An intelligent prediction-based energy storage system charging and discharging optimization control method

Abstract

The application discloses a kind of based on intelligent prediction's energy storage system charge-discharge optimization control method, it is related to energy storage control technical field, control steps are:S1, obtain energy storage system charge-discharge historical data, after pre-processing, historical data is uploaded storage;S2, construct state space model and optimization objective function, obtain historical data and real-time monitoring data, disturbance variable of future at least 2 control periods is predicted using intelligent algorithm, with multi-period optimization objective function as criterion, the optimal control sequence of future N periods is solved, based on actual system state update model, monitoring data is input into model and output.This application breaks through the limitation of traditional single-period optimization, solves the optimal control sequence of future N periods through multi-period optimization objective function, plans ahead the disturbance and constraint within multi-period, avoids the local solution trap caused by only paying attention to single-period optimization, can adjust control rhythm in advance through multi-period prediction, effectively suppresses system oscillation, prevents divergence.

Description

Technical Field

[0001] This invention relates to the field of energy storage control technology, specifically to a method for optimizing the charging and discharging control of an energy storage system based on intelligent prediction. Background Technology

[0002] As the proportion of renewable energy in the power system continues to increase, the optimization of charging and discharging control strategies for energy storage systems, as key devices for mitigating new energy fluctuations and improving grid stability, is becoming increasingly important. Model predictive control (MPC), as an advanced control method, has been widely applied to the charging and discharging optimization of energy storage systems. However, existing model predictive control (MMDC)-based optimization methods for energy storage system charging and discharging have significant drawbacks: traditional MMDC only considers the optimal output state within a single control cycle, neglecting the optimal solution of the predictive model over multiple control cycles. This makes it prone to getting trapped in local optima, which can lead to increased oscillations or even system divergence under conditions of significant disturbances (such as sudden changes in renewable energy output or drastic load fluctuations) or modeling errors, severely impacting the stability and economics of the energy storage system. Furthermore, balancing real-time performance and optimization effectiveness is often difficult when dealing with long-term optimization problems, resulting in a significant gap between the actual performance of the control strategy and theoretical expectations. To address these issues, we propose an intelligent predictive-based optimization control method for energy storage system charging and discharging. Summary of the Invention

[0003] To address the aforementioned technical problems, a method for optimizing the charging and discharging of an energy storage system based on intelligent prediction is provided. This technical solution solves the problems mentioned above.

[0004] To achieve the above objectives, the technical solution adopted by this invention is: a charging and discharging optimization control method for an energy storage system based on intelligent prediction, wherein the control steps are as follows: S1. Obtain historical charging and discharging data of the energy storage system, preprocess the historical data, and then upload and store it. S2. Construct a state-space model and optimize the objective function, obtain historical data and real-time monitoring data, use intelligent algorithms to predict the disturbance variables for at least two control cycles in the future, solve the optimal control sequence for the next N cycles based on the multi-cycle optimization objective function, and input the monitoring data into the model output based on the actual system state update model to obtain the energy storage system charging and discharging optimization control strategy. S3. Convert the output control strategy into a device-compatible physical control signal to adjust the charging and discharging process; S4. Continuously monitor the charging and discharging status, provide timely feedback on any abnormalities, write the control process into the log, and perform subsequent access monitoring.

[0005] Preferably, the specific operation steps in step S2 are as follows: Establish a state-space model of the energy storage system, which includes the state variables, control variables, and disturbance variables of the energy storage system; A multi-cycle optimization objective function is constructed, which comprehensively considers the system operating cost, energy storage lifetime loss, and control stability within multiple control cycles. Based on historical data and real-time monitoring data, an intelligent prediction algorithm is used to predict the changes of disturbance variables in the next N control cycles, where N≥2; At each control moment, the optimal control sequence for the next N control cycles is solved using the multi-cycle optimization objective function as the optimization criterion. Only the first control action in the optimal control sequence is executed, and then the process proceeds to the next control moment; At the next control moment, the state space model is updated based on the actual system state. The monitoring data is input into the model to output the control strategy.

[0006] Preferably, the state-space model is established by clarifying the physical mechanism boundaries and modeling objectives, defining the dynamic processes and application scenarios to be considered, selecting variables, including state variables which are the smallest set describing the core dynamics, control variables which are actively regulated inputs, and disturbance variables which are uncontrollable disturbances; establishing nonlinear state equations based on the law of conservation of charge / energy to describe the rate of change of state variables, defining output equations, associating measurable outputs with variables, identifying parameters, verifying dynamics and robustness through experiments, and correcting deviations.

[0007] Preferably, the state equation is in the form of:

[0008] in The time derivative of the state variable; It is a state vector; For control vectors; The perturbation vector; It is a nonlinear function; The output equation is defined in the following form:

[0009] in The output vector includes battery terminal voltage and output power.

