Optimization method and optimization device for energy storage charging and discharging scheduling algorithm

Abstract

The invention relates to the technical field of energy storage, in particular to an optimization method and optimization device for an energy storage charge and discharge scheduling algorithm, and the method comprises the steps: generating a light-temperature coupling feature vector through collecting photovoltaic power and temperature control load data, dynamically judging a supercritical state based on the vector, and loading corresponding mode parameters and slack variables, a dynamic weight MPC model is constructed according to mode parameters and slack variables, a historical optimal solution is used for preheating an initial solution to accelerate model prediction control solution, the slack variables are dynamically adjusted in combination with power grid frequency deviation feedback, supercritical criterion conditions are optimized in a closed-loop mode, and rapid generation and dynamic correction of an energy storage scheduling strategy in extreme weather are achieved. The problem of response lag caused by mode switching delay and parameter solidification in a traditional method is solved, the real-time performance of power supply guarantee and the strategy reliability are remarkably improved, and meanwhile the economical efficiency target and the power grid stability requirement are considered.

Description

Technical Field

[0001] This invention belongs to the field of energy storage technology, specifically an optimization method and device for energy storage charging and discharging scheduling algorithm. Background Technology

[0002] With the increasing prevalence of new energy power generation in high-proportion grid connections, the intermittency and volatility of renewable energy sources such as photovoltaic and wind power place higher demands on the dispatch capabilities of energy storage systems. Especially in extreme weather scenarios, such as short-term severe convective weather causing a sudden drop in photovoltaic output, or high temperatures triggering a surge in temperature-controlled loads, the grid needs to achieve power balance and power supply guarantee within a time scale of seconds.

[0003] Traditional model predictive control algorithms generate charging and discharging strategies through multi-objective optimization. However, when dealing with such sudden scenarios, the optimization objectives need to be switched frequently. Existing technologies usually use fixed weight adjustment or manual intervention mode switching, which causes the solver to have computational delays due to objective function reconstruction and constraint changes, making it difficult to respond to dynamic changes in a timely manner.

[0004] In addition, existing methods lack quantitative analysis of the coupling effects of extreme weather at the feature extraction level. For example, the synergistic effect of a sudden drop in photovoltaic output and a sudden change in temperature control load has not been modeled as a unified criterion, which increases the risk of delayed or misjudged mode switching decisions.

[0005] On the other hand, the lack of closed-loop verification and parameter self-optimization mechanisms for energy storage scheduling strategies leads to the ineffective use of historical operating data, making it difficult to dynamically correct strategy deviations.

[0006] Therefore, there is an urgent need for an energy storage scheduling optimization method that can integrate multi-source feature dynamic judgment, accelerate model solving, and achieve parameter adaptation, in order to solve the contradiction between real-time performance and reliability under extreme weather conditions. Summary of the Invention

[0007] The purpose of this invention is to address the aforementioned shortcomings in the prior art by providing an optimization method and device for energy storage charging and discharging scheduling algorithms.

[0008] The objective of this invention is achieved through the following technical solution: an optimization method for an energy storage charging and discharging scheduling algorithm, comprising the following steps: S1. Based on the collected photovoltaic power data and temperature control load data, generate a light-temperature coupling feature vector; S2. Based on the light-temperature coupling feature vector, determine whether a supercritical state is triggered through predefined supercritical judgment conditions, and load the corresponding scheduling mode parameters and relaxation variables; wherein, the scheduling mode parameters are used to adjust the economic priority and power supply guarantee priority of energy storage charging and discharging, and the relaxation variables are used to adjust the power balance constraints of the power grid. S3. Based on the scheduling mode parameters and the power balance constraints, a dynamic weight model is constructed; the initial solution of the dynamic weight model is preheated using the historical optimal solution that matches the current scheduling mode parameters, and the charging and discharging power command is obtained by solving the dynamic weight model; the scheduling mode parameters include a supply guarantee mode and an economic mode. S4. Execute the charging and discharging power command, dynamically adjust the relaxation variable according to the grid frequency deviation, and return to step S1 to re-execute; S5. Based on the records of power grid frequency deviation and supercritical events, optimize the supercritical judgment conditions and update the backup strategy library; the backup strategy library stores the scheduling mode parameters, the slack variables, the supercritical judgment conditions, and the historical optimal solutions.

[0009] The present invention is further configured such that generating the light-temperature coupling feature vector based on the collected photovoltaic power data and temperature control load data includes: A1. Collect photovoltaic power and temperature-controlled load power at a preset cycle; A2. Calculate the rate of change of photovoltaic power to obtain the instantaneous rate of change of photovoltaic power. The calculation formula is as follows: in The instantaneous rate of change of photovoltaic power. For photovoltaic power, For the preset time window, For time; A3. The temperature-controlled load power undergoes abrupt change detection to generate a temperature-controlled load abrupt change marker. ;when At that time, the temperature control load sudden change flag quantity The value is 1 if it is 1 otherwise it is 0; where, Let k be the power of the temperature-controlled load, and k be a preset mutation coefficient. The time constant is the thermal inertia. A4. Combine the instantaneous change rate of photovoltaic power and the sudden change flag of temperature control load to obtain the light-temperature coupling feature vector.

