Power grid interaction multi-target scheduling strategy considering pumped storage start-stop strategy

Abstract

The invention discloses a power grid interaction multi-target scheduling strategy considering a pumped storage start-stop strategy. The power grid interaction multi-target scheduling strategy comprises the following steps: modeling for power grid distributed resource output characteristics; modeling based on the power constraint of the pumped storage start-stop strategy; constructing a power grid interaction multi-target optimization scheduling model considering a pumped storage operation strategy and power grid peak regulation and frequency modulation; an improved multi-target squala optimization algorithm is used for carrying out power grid optimization scheduling, multi-target expansion is carried out on an original squala optimization algorithm based on a Pareto theory, and a basic multi-target squala optimization algorithm framework is designed; levy flight is added in collaborative predation in the position updating stage to enhance population diversity and avoid local optimum, and secondary updating is performed on the population in the population updating process to perform local search. And solving the optimal scheduling strategy of the system, so that the thermal power generating unit can stably operate, and meanwhile, the frequency deviation is inhibited, thereby realizing safe and stable output of the system.

Description

A multi-objective scheduling strategy for power grid interaction that takes into account pumped storage start-up and shutdown strategies Technical Field

[0001] A multi-objective grid interaction dispatch strategy that takes into account pumped storage start-up and shutdown strategies aims to achieve optimal grid operation under multiple objectives of economy, reliability and renewable energy consumption by rationally configuring and coordinating the optimization of related equipment. Background Technology

[0002] Due to the high randomness and volatility of new energy sources, the balance between power supply and demand becomes more complex. The inherent characteristics of new energy sources severely reduce the grid's balancing capacity, and traditional technologies and production organization models can no longer meet the normal operation requirements of grids with a high proportion of new energy sources. With the advancement of energy transition, energy storage devices, as important carriers for integrating distributed energy and optimizing the energy structure, have become a research hotspot for optimized scheduling, aiming to achieve a balance between economic efficiency, reliability, and renewable energy absorption. Pumped storage hydroelectric power stations have the characteristics of rapid energy storage and release, effectively addressing the impact caused by the randomness of renewable energy supply-side generation, while also having the ability to respond promptly to dynamic load changes, thus improving the overall stability and operating characteristics of the power system. As a mature energy storage technology, pumped storage has advantages such as large storage capacity and stable operation, effectively balancing power fluctuations in the grid and improving power supply reliability. Wind-solar hybridization utilizes the temporal and spatial complementarity of wind and solar energy, improving the utilization efficiency and stability of renewable energy. However, the intermittency and uncertainty of wind and solar power generation still pose challenges to the stable operation of the grid, requiring optimized scheduling strategies to address these challenges.

[0003] Current energy storage devices and grid optimization scheduling methods mostly target single objectives, such as peak shaving or frequency regulation, which are insufficient to meet the multi-objective optimization needs of grid interaction. Therefore, it is necessary to design a grid-oriented, multi-time-scale, multi-objective optimization scheduling method based on pumped storage start-up and shutdown, and pumped storage peak shaving and frequency regulation. This method can comprehensively consider multiple objectives such as grid economy, thermal power unit generation stability, and renewable energy consumption to achieve optimal microgrid operation. Summary of the Invention

[0004] The present invention aims to provide a multi-objective optimization scheduling strategy for power grids to overcome the problems of insufficient consideration of pumped storage start-up and shutdown planning and difficulty in balancing the multi-objective optimization needs of power grids in the existing technology, so as to achieve synergistic optimization of economic efficiency, reliability and renewable energy consumption targets.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0006] This invention discloses a multi-objective scheduling strategy for power grid interaction that takes into account pumped storage start-up and shutdown strategies, comprising the following steps:

[0007] S1: Modeling the output characteristics of distributed power grid resources;

[0008] S2: Modeling based on power constraints of pumped storage start-stop strategy;

[0009] S3: Construct a multi-objective optimization scheduling model for power grid interaction that takes into account both pumped storage operation strategy and power grid peak shaving and frequency regulation.

[0010] S4: Power grid optimization scheduling is performed using an improved multi-objective white shark optimization algorithm. Based on Pareto theory, the original white shark optimization algorithm is extended to multiple objectives, and a basic multi-objective white shark optimization algorithm framework is designed. Levy flight is added to the cooperative predation in the position update stage to enhance population diversity and avoid local optima. Secondary updates are performed on the population during the population update process to conduct local search.