[0010] Preferably, the objective function construction steps are as follows: Determine the total optimization duration, divide the control cycle into N discrete control cycles according to the control frequency, clarify the operating constraints of the energy storage system in each cycle, and unify the correspondence between the time scale and the time domain of the variables. Define the mathematical expressions for the three sub-objectives, where the system operating cost expression is: Operating cost = Electricity price × Electricity purchased + Maintenance cost The expression for energy storage lifetime loss is: Lifetime loss = Number of cycles / SOC decay model By using the depth of charge and discharge and the current-related lifetime loss coefficient, it can be converted into equivalent economic cost or lifetime decay. The control stability expression is: Control fluctuation = Energy storage power in adjacent cycles / SOC fluctuation The square of the fluctuation value is used to characterize the control fluctuation; Normalization was used to eliminate the influence of the dimensions of the results; the weights of each sub-objective were determined by the analytic hierarchy process in combination with engineering requirements. Taking the total cumulative sum of the weighted sub-objectives of each period as the optimization objective, the objective function is obtained, and its expression is: Overall goal = Σ (cycle k operating cost × weight 1 + cycle k lifespan loss × weight 2 + cycle k control fluctuation × weight 3).

[0011] Preferably, the intelligent prediction algorithm predicts by acquiring historical data, constructing feature engineering based on the historical data, using the ARIMA model for prediction, training and optimizing the ARIMA model, fusing the prediction results of a single model, inputting real-time data into the optimized model, and outputting the predicted values ​​of the disturbance variables for the next N control cycles; comparing the actual disturbance value with the predicted value in each cycle, calculating the error and feeding it back, and if the error exceeds the threshold, continuously optimizing the model parameters.

[0012] Preferably, at each control moment, the current system state, the predicted disturbances for the next N periods, and the constraints are integrated. The multi-period optimization objective function is transformed into a mathematical programming problem. The optimal control sequence is solved based on the LP algorithm. The first control action is executed and the command is issued. During execution, anomalies are monitored and temporary compensation is provided. At the next control moment, the actual operating data of the previous period is collected, the state variables are updated after verification, the error of the state transition equation is corrected using Kalman filtering, the time-varying parameters of internal resistance are re-identified using the least squares method, and the model structure is adjusted to match the system changes. The new monitoring data and the re-predicted disturbance values ​​are input, and the optimization process is repeated to solve for the new optimal sequence in combination with the updated state-space model.

[0013] Preferably, the historical data in step S1 includes charging and discharging power, charging and discharging capacity, charging and discharging duration, state of charge, environmental data, and time log data; data preprocessing includes data cleaning, data format standardization, and normalization.

[0014] Preferably, step S3 includes analyzing the control strategy, extracting the charging and discharging mode and core power parameters, verifying compliance against the device's rated values, and automatically reducing the rating if the limit is exceeded; matching the device's communication protocol to encapsulate the instructions, converting the digital instructions into physical signals of the appropriate level through D / A conversion, and then outputting them for execution.

[0015] Preferably, in step S4, the abnormal feedback is achieved by establishing quantitative abnormal judgment rules to avoid misjudgment or omission, and feedback is provided through pop-up window feedback, warning sound feedback, and alarm sound feedback.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention predicts disturbance variables for at least two control cycles in advance, avoiding control lag caused by sudden disturbances in traditional passive response modes. The multi-cycle optimization objective function can plan the control sequence for N cycles in advance based on the prediction, allowing for more charging during low-price periods, reducing electricity purchase costs and avoiding hasty adjustments when disturbances occur, thus improving the system's ability to cope with uncertainty. It breaks through the limitations of traditional single-cycle optimization by solving the optimal control sequence for the next N cycles through the multi-cycle optimization objective function, coordinating disturbances and constraints in advance across multiple cycles, avoiding local solution traps caused by focusing only on single-cycle optima. When faced with sudden changes in renewable energy output, large disturbances such as drastic load fluctuations, or modeling errors, it can adjust the control rhythm in advance through multi-cycle prediction, effectively suppressing system oscillations, preventing divergence, and significantly improving the stability of energy storage operation. Attached Figure Description

[0017] Figure 1 This is a flowchart of the control steps of the present invention. Detailed Implementation

[0018] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0019] Reference Figure 1 As shown, a charging and discharging optimization control method for an energy storage system based on intelligent prediction is described, and the control steps are as follows: S1. Obtain historical charging and discharging data of the energy storage system, preprocess the historical data, and then upload and store it. S2. Construct a state-space model and optimize the objective function, obtain historical data and real-time monitoring data, use intelligent algorithms to predict the disturbance variables for at least two control cycles in the future, solve the optimal control sequence for the next N cycles based on the multi-cycle optimization objective function, and input the monitoring data into the model output based on the actual system state update model to obtain the energy storage system charging and discharging optimization control strategy. S3. Convert the output control strategy into a device-compatible physical control signal to adjust the charging and discharging process; S4. Continuously monitor the charging and discharging status, provide timely feedback on any abnormalities, write the control process into the log, and perform subsequent access monitoring.

[0020] This application uses S1 to preprocess and store historical data, combined with S2 real-time data, to provide high-quality support for decision-making and reduce blind spots. In terms of strategy optimization, intelligent algorithms predict disturbance variables in advance, use multi-period objective functions to find the optimal sequence, and can also dynamically update the model to adapt to system