[0010] The present invention is further configured such that, based on the light-temperature coupling feature vector, a predefined supercritical judgment condition is used to determine whether a supercritical state is triggered, and corresponding scheduling mode parameters and relaxation variables are loaded; wherein, the scheduling mode parameters are used to adjust the economic priority and power supply guarantee priority of energy storage charging and discharging, and the relaxation variables are used to adjust the power balance constraints of the power grid, including: B1. Based on the aforementioned light-temperature coupling feature vector, if and If a condition is triggered, a supercritical state is established; otherwise, the state is considered normal. The supercritical judgment condition is the power drop threshold. and ; B2. If a supercritical state is triggered, load the corresponding scheduling mode parameters and relaxation variables from the backup strategy library; wherein, in the supercritical state, the scheduling mode parameters are the supply guarantee mode. B3. If it is the normal state, load the corresponding scheduling mode parameters and slack variables from the backup strategy library; wherein, in the normal state, the scheduling mode parameters are the economy mode; The scheduling mode parameters include a power supply guarantee priority coefficient and an economic priority coefficient; if the scheduling mode parameters are for the power supply guarantee mode, the power supply guarantee priority coefficient is greater than the economic priority coefficient; if the scheduling mode parameters are for the economic mode, the power supply guarantee priority coefficient is less than the economic priority coefficient.

[0011] The present invention is further configured such that, based on the scheduling mode parameters and the power balance constraints, a dynamic weight model is constructed; the initial solution of the dynamic weight model is preheated using the historical optimal solution matching the current scheduling mode parameters; and the charging and discharging power command is obtained by solving the dynamic weight model; the scheduling mode parameters include a supply guarantee mode and an economic mode, including: C1. Construct an objective function based on the power supply guarantee priority coefficient and the economic priority coefficient: in, For discharge power, For charging power, For time intervals, For grid interaction power, The real-time electricity price of the power grid, α is the power supply guarantee priority coefficient, and β is the economic priority coefficient; C2. Define the power balance constraint condition based on the slack variables: in, For the power of the energy storage system, The power of the temperature-controlled load, For the slack variable; C3. Based on the objective function and the power balance constraints, the dynamic weight model is obtained; C4. Retrieve the historical best solution from the backup strategy library that matches the current scheduling mode parameters. The historical best solution is used as the initial solution for the dynamic weight model; where, This is the optimal power solution for the energy storage system, including the optimal charging power solution and the optimal discharging power solution; This is the optimal solution for power grid interaction. C5. Solve the dynamic weight model using the branch and bound method, and output the charging and discharging power command.

[0012] The present invention is further configured such that the historical optimal solution matching the current scheduling mode parameters is retrieved from the backup strategy library. The historical best solution is used as the initial solution for the dynamic weight model; where, This is the optimal power solution for the energy storage system, including the optimal charging power solution and the optimal discharging power solution; The optimal solution for grid interaction power includes: D1. Read the historical optimal solution from the backup strategy library. ; D2. Verify whether the historical optimal solution satisfies the power balance constraint condition. ; D3. If the verification passes, the historical best solution will be used as the initial solution; otherwise, the default initial solution will be used.

[0013] The present invention is further configured such that dynamically adjusting the slack variable according to the power grid frequency deviation includes: E1. Acquire the power grid frequency and calculate the power grid frequency deviation based on the power grid frequency. ,in, The power grid frequency, The rated frequency of the power grid; E2. If the power grid frequency deviation is greater than or equal to the frequency deviation threshold and continues for a preset duration, the relaxation variable is adjusted according to the following formula: in, For the adjusted slack variables, The slack variables before adjustment , representing the cumulative amount of frequency deviation exceeding the threshold; The frequency deviation threshold is... For the preset duration, This represents the cumulative number of adjustments made to the slack variable during the current event, where the event is either in the supercritical state or in the normal state. γ is the preset maximum number of adjustments allowed, λ is the frequency deviation sensitivity coefficient, and λ is the adjustment attenuation coefficient.

[0014] The present invention is further configured such that, based on the records of power grid frequency deviation and supercritical events, the supercritical judgment conditions are optimized and the backup strategy library is updated; the backup strategy library stores the scheduling mode parameters, the slack variables, the supercritical judgment conditions, and the historical optimal solutions, including: F1. In events categorized as being in a supercritical state, record the number of times the grid frequency deviation is greater than or equal to the frequency deviation threshold, and record these as supercritical events. F2. If the number of times the same event is triggered reaches the preset number of triggers, the current supercritical judgment condition is determined to be inaccurate. F3. When the supercriticality criterion is inaccurate, update the power descent threshold in the supercriticality criterion using the gradient descent method; the update formula for the power descent threshold is: in, The updated power descent threshold. The previous power sag threshold was used. To preset the learning rate, This represents the actual frequency deviation. This is due to prediction bias based on historical data; F4. Write the updated power descent threshold to the backup policy library.