[0011] Furthermore, in step S1, power data from the load side and the generator side of the power system are obtained, and constraint models of the output characteristics of the system are constructed respectively. Specifically, this includes the following steps:

[0012] S11: Obtain 24-hour wind power generation, photovoltaic power generation, and total load power to obtain 24-hour net load data.

[0013] S12: Add constraints to the power system, specifically including:

[0014] System power balance constraints:

[0015]

[0016] Where, N c N w N p N phs These represent the number of each type of thermal power unit, wind power unit, photovoltaic unit, and pumped storage unit, respectively. c P w P p P phs P load These are the output power of thermal power units, wind power units, photovoltaic power units, pumped storage units, and total load power, respectively.

[0017] Thermal power unit ramping constraints:

[0018] .

[0019] in, , These are the output power of the i-th thermal power unit at time t and the output power of the thermal power unit at time t-1, respectively.

[0020] Wind turbine output constraints:

[0021]

[0022] Among them, P w It refers to the output power of wind power generation.

[0023] Photovoltaic generator output constraints:

[0024]

[0025] Among them, P p It refers to the output power of wind power generation.

[0026] S13: Model the operating costs and volatility of each power generation unit, specifically including:

[0027] Cost of generating electricity from thermal power units:

[0028]

[0029] Among them, P i,t Let C be the output power of the i-th thermal power unit at time t. thermal This represents the total power generation cost of the thermal power unit.

[0030] Pumped storage operation and maintenance costs:

[0031]

[0032] Among them, P phs,t Let C be the output power of the pumped storage unit at time t. PHS This refers to the operation and maintenance costs of pumped storage hydroelectric power.

[0033] Total operating costs:

[0034]

[0035] Variance of thermal power unit output:

[0036]

[0037] Furthermore, in step S2, the actual operation of pumped storage involves two start-up and shutdown actions: startup and shutdown. Start-up is further divided into pumping and power generation. Therefore, 1, 0, and -1 are introduced to represent their respective operating conditions. Based on the actual operating characteristics of pumped storage, a pumped storage start-up and shutdown method is generated by classifying the indicators and allocating power. This includes the following steps:

[0038] S21: The following output constraints are added for pumped storage units:

[0039]

[0040] in, Let be the reservoir capacity at time t. Let t be the generator power of the pumped storage unit. Let t be the power of the pump unit of the pumped storage unit. For unit efficiency.

[0041] S22: The load fluctuation component and net load standard deviation are calculated from the net load data. The degree of net load fluctuation in each period can be known from the above two data.

[0042]

[0043]

[0044]

[0045]

[0046] S23: J is obtained by linear weighting the above two indicators. Start-up and shutdown thresholds are set based on J. The distribution of J in each time period yields three levels: strong, medium, and weak. The PHS operating status is divided into 7 intervals, namely 3 discharge levels, 3 charging levels, and 1 shutdown state. The optimal peak-shaving configuration of the power system is obtained through a multi-objective improved optimization algorithm.

[0047]

[0048] Furthermore, in step S3, after the overall peak-shaving power is determined, frequency deviations are prone to occur during periods of rapid load increase or decrease, requiring more frequent adjustments. Therefore, frequency adjustments are performed between 12:00 and 13:00 and between 18:00 and 19:00, respectively. This includes the following steps:

[0049] S31: Obtain the raw data of the corresponding load data, wind power generation data, and photovoltaic power generation data for each 1-minute time period.

[0050] S32: Perform primary frequency regulation droop control on the deviation frequency, set the droop coefficient and the energy storage participation capacity to obtain the primary frequency regulation coefficient, and the pumped storage system quickly adjusts the power according to the droop characteristics to offset the frequency difference.

[0051]

[0052]

[0053] Among them, P part For participation capacity.

[0054]

[0055] S33: Primary frequency modulation can only suppress frequency deviation, not completely eliminate it. Therefore, an auxiliary frequency modulation module is added to eliminate the blind spot that cannot be adjusted in droop control. Through the integral element, the steady-state frequency difference is gradually eliminated, restoring the frequency to the rated value.

[0056]

[0057]

[0058] Where P0 is the AGC baseline and I(t) is the integral state.

[0059]

[0060]

[0061] S34: The total power of primary frequency modulation for the corresponding time period can be obtained by adding the power obtained from the above droop control and the auxiliary frequency modulation.

