction parameters, read the particles to determine and select the optimal intraday scheduling result, construct the initial particles for cyclic rolling optimization, execute the short-term scheduling rolling optimization strategy, determine whether the current rolling optimization result meets the preset constraints, perform cyclic rolling optimization until the preset iteration conditions are met, update the optimal position of the individual and output the optimal solution set.

Images