[0015] An optimization device for an energy storage charging and discharging scheduling algorithm, comprising: The light-temperature coupling module is used to generate a light-temperature coupling feature vector based on the collected photovoltaic power data and temperature control load data; The mode management module is used to determine whether a supercritical state is triggered based on the light-temperature coupling feature vector and predefined supercritical judgment conditions, and to load the corresponding scheduling mode parameters and relaxation variables; wherein, the scheduling mode parameters are used to adjust the economic priority and power supply guarantee priority of energy storage charging and discharging, and the relaxation variables are used to adjust the power balance constraints of the power grid. The dynamic optimization module is used to construct a dynamic weight model based on the scheduling mode parameters and the power balance constraints; to preheat the initial solution of the dynamic weight model using the historical best solution that matches the current scheduling mode parameters; and to obtain the charging and discharging power command by solving the dynamic weight model; the scheduling mode parameters include a supply guarantee mode and an economic mode. The power regulation module is used to execute the charging and discharging power command, dynamically adjust the relaxation variable according to the grid frequency deviation, and re-execute the operations of the optical-temperature coupling module, the mode management module, and the dynamic optimization module. The strategy optimization module is used to optimize the supercritical judgment conditions and update the backup strategy library based on the records of power grid frequency deviation and supercritical events; the backup strategy library stores the scheduling mode parameters, the slack variables, the supercritical judgment conditions, and the historical optimal solutions.

[0016] The beneficial effects of this invention are as follows: This invention generates a light-temperature coupled feature vector by collecting photovoltaic power and temperature-controlled load data. Based on this vector, the supercritical state is dynamically determined and corresponding mode parameters and relaxation variables are loaded. A dynamic weighted MPC model is constructed according to the mode parameters and relaxation variables. The model predictive control solution is accelerated by using the historical best solution to preheat the initial solution. Combined with the grid frequency deviation feedback, the relaxation variables are dynamically adjusted and the supercritical criterion conditions are optimized in a closed loop. This enables the rapid generation and dynamic correction of energy storage scheduling strategies under extreme weather conditions, solves the response lag problem caused by mode switching delay and parameter solidification in traditional methods, significantly improves the real-time performance and strategy reliability of power supply security, and takes into account both economic objectives and grid stability requirements. Attached Figure Description

[0017] The invention will be further illustrated with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the invention. For those skilled in the art, other drawings can be obtained based on the following drawings without any creative effort.

[0018] Figure 1 This is a flowchart of the present invention; Detailed Implementation

[0019] The present invention will be further described in conjunction with the following embodiments.

[0020] Depend on Figure 1 As can be seen, this embodiment provides an optimization method for an energy storage charging and discharging scheduling algorithm, including the following steps: S1. Based on the collected photovoltaic power data and temperature control load data, a photovoltaic-temperature coupling feature vector is generated. High-precision sensors can be used to collect photovoltaic power data (such as the DC power output of the photovoltaic array) and temperature control load data (such as the real-time power consumption of air conditioners and cooling equipment) in real time. This data can be stored in a local database in time-series format. When generating the photovoltaic-temperature coupling feature vector, multivariate feature extraction algorithms, such as principal component analysis (PCA) or an autoencoder, can be used to reduce the dimensionality of the original data and extract key features that characterize the coupling relationship between photovoltaic power output fluctuations and temperature control load abrupt changes.

[0021] S2. Based on the light-temperature coupling feature vector, determine whether a supercritical state has been triggered using predefined supercritical judgment conditions, and load the corresponding scheduling mode parameters and relaxation variables. The scheduling mode parameters are used to adjust the economic priority and power supply guarantee priority of energy storage charging and discharging, while the relaxation variables are used to adjust the power balance constraints of the power grid. The determination of the supercritical state is based on predefined supercritical judgment conditions, which can be obtained through training on historical extreme weather scenario data. Specifically, a support vector machine (SVM) or a deep neural network (DNN) can be used as a classifier to map the current light-temperature coupling feature vector to the supercritical state space. When a supercritical state is determined, the system loads the corresponding scheduling mode parameters and relaxation variables from the backup strategy library. The scheduling mode parameters include economic weights (such as the cost per kilowatt-hour of energy storage charging and discharging) and power supply guarantee weights (such as the priority coefficient of power grid frequency stability), while the relaxation variables are used to dynamically adjust the power balance constraints (such as the allowable power deviation range).

[0022] S3. Based on the scheduling mode parameters and the power balance constraints, a dynamic weighting model is constructed. The initial solution of the dynamic weighting model is preheated using historical optimal solutions matching the current scheduling mode parameters, and the charging / discharging power command is obtained by solving the dynamic weighting model. The scheduling mode parameters include a supply guarantee mode and an economic mode. The core of the dynamic weighting model lies in integrating economic and supply guarantee objectives into a unified optimization framework. In specific implementation, economic objectives (such as the economic cost of energy storage charging and discharging) and supply guarantee objectives (such as the guarantee cost of grid frequency stability) can be weighted and summed using dynamic weighting coefficients. The dynamic weighting coefficients can be adjusted in real time according to the current grid state and energy storage state to balance the priorities between economic efficiency and supply guarantee.