[0062]

[0063] Furthermore, in step S4, the improved multi-objective great white shark optimization algorithm includes several aspects: initializing the great white shark population, fitness evaluation, random relocation, wave-like swimming search, cooperative predation search, and population update. The specific steps are as follows:

[0064] Initialize the white shark population: Set the maximum number of iterations, introduce initialization to the decision variables to generate a group of white shark individuals, each representing a power grid interaction optimization scheduling strategy. The initialization formula is as follows:

[0065]

[0066] Where, x i,j Let represent the position information of the i-th shark individual in the j-th decision variable dimension, where LB is the lower bound and UB is the upper bound. rand is a random function.

[0067] Fitness assessment: The fitness of each individual great white shark is calculated, taking into account constraints. The formula is as follows:

[0068]

[0069] Non-dominated sorting and external archive management: Sort the population according to Pareto dominance, mark non-dominated solutions, and store non-dominated solutions in an external archive. After each iteration, compare the newly generated non-dominated solutions with the original solutions in the archive, retain all non-dominated solutions, and replace them with crowding distance if the archive exceeds the capacity limit.

[0070]

[0071] Where, x i Let i be the i-th shark individual, and k be the number of objective functions. Let p be the function value of the individual on the p-th objective function. This represents the maximum value in the current population. It is the minimum value in the current population.

[0072] Random values ​​are used to determine the update of individual shark particles.

[0073] If rand < 0.5, perform a global search for individual sharks:

[0074] Random relocation:

[0075]

[0076] Where D is the number of dimensions of the decision variable.

[0077] Wave movement:

[0078]

[0079] If rand > 0.5, perform localized development of the individual shark:

[0080] Levy flight:

[0081]

[0082] Where L is the Levy flight step size coefficient. The location of a random reference individual.

[0083] Cooperative predation:

[0085] The secondary local search mechanism includes: first, selecting 20% ​​of individuals and a non-dominated solution, then comparing their fitness and using differential mutation to maintain diversity for individuals with higher fitness; and using elite attraction and Gaussian perturbation to maintain convergence for individuals with low fitness.

[0086] If the new solution is better than the current solution, then accept it:

[0087]

[0088] Otherwise, move in the opposite direction to the inferior solution:

[0089]

[0090] As the number of iterations increases, the output power of each unit carried by the shark population is updated and optimized. After merging the population and archiving, the population is sorted non-dominated and then the next generation of population is filled according to the leading level. This yields a high-quality, highly diverse solution set of approximate Pareto optimal solutions, thus obtaining the optimal output of the multi-energy complementary power system.

[0091] The technical effects achieved by this invention are as follows:

[0092] (1) Based on the construction of a multi-energy complementary system of wind, solar, thermal, and pumped storage, a pumped storage start-up and shutdown strategy is added. The start-up and shutdown time distribution and power planning of pumped storage are obtained through comprehensive indicators based on electricity price and net load fluctuations. (2) A multi-time-scale scheduling model is constructed by coupling frequency regulation time during peak shaving time to better cope with the problem of frequent frequency changes caused by large changes in net load after the addition of new energy output. (3) Non-dominated sorting and congestion comparison are added to the white shark optimization algorithm to generate a multi-objective white shark algorithm. On this basis, Levy flight and secondary local search mechanisms are added to increase the diversity and convergence of the algorithm solution. Attached Figure Description

[0093] Figure 1 is a flowchart of a multi-objective scheduling strategy for power grid interaction that takes into account the start-up and shutdown strategy of pumped storage as described in this invention.

[0094] Figure 2 is a flowchart of the pumped storage start-up and shutdown strategy described in this invention.

[0095] Figure 3 is a flowchart of the solution process for the multi-objective optimization scheduling model of the power grid described in this invention. Detailed implementation methods

[0097] With reference to the accompanying drawings of the embodiments of the present invention, the relevant technical solutions will be systematically and comprehensively described below. Figure 1 illustrates the overall process of power grid optimization scheduling. Based on this process, a general framework for the optimization scheduling problem is constructed. First, a multi-energy complementary model of wind-solar-thermal-pumped-storage power stations is built using abandoned mines as a scenario. Second, a state prediction method for pumped-storage units based on net load is added to the model to improve the flexible peak-shaving capability of pumped-storage. On the basis of peak shaving, refined power allocation is performed for periods with frequent frequency changes. Finally, the multi-objective white shark optimization algorithm is improved to find the optimal scheduling allocation. Figure 2 shows the pumped-storage start-up and shutdown strategy under actual conditions. Based on the changes in net load in the system, start-up and shutdown planning and power zoning of pumped-storage units are carried out, which is more in line with the actual situation. Figure 3 introduces the flowchart of the self-improved multi-objective white shark algorithm. Global optimization is achieved through random relocation and sinusoidal swimming. The convergence of the algorithm is improved through cooperative predation and Levy flight. Finally, secondary updates are performed on the obtained white shark individuals, and the population is optimized based on the sum of the fitness of different individuals. The calculation method is reasonably designed and improved to better solve the power grid optimization scheduling problem constructed in this paper. It should be noted that the embodiments represent only some implementation examples of the present invention and do not cover all possible technical forms. The specific implementation examples listed herein are only for illustrative purposes and should not be considered as limiting definitions of the present invention. Based on the embodiments disclosed in this invention, all technical solutions derived by those skilled in the art without creative effort are within the scope of protection of this invention.

[0098] Referring to Figure 1, the overall process of the power grid optimization scheduling problem is established; based on the content of Figure 2, a constraint model for the pumped storage start-up and shutdown strategy is constructed; according to Figure 3, the White Shark optimization algorithm is improved into a multi-objective White Shark optimization algorithm, and its working principle is explained. A multi-objective scheduling method for power grid interaction considering the pumped storage start-up and shutdown strategy includes the following steps:

[0099] S1: Modeling the output characteristics of distributed power grid resources;

[0100] In this embodiment of the invention, modeling of the power output of wind and solar power systems, pumped storage power output, and thermal power system output and constraints is included.

[0101] S2: Modeling based on power constraints of pumped storage start-stop strategy;

[0102] In this embodiment of the invention, the operating status of pumped storage is affected by a comprehensive index based on electricity price and net load fluctuations. The power range is defined by three levels of the comprehensive index, thereby fully realizing the peak-shaving function of pumped storage in the multi-energy complementary power system and providing a basis for optimized scheduling. The constructed multi-objective optimization scheduling model of the power grid aims at minimizing operating costs and the variance of thermal power unit fluctuations, obtaining the optimal peak-shaving output of each unit over 24 hours.

[0103] S3: Construct a multi-objective optimization scheduling model for power grid interaction that takes into account both pumped storage operation methods and power grid peak shaving and frequency regulation;

[0104] In this embodiment of the invention, after the overall peak-shaving power is determined, frequency deviations are prone to occur during periods of rapid load increase or decrease, requiring more frequent adjustments. Therefore, frequency modulation is performed between 12:00 and 13:00 and between 18:00 and 19:00, respectively. First, a frequency modulation droop control is performed on the offset frequency, and then auxiliary frequency modulation is used to adjust the blind zone missed in the droop control, thus obtaining the complete frequency modulation power.

[0105] S4: An improved multi-objective white shark optimization algorithm is used for power grid optimal scheduling. Based on Pareto theory, the original white shark optimization algorithm is extended to a multi-objective model, and a basic multi-objective white shark optimization algorithm framework is designed. In the position update phase, the Levy-flight method is introduced to allow the population to jump around, thereby improving population diversity and global search capability. During the population update process, a secondary local search mechanism is implemented to better balance the algorithm's exploration and development capabilities and avoid getting trapped in local optima. This series of improvements significantly enhances the algorithm's performance and applicability, providing a more effective solution to the power grid optimal scheduling problem.

[0106] Specifically, step S1 includes the following steps:

[0107] S11: Obtain 24-hour wind power generation, photovoltaic power generation, and total load power to obtain 24-hour net load data;

[0108] S12: Add constraints to the power system, specifically including:

[0109] System power balance constraints:

[0110]

[0111] Where, N c N w N p N phs These represent the number of each type of thermal power unit, wind power unit, photovoltaic unit, and pumped storage unit, respectively. c Pw P p P phs P load These are the output power of thermal power units, the output power of wind turbine generator units, the output power of photovoltaic generator units, the output power of pumped storage units, and the total load power, respectively.

[0112] Thermal power unit ramping constraints:

[0113] .

[0114] in, , These are the output power of the i-th thermal power unit at time t and the output power of the thermal power unit at time t-1, respectively.