[0023] S4. Execute the charging / discharging power command, dynamically adjust the relaxation variable according to the grid frequency deviation, and return to step S1 to re-execute. Specifically, the charging / discharging power command can be executed through a power electronic interface (such as a bidirectional inverter), while simultaneously monitoring the grid frequency deviation in real time (e.g., through PMU measurement). When the frequency deviation exceeds a preset threshold, the relaxation variable can be dynamically adjusted through the following mechanism: mapping the frequency deviation to the adjustment range of the relaxation variable, for example, using a PID controller or fuzzy logic inference; the adjusted relaxation variable re-triggers the iterative process from S1 to S3 to ensure the system remains stable in a dynamic environment.

[0024] S5. Based on the records of power grid frequency deviation and supercritical events, optimize the supercritical judgment conditions and update the backup strategy library; the backup strategy library stores the scheduling mode parameters, the slack variables, the supercritical judgment conditions, and the historical optimal solutions. Based on the operational data of each supercritical event, Bayesian optimization or a genetic algorithm can be used to optimize the threshold parameters of the supercritical judgment conditions. Simultaneously, the newly generated scheduling mode parameters, slack variables, and historical optimal solutions are updated to the backup strategy library, forming a closed-loop mechanism for continuous learning.

[0025] This embodiment generates a light-temperature coupled feature vector by collecting photovoltaic power and temperature-controlled load data. Based on this vector, the supercritical state is dynamically determined and corresponding mode parameters and relaxation variables are loaded. A dynamic weighted MPC model is constructed according to the mode parameters and relaxation variables. The model predictive control solution is accelerated by using the historical best solution to preheat the initial solution. Combined with grid frequency deviation feedback, the relaxation variables are dynamically adjusted and the supercritical criterion conditions are optimized in a closed loop. This enables the rapid generation and dynamic correction of energy storage dispatch strategies under extreme weather conditions, solving the response lag problem caused by mode switching delay and parameter solidification in traditional methods. It significantly improves the real-time performance and strategy reliability of power supply security, while taking into account both economic objectives and grid stability requirements.

[0026] This embodiment provides an optimization method for an energy storage charging and discharging scheduling algorithm. The generation of a light-temperature coupling feature vector based on collected photovoltaic power data and temperature-controlled load data includes: A1. Collect photovoltaic power and temperature-controlled load power at a preset cycle; A2. Calculate the rate of change of photovoltaic power to obtain the instantaneous rate of change of photovoltaic power. The calculation formula is as follows: in The instantaneous rate of change of photovoltaic power. For photovoltaic power, For the preset time window, For time; A3. The temperature-controlled load power undergoes abrupt change detection to generate a temperature-controlled load abrupt change marker. ;when At that time, the temperature control load sudden change flag quantity The value is 1 if it is 1 otherwise it is 0; where, Let k be the power of the temperature-controlled load, and k be a preset mutation coefficient. The time constant is the thermal inertia. A4. Combine the instantaneous change rate of photovoltaic power and the sudden change flag of temperature control load to obtain the light-temperature coupling feature vector.

[0027] Specifically, this embodiment collects photovoltaic power and temperature-controlled load power data in real time at a preset period (e.g., seconds or minutes). This data is acquired through high-precision sensors and stored in a local database. The data acquisition period and accuracy can be optimized according to the dynamic response requirements of the power grid to ensure timely capture of power changes in extreme weather scenarios. The rate of change of the collected photovoltaic power data is calculated to quantify the volatility of photovoltaic output. The formula for calculating the rate of change is: .in, Indicates the instantaneous rate of change of photovoltaic power. The photovoltaic power at the current moment, A preset time window is used. This calculation can reflect the changing trend of photovoltaic power within the time window, such as the sudden drop in photovoltaic output during short-term severe convective weather. Abrupt change detection is performed on the power of temperature-controlled loads to generate a temperature-controlled load abrupt change marker. When the temperature-controlled load power meets the conditions At that time, the marker quantity The value is 1 if it is 1, otherwise it is 0. The current temperature-controlled load power is given by k, where k is the preset mutation coefficient. This is the thermal inertia time constant. This detection mechanism can identify abnormal surges in temperature-controlled loads, such as a sudden increase in air conditioning load during hot weather. The instantaneous rate of change of photovoltaic power is also considered. and temperature control load mutation marker These are combined to form a light-temperature coupling feature vector. This feature vector comprehensively reflects the coupling effect between photovoltaic power output fluctuations and temperature-controlled load abrupt changes, providing a quantitative basis for subsequent supercritical state determination.

[0028] This embodiment provides an optimization method for an energy storage charging and discharging scheduling algorithm. Based on the optical-thermal coupling feature vector, a predefined supercritical judgment condition is used to determine whether a supercritical state has been triggered, and corresponding scheduling mode parameters and relaxation variables are loaded. The scheduling mode parameters are used to adjust the economic priority and power supply guarantee priority of energy storage charging and discharging, and the relaxation variables are used to adjust the power balance constraints of the power grid, including: B1. Based on the aforementioned light-temperature coupling feature vector, if and If a condition is triggered, a supercritical state is established; otherwise, the state is considered normal. The supercritical judgment condition is the power drop threshold. and ; B2. If a supercritical state is triggered, load the corresponding scheduling mode parameters and relaxation variables from the backup strategy library; wherein, in the supercritical state, the scheduling mode parameters are the supply guarantee mode. B3. If it is the normal state, load the corresponding scheduling mode parameters and slack variables from the backup strategy library; wherein, in the normal state, the scheduling mode parameters are the economy mode; The scheduling mode parameters include a power supply guarantee priority coefficient and an economic priority coefficient; if the scheduling mode parameters are for the power supply guarantee mode, the power supply guarantee priority coefficient is greater than the economic priority coefficient; if the scheduling mode parameters are for the economic mode, the power supply guarantee priority coefficient is less than the economic priority coefficient.