[0115] Wind turbine output constraints:

[0116]

[0117] Among them, P w It refers to the power output of wind power generation;

[0118] Photovoltaic generator output constraints:

[0119]

[0120] Among them, P p It refers to the power output of wind power generation;

[0121] S13: Model the operating costs and volatility of each power generation unit, specifically including:

[0122] Cost of generating electricity from thermal power units:

[0123]

[0124] Among them, P i,t Let C be the output power of the i-th thermal power unit at time t. thermal This represents the total power generation cost of the thermal power unit.

[0125] Pumped storage operation and maintenance costs:

[0126]

[0127] Among them, P phs,t Let C be the output power of the pumped storage unit at time t. PHS This refers to the operation and maintenance costs of pumped storage hydroelectric power.

[0128] Total operating costs:

[0129]

[0130] Variance of thermal power unit output:

[0131]

[0132] Specifically, step S2 includes the following steps:

[0133] S21: Add output constraints for pumped storage units;

[0134]

[0135] in, Let be the reservoir capacity at time t. Let t be the generator power of the pumped storage unit. Let t be the power of the pump unit of the pumped storage unit. For unit efficiency;

[0136] S22: The load fluctuation component and net load standard deviation are calculated from the net load data. The degree of net load fluctuation in each period can be known through the above two data.

[0137]

[0138]

[0139]

[0140]

[0141] S23: J is obtained by linear weighting the above two indicators. Start-up and shutdown thresholds are set based on J. The distribution of J in each time period yields three levels: strong, medium, and weak. The PHS operating status is divided into 7 intervals, namely 3 discharge levels, 3 charging levels, and 1 shutdown state. The optimal peak-shaving configuration of the power system is obtained through a multi-objective improved optimization algorithm.

[0142]

[0143] Specifically, step S3 includes the following steps:

[0144] S31: Obtain the raw data of corresponding load data, wind power generation data, and photovoltaic power generation data for each 1-minute time period;

[0145] S32: Perform primary frequency regulation droop control on the deviation frequency, set the droop coefficient and the energy storage participation capacity to obtain the primary frequency regulation coefficient, and the pumped storage system quickly adjusts the power according to the droop characteristics to offset the frequency difference;

[0146]

[0147]

[0148] Among them, P part For participation capacity;

[0149]

[0150] S33: Primary frequency modulation can only suppress frequency deviation, not completely eliminate it. Therefore, an auxiliary frequency modulation module is added to eliminate the blind spot that cannot be adjusted in droop control. Through the integral element, the steady-state frequency difference is gradually eliminated, restoring the frequency to the rated value.

[0151]

[0152]

[0153] Where P0 is the AGC baseline and I(t) is the integral state;

[0154]

[0155]

[0156] S34: The total power of primary frequency modulation for the corresponding time period can be obtained by adding the power obtained from the above droop control and the auxiliary frequency modulation.

[0157]

[0158] Specifically, step S4 includes the following steps:

[0159] S41: Input relevant data of the power grid, including the objective function parameters of the multi-objective optimization model and the relevant variable constraints, and further derive the decision variables of the power grid optimization scheduling strategy;

[0160] S42: Solving multi-objective models;

[0161] S421: Initialize the white shark population: Set the maximum number of iterations, introduce initialization to the decision variables to generate a group of white shark individuals, each representing a power grid interactive optimization scheduling strategy. The initialization formula is as follows:

[0162]

[0163] Where, x i,j Let represent the position information of the i-th shark individual in the j-th decision variable dimension, where LB is the lower bound and UB is the upper bound. rand is a random function.

[0164] S422: Fitness Assessment: Calculate the fitness of each individual great white shark, taking constraints into account. The formula is as follows:

[0165]

[0166] S423: Non-dominated sorting and external archive set management: Sort the population according to Pareto dominance, mark non-dominated solutions, store non-dominated solutions in an external archive set, and after each iteration, compare the newly generated non-dominated solutions with the original solutions in the archive set, retain all non-dominated solutions, and if the archive set exceeds the capacity limit, use the crowding distance to replace them.