[0029] Specifically, the power drop threshold can be pre-set based on historical extreme weather data and grid stability requirements to identify abnormal declines in photovoltaic output. The supercritical state determination mechanism ensures the system can respond quickly under extreme weather conditions, avoiding grid power imbalance. When a supercritical state is determined, the corresponding power supply guarantee mode parameters and relaxation variables are loaded from the backup strategy library. The power supply guarantee mode parameters indicate that in a supercritical state, the scheduling mode parameters prioritize power supply security, with priority over economic efficiency. Specifically, the power supply security priority coefficient is greater than the economic efficiency priority coefficient. Relaxation variables are used to adjust grid power balance constraints, allowing for a moderate relaxation of power balance requirements under extreme conditions to ensure power supply stability. After loading these parameters and variables, grid power supply stability will be prioritized, even if this means compromising short-term economic objectives. If a normal state is determined, the corresponding economic mode parameters and relaxation variables are loaded from the backup strategy library. The economic mode parameters indicate that in a normal state, the scheduling mode parameters prioritize economic objectives, with priority over power supply security. Specifically, the economic efficiency priority coefficient is greater than the power supply security priority coefficient. Under normal conditions, the slack variables can be set relatively strictly to ensure that the power balance constraints of the power grid are strictly satisfied, while pursuing the optimal economic efficiency of energy storage dispatch. After loading these parameters and variables, priority will be given to pursuing the economic efficiency of energy storage dispatch, while maintaining the stable operation of the power grid.

[0030] This embodiment provides an optimization method for an energy storage charging and discharging scheduling algorithm. Based on the scheduling mode parameters and the power balance constraints, a dynamic weight model is constructed. The initial solution of the dynamic weight model is preheated using historical optimal solutions matching the current scheduling mode parameters, and the charging and discharging power command is obtained by solving the dynamic weight model. The scheduling mode parameters include a supply guarantee mode and an economic mode, including: C1. Construct an objective function based on the power supply guarantee priority coefficient and the economic priority coefficient: in, For discharge power, For charging power, For time intervals, For grid interaction power, The real-time electricity price of the power grid, α is the power supply guarantee priority coefficient, and β is the economic priority coefficient; C2. Define the power balance constraint condition based on the slack variables: in, For the power of the energy storage system, The power of the temperature-controlled load, For the slack variable; C3. Based on the objective function and the power balance constraints, the dynamic weight model is obtained; C4. Retrieve the historical best solution from the backup strategy library that matches the current scheduling mode parameters. The historical best solution is used as the initial solution for the dynamic weight model; where, This is the optimal power solution for the energy storage system, including the optimal charging power solution and the optimal discharging power solution; This is the optimal solution for power grid interaction. C5. Solve the dynamic weight model using the branch and bound method, and output the charging and discharging power command.

[0031] Specifically, the system constructs an optimization objective function based on the power supply guarantee priority coefficient (α) and the economic priority coefficient (β). This objective function aims to minimize the operating cost of the energy storage system while ensuring a stable power supply from the grid. The objective function takes the following form: .in, The discharge power of the energy storage system. For charging power, For time intervals, For grid interaction power, This represents the real-time electricity price of the power grid. By adjusting the values ​​of α and β, a dynamic balance can be achieved between power supply security and economic efficiency. The system defines power balance constraints based on the slack variable (ε) to ensure the balance of power in the power grid. The constraint conditions are in the following form: .in, For the power of the energy storage system, For temperature-controlled load power, The slack variable (ε) allows for a moderate relaxation of power balance constraints under extreme conditions to ensure stable system operation. This constraint ensures that grid power fluctuates within an acceptable range, avoiding grid instability caused by power imbalance. Based on the above objective function and power balance constraints, a dynamic weight model is constructed. This model optimizes the charging and discharging strategy of the energy storage system by predicting the grid state over a future period, achieving an optimal balance between power supply security and economic efficiency. The system retrieves historical optimal solutions matching the current scheduling mode from the backup strategy library. These historical optimal solutions are used as the initial solutions for the dynamic weighting model. The historical optimal solutions include the energy storage system's power optimal solution (…). ) and the optimal solution for grid interaction power ( By using historical best solutions as initial guesses, the solution process for the dynamic weight model can be accelerated, improving real-time response capabilities. The branch-and-bound method is employed to solve the dynamic weight model, yielding the optimal charging and discharging power command. The branch-and-bound method systematically searches the solution space to ensure the finding of the global optimum while avoiding getting trapped in local optima.