[0167]

[0168] Where, x i Let i be the i-th shark individual, and k be the number of objective functions. Let p be the function value of the individual on the p-th objective function. This represents the maximum value in the current population. It is the minimum value in the current population;

[0169] S424: Random values ​​are used to determine the update of individual shark particles;

[0170] If rand < 0.5, perform a global search for individual sharks:

[0171] Random relocation:

[0172]

[0173] Where D is the number of dimensions of the decision variables;

[0174] Wave movement:

[0175]

[0176] S425: If rand > 0.5, perform local development of the individual shark:

[0177] Levy flight:

[0178]

[0179] Where L is the Levy flight step size coefficient. The location of a random reference individual;

[0180] Cooperative predation:

[0181]

[0182] S428: The secondary local search mechanism includes: first, selecting 20% ​​of individuals and a non-dominated solution, then comparing fitness and using differential mutation to maintain diversity for individuals with higher fitness; and using elite attraction and Gaussian perturbation to maintain convergence for individuals with low fitness.

[0183] If the new solution is better than the current solution, then accept it:

[0184]

[0185] Otherwise, move in the opposite direction to the inferior solution:

[0186]

[0187] Population updates are achieved by merging populations and archives, performing non-dominated sorting, and then filling the next generation of populations according to the frontier level, ultimately obtaining a high-quality, highly diverse solution set of approximate Pareto optimal solutions.

[0188] S43: Finally, a high-quality, highly diverse set of solutions for approximate Pareto optimality is obtained, from which users can select the appropriate solution based on their actual needs.

[0189] Obviously, for those skilled in the art, this invention is not limited to the specific details involved in the above exemplary embodiments. Other specific forms may be derived without departing from the spirit or essential characteristics of the invention. These embodiments are illustrative only and not limiting. The scope of the invention is defined by the appended claims and is intended to cover all variations within the meaning and scope of the equivalents of the claims. Reference numerals in the claims do not affect the scope of the claims. The specification describes embodiments, but the technical solutions of the various embodiments can be combined to form other embodiments that will be understood by those skilled in the art.

Claims

1. A multi-objective scheduling strategy for power grid interaction that takes into account pumped storage start-up and shutdown strategies, characterized in that, The process includes the following steps: S1: Modeling the output characteristics of distributed resources in the power grid; S2: Modeling based on the power constraints of pumped storage start-up and shutdown strategies; S3: Constructing a multi-objective optimization scheduling model for power grid interaction that takes into account both pumped storage operation strategies and power grid peak shaving and frequency regulation. S4: Power grid optimization scheduling is performed using an improved multi-objective white shark optimization algorithm. Based on Pareto theory, the original white shark optimization algorithm is extended to multiple objectives, and a basic multi-objective white shark optimization algorithm framework is designed. Levy flight is added to the cooperative predation in the position update stage to enhance population diversity and avoid local optima. Secondary updates are performed on the population during the population update process to conduct local search.

2. The multi-objective scheduling strategy for power grid interaction considering pumped storage start-up and shutdown strategies according to claim 1, characterized in that: In S1, power data from the load side and generator side of the power system are acquired, and constraint models for the system's output characteristics are established respectively; S11: 24-hour wind power generation, photovoltaic power generation, and total load power are acquired to obtain 24-hour net load data; S12: Constraints are added to the power system, specifically including: system power balance constraints: Where, N c N w N p N phs These represent the number of each type of thermal power unit, wind power unit, photovoltaic unit, and pumped storage unit, respectively. c P w P p P phs P load These represent the output power of thermal power units, wind turbine generator units, photovoltaic generator units, pumped storage units, and total load power; thermal power unit ramping constraints: .in, , These represent the output power of the i-th thermal power unit at time t and the output power of the thermal power unit at time t-1, respectively; output constraints for wind turbines: Among them, P w This refers to the output power of wind power generation. Photovoltaic generator output constraints: Among them, P p This refers to the output power of wind power generation. S13: Model the operating costs of each power generation device and the volatility of thermal power units, specifically including: power generation cost of thermal power units: Among them, P i,t Let C be the output power of the i-th thermal power unit at time t. thermal Total power generation cost of thermal power units; Operation and maintenance cost of pumped storage hydroelectric power units: Among them, P phs,t Let C be the output power of the pumped storage unit at time t. PHS This refers to the operation and maintenance costs of pumped storage hydroelectric power plants. Total operating costs: Variance of thermal power unit output:

3. The multi-objective scheduling strategy for power grid interaction considering pumped storage start-up and shutdown strategies according to claim 1, characterized in that: In S2, the actual operation of pumped storage involves two start-up and shutdown actions: startup and shutdown. Start-up is further divided into pumping and power generation. Therefore, 1, 0, and -1 are introduced to represent their respective operating conditions. Based on the actual operating characteristics of pumped storage, a pumped storage start-up and shutdown strategy is generated by classifying the indicators and allocating power. The steps include: S21: Adding output constraints to the pumped storage unit; in, Let be the reservoir capacity at time t. Let t be the generator power of the pumped storage unit. Let t be the power of the pump unit of the pumped storage unit. For unit efficiency; S22: Calculate the load fluctuation component and net load standard deviation from the net load data. The degree of net load fluctuation in each time period can be known from the above two data. S23: J is obtained by linear weighting the above two indicators. Start-up and shutdown thresholds are set based on J. The distribution of J in each time period yields three levels: strong, medium, and weak. The PHS operating status is divided into 7 intervals, namely 3 discharge levels, 3 charging levels, and 1 shutdown state. The optimal peak-shaving configuration of the power system is obtained through a multi-objective improved optimization algorithm.

4. A multi-objective grid interaction scheduling strategy considering pumped storage start-up and shutdown strategies as described in claim 1, characterized in that: In S3, after the overall peak-shaving power is determined, frequency deviations are prone to occur during periods of rapid load increase or decrease, requiring more frequent adjustments. Therefore, frequency adjustments are performed between 12:00 and 13:00 and between 18:00 and 19:

00. The process includes the following steps: S31: Acquire the raw data of the corresponding load data, wind power generation data, and photovoltaic power generation data for each 1-minute time period; S32: Perform primary frequency regulation droop control on the deviation frequency, set the droop coefficient R and the energy storage participation capacity to obtain the primary frequency regulation coefficient, and the pumped storage system quickly adjusts the power according to the droop characteristics to offset the frequency difference; Among them, P part For participation capacity; S33: Primary frequency modulation can only suppress frequency deviation, not completely eliminate it. Therefore, an auxiliary frequency modulation module is added to eliminate the blind spot that cannot be adjusted in droop control. Through the integral element, the steady-state frequency difference is gradually eliminated, restoring the frequency to the rated value. Where P0 is the AGC baseline and I(t) is the integral state; S34: The total power of primary frequency modulation for the corresponding time period can be obtained by adding the power obtained from the above droop control and the auxiliary frequency modulation.

5. A multi-objective grid interaction scheduling strategy considering pumped storage start-up and shutdown strategies according to claim 1, characterized in that: In S4, the improved multi-objective white shark optimization algorithm includes several aspects: initializing the white shark population, fitness evaluation, random relocation, wave-like swimming search, cooperative predation search, and population update. The specific steps are as follows: Initializing the white shark population: Set the maximum number of iterations, introduce initialization to the decision variables to generate a group of white shark individuals, each representing a power grid interactive optimization scheduling strategy. The initialization formula is as follows: Where, x i,j Let represent the position information of the i-th shark individual in the j-th decision variable dimension, where LB is the lower bound and UB is the upper bound. `rand` is a random function. Fitness evaluation: Calculate the fitness of each white shark individual, considering constraints. The formula is as follows: Non-dominated sorting and external archive management: Sort the population according to Pareto dominance, mark non-dominated solutions, and store non-dominated solutions in an external archive. After each iteration, compare the newly generated non-dominated solutions with the original solutions in the archive, retain all non-dominated solutions, and replace them with crowding distance if the archive exceeds the capacity limit. Where, x i Let i be the i-th shark individual, and k be the number of objective functions. Let p be the function value of the individual on the p-th objective function. This represents the maximum value in the current population. The minimum value in the current population is used; random values ​​are used to determine the update of individual shark particles; if rand < 0.5, a global search for individual sharks is performed: random relocation is then implemented. Where D is the number of dimensions of the decision variables; wave motion: If rand > 0.5, execute individual shark development: Levy flight: Where L is the Levy flight step size coefficient. The location of random reference individuals; cooperative predation: The secondary local search mechanism includes: first, selecting 20% ​​of individuals and a non-dominated solution; then comparing their fitness; based on the results, differential mutation is used to maintain diversity for individuals with higher fitness; elite attraction and Gaussian perturbation are used to maintain convergence for individuals with low fitness; if the new solution is better than the current solution, it is accepted. Otherwise, move in the opposite direction to the inferior solution: As the number of iterations increases, the output power of each unit carried by the shark population is updated and optimized. After merging the population and archiving, the population is sorted non-dominated and then the next generation of population is filled according to the leading level. This yields a high-quality, highly diverse solution set of approximate Pareto optimal solutions, thus obtaining the optimal output of the multi-energy complementary power system.

Images