[0032] This embodiment provides an optimization method for an energy storage charging and discharging scheduling algorithm, wherein the method involves calling the historical optimal solution from the backup strategy library that matches the current scheduling mode parameters. The historical best solution is used as the initial solution for the dynamic weight model; where, This is the optimal power solution for the energy storage system, including the optimal charging power solution and the optimal discharging power solution; The optimal solution for grid interaction power includes: D1. Read the historical optimal solution from the backup strategy library. ; D2. Verify whether the historical optimal solution satisfies the power balance constraint condition. ; D3. If the verification passes, the historical best solution will be used as the initial solution; otherwise, the default initial solution will be used.

[0033] Specifically, the system reads historical optimal solutions from the backup strategy library, including the energy storage system power optimal solution (…). ) and the optimal solution for grid interaction power ( These historical optimal solutions are generated based on operational data from similar past scenarios and are stored in a backup strategy library for quick recall in future scheduling. After reading the historical optimal solutions, it is verified whether these solutions satisfy the current power balance constraints: .in, and It is the historical best solution. ϵ is the current temperature-controlled load power, and ϵ is a relaxation variable. The purpose of verification is to ensure that the historical optimal solution remains effective under the current grid conditions, avoiding strategy failure due to changes in grid conditions. This verification mechanism checks whether the allowable power deviation range is met by substituting the historical optimal solution into the current power balance constraints. Based on the verification results, it is decided whether to use the historical optimal solution as the initial solution of the dynamic weighted MPC model. If the historical optimal solution meets the power balance constraints, it is used as the initial solution of the MPC model. This helps to accelerate the model's solution process because the initial solution is close to the optimal solution. If the historical optimal solution does not meet the constraints, the default initial solution is used. The default initial solution can be set based on basic operating parameters and safety constraints to ensure that the model can solve stably without historical data reference.

[0034] This embodiment provides an optimization method for an energy storage charging and discharging scheduling algorithm, wherein dynamically adjusting the relaxation variable based on the grid frequency deviation includes: E1. Acquire the power grid frequency and calculate the power grid frequency deviation based on the power grid frequency. ,in, The power grid frequency, The rated frequency of the power grid; E2. If the power grid frequency deviation is greater than or equal to the frequency deviation threshold and continues for a preset duration, the relaxation variable is adjusted according to the following formula: in, For the adjusted slack variables, The slack variables before adjustment , representing the cumulative amount of frequency deviation exceeding the threshold; The frequency deviation threshold is... For the preset duration, This represents the cumulative number of adjustments made to the slack variable during the current event, where the event is either in the supercritical state or in the normal state. γ is the preset maximum number of adjustments allowed, λ is the frequency deviation sensitivity coefficient, and λ is the adjustment attenuation coefficient.

[0035] Specifically, by dynamically adjusting slack variables, the system can moderately relax power balance constraints when the grid frequency deviation is large, ensuring the stable operation of the grid. At the same time, by limiting the cumulative deviation and the number of adjustments, the system can gradually restore strict power balance constraints after the frequency deviation returns to normal, avoiding performance degradation caused by over-adjustment.

[0036] This embodiment provides an optimization method for an energy storage charging and discharging scheduling algorithm. The method involves optimizing the supercriticality judgment condition and updating the backup strategy library based on records of grid frequency deviation and supercritical events. The backup strategy library stores the scheduling mode parameters, the relaxation variables, the supercriticality judgment condition, and the historical optimal solution, including: F1. In events categorized as being in a supercritical state, record the number of times the grid frequency deviation is greater than or equal to the frequency deviation threshold, and record these as supercritical events. F2. If the number of times the same event is triggered reaches the preset number of triggers, the current supercritical judgment condition is determined to be inaccurate. F3. When the supercriticality criterion is inaccurate, update the power descent threshold in the supercriticality criterion using the gradient descent method; the update formula for the power descent threshold is: in, The updated power descent threshold. The previous power sag threshold was used. To preset the learning rate, This represents the actual frequency deviation. This is due to prediction bias based on historical data; F4. Write the updated power descent threshold to the backup policy library.

[0037] Specifically, in the same supercritical event, if the number of triggers reaches a preset trigger count, the current supercritical judgment condition is determined to be inaccurate. A threshold preset based on historical data and system design requirements can be used to determine whether the current judgment condition needs optimization. By comparing the actual number of triggers with the preset trigger count, it is possible to identify whether the judgment condition is too sensitive or inaccurate, thereby triggering the optimization process. When the supercritical judgment condition is determined to be inaccurate, the system updates the power descent threshold using the gradient descent method. The update formula is as follows: in, The updated power descent threshold. The previous power sag threshold was used. To preset the learning rate, This represents the actual frequency deviation. This addresses the prediction bias based on historical data. By calculating the gradient between the predicted and actual biases, the power descent threshold is dynamically adjusted to minimize the prediction error. Continuously optimizing the power descent threshold improves the accuracy and adaptability of supercritical state detection, ensuring stable system operation under various scenarios.

[0038] This embodiment provides an optimization device for an energy storage charging and discharging scheduling algorithm, comprising: The light-temperature coupling module is used to generate a light-temperature coupling feature vector based on the collected photovoltaic power data and temperature control load data; The mode management module is used to determine whether a supercritical state is triggered based on the light-temperature coupling feature vector and predefined supercritical judgment conditions, and to load the corresponding scheduling mode parameters and relaxation variables; wherein, the scheduling mode parameters are used to adjust the economic priority and power supply guarantee priority of energy storage charging and discharging, and the relaxation variables are used to adjust the power balance constraints of the power grid. The dynamic optimization module is used to construct a dynamic weight model based on the scheduling mode parameters and the power balance constraints; to preheat the initial solution of the dynamic weight model using the historical best solution that matches the current scheduling mode parameters; and to obtain the charging and discharging power command by solving the dynamic weight model; the scheduling mode parameters include a supply guarantee mode and an economic mode. The power regulation module is used to execute the charging and discharging power command, dynamically adjust the relaxation variable according to the grid frequency deviation, and re-execute the operations of the optical-temperature coupling module, the mode management module, and the dynamic optimization module. The strategy optimization module is used to optimize the supercritical judgment conditions and update the backup strategy library based on the records of power grid frequency deviation and supercritical events; the backup strategy library stores the scheduling mode parameters, the slack variables, the supercritical judgment conditions, and the historical optimal solutions.

[0039] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. An optimization method for an energy storage charging and discharging scheduling algorithm, characterized in that: Includes the following steps: S1. Based on the collected photovoltaic power data and temperature control load data, generate a light-temperature coupling feature vector; S2. Based on the light-temperature coupling feature vector, determine whether a supercritical state is triggered through predefined supercritical judgment conditions, and load the corresponding scheduling mode parameters and relaxation variables; wherein, the scheduling mode parameters are used to adjust the economic priority and power supply guarantee priority of energy storage charging and discharging, and the relaxation variables are used to adjust the power balance constraints of the power grid. S3. Based on the scheduling mode parameters and the power balance constraints, a dynamic weight model is constructed; the initial solution of the dynamic weight model is preheated using the historical optimal solution that matches the current scheduling mode parameters, and the charging and discharging power command is obtained by solving the dynamic weight model; the scheduling mode parameters include a supply guarantee mode and an economic mode. S4. Execute the charging and discharging power command, dynamically adjust the relaxation variable according to the grid frequency deviation, and return to step S1 to re-execute; S5. Based on the records of power grid frequency deviation and supercritical events, optimize the supercritical judgment conditions and update the backup strategy library; the backup strategy library stores the scheduling mode parameters, the slack variables, the supercritical judgment conditions, and the historical optimal solutions.

2. The optimization method of the energy storage charging and discharging scheduling algorithm according to claim 1, characterized in that: The generation of the light-temperature coupling feature vector based on the collected photovoltaic power data and temperature control load data includes: A1. Collect photovoltaic power and temperature-controlled load power at a preset cycle; A2. Calculate the rate of change of photovoltaic power to obtain the instantaneous rate of change of photovoltaic power. The calculation formula is as follows: in The instantaneous rate of change of photovoltaic power. For photovoltaic power, For the preset time window, For time; A3. The temperature-controlled load power undergoes abrupt change detection to generate a temperature-controlled load abrupt change marker. ;when At that time, the temperature control load sudden change flag quantity The value is 1 if it is 1 otherwise it is 0; where, The power of the temperature-controlled load is given by k, which is a preset mutation coefficient. The time constant is the thermal inertia. A4. Combine the instantaneous change rate of photovoltaic power and the sudden change flag of temperature control load to obtain the light-temperature coupling feature vector.

3. The optimization method of the energy storage charging and discharging scheduling algorithm according to claim 2, characterized in that: The process involves determining whether a supercritical state has been triggered based on the light-temperature coupling feature vector and predefined supercritical judgment conditions, and then loading corresponding scheduling mode parameters and relaxation variables. The scheduling mode parameters are used to adjust the economic priority and power supply guarantee priority of energy storage charging and discharging, and the relaxation variables are used to adjust the power balance constraints of the power grid, including: B1. Based on the aforementioned light-temperature coupling feature vector, if and If a condition is triggered, a supercritical state is established; otherwise, the state is considered normal. The supercritical judgment condition is the power drop threshold. and ; B2. If a supercritical state is triggered, load the corresponding scheduling mode parameters and relaxation variables from the backup strategy library; wherein, in the supercritical state, the scheduling mode parameters are the supply guarantee mode. B3. If it is the normal state, load the corresponding scheduling mode parameters and slack variables from the backup strategy library; wherein, in the normal state, the scheduling mode parameters are the economy mode; The scheduling mode parameters include a power supply guarantee priority coefficient and an economic priority coefficient; if the scheduling mode parameters are for the power supply guarantee mode, the power supply guarantee priority coefficient is greater than the economic priority coefficient; if the scheduling mode parameters are for the economic mode, the power supply guarantee priority coefficient is less than the economic priority coefficient.

4. The optimization method of the energy storage charging and discharging scheduling algorithm according to claim 3, characterized in that: A dynamic weighting model is constructed based on the scheduling mode parameters and the power balance constraints. The initial solution of the dynamic weighting model is preheated using historical optimal solutions that match the current scheduling mode parameters, and the charging / discharging power command is obtained by solving the dynamic weighting model. The scheduling mode parameters include a supply guarantee mode and an economic mode, including: C1. Construct an objective function based on the power supply guarantee priority coefficient and the economic priority coefficient: in, For discharge power, For charging power, For time intervals, For grid interaction power, The real-time electricity price of the power grid, α is the power supply guarantee priority coefficient, and β is the economic priority coefficient; C2. Define the power balance constraint condition based on the slack variables: in, For the power of the energy storage system, The power of the temperature-controlled load, For the slack variable; C3. Based on the objective function and the power balance constraints, the dynamic weight model is obtained; C4. Retrieve the historical best solution from the backup strategy library that matches the current scheduling mode parameters. The historical optimal solution is used as the initial solution for the dynamic weight model; where, This is the optimal power solution for the energy storage system, including the optimal charging power solution and the optimal discharging power solution; This is the optimal solution for power grid interaction. C5. Solve the dynamic weight model using the branch and bound method, and output the charging and discharging power command.

5. The optimization method of the energy storage charging and discharging scheduling algorithm according to claim 4, characterized in that: The step of retrieving the historical best solution from the backup strategy library that matches the current scheduling mode parameters. The historical optimal solution is used as the initial solution for the dynamic weight model; where, This is the optimal power solution for the energy storage system, including the optimal charging power solution and the optimal discharging power solution; The optimal solution for grid interaction power includes: D1. Read the historical optimal solution from the backup strategy library. ; D2. Verify whether the historical optimal solution satisfies the power balance constraint condition. ; D3. If the verification passes, the historical best solution will be used as the initial solution; otherwise, the default initial solution will be used.

6. The optimization method of the energy storage charging and discharging scheduling algorithm according to claim 4, characterized in that: The dynamic adjustment of the relaxation variable based on the power grid frequency deviation includes: E1. Acquire the power grid frequency and calculate the power grid frequency deviation based on the power grid frequency. ,in, The power grid frequency, The rated frequency of the power grid; E2. If the power grid frequency deviation is greater than or equal to the frequency deviation threshold and continues for a preset duration, the relaxation variable is adjusted according to the following formula: in, For the adjusted slack variables, The slack variables before adjustment , representing the cumulative amount of frequency deviation exceeding the threshold; The frequency deviation threshold is... For the preset duration, This represents the cumulative number of adjustments made to the slack variable during the current event, where the event is either in the supercritical state or in the normal state. γ is the preset maximum number of adjustments allowed, λ is the frequency deviation sensitivity coefficient, and λ is the adjustment attenuation coefficient.

7. The optimization method of the energy storage charging and discharging scheduling algorithm according to claim 6, characterized in that: The supercritical judgment conditions are optimized and the backup strategy library is updated based on the records of power grid frequency deviation and supercritical events. The backup strategy library stores the scheduling mode parameters, the slack variables, the supercritical judgment conditions, and the historical optimal solutions, including: F1. In events categorized as being in a supercritical state, record the number of times the grid frequency deviation is greater than or equal to the frequency deviation threshold, and record these as supercritical events. F2. If the number of times the same event is triggered reaches the preset number of triggers, the current supercritical judgment condition is determined to be inaccurate. F3. When the supercriticality criterion is inaccurate, update the power descent threshold in the supercriticality criterion using the gradient descent method; the update formula for the power descent threshold is: in, The updated power descent threshold. The previous power sag threshold was used. To preset the learning rate, This represents the actual frequency deviation. This is due to prediction bias based on historical data; F4. Write the updated power descent threshold to the backup policy library.

8. An optimization device for an energy storage charging and discharging scheduling algorithm, characterized in that: include: The light-temperature coupling module is used to generate a light-temperature coupling feature vector based on the collected photovoltaic power data and temperature control load data; The mode management module is used to determine whether a supercritical state is triggered based on the light-temperature coupling feature vector and predefined supercritical judgment conditions, and to load the corresponding scheduling mode parameters and relaxation variables; wherein, the scheduling mode parameters are used to adjust the economic priority and power supply guarantee priority of energy storage charging and discharging, and the relaxation variables are used to adjust the power balance constraints of the power grid. The dynamic optimization module is used to construct a dynamic weight model based on the scheduling mode parameters and the power balance constraints; to preheat the initial solution of the dynamic weight model using the historical best solution that matches the current scheduling mode parameters; and to obtain the charging and discharging power command by solving the dynamic weight model; the scheduling mode parameters include a supply guarantee mode and an economic mode. The power regulation module is used to execute the charging and discharging power command, dynamically adjust the relaxation variable according to the grid frequency deviation, and re-execute the operations of the optical-temperature coupling module, the mode management module, and the dynamic optimization module. The strategy optimization module is used to optimize the supercritical judgment conditions and update the backup strategy library based on the records of power grid frequency deviation and supercritical events; the backup strategy library stores the scheduling mode parameters, the slack variables, the supercritical judgment conditions, and the historical optimal solutions.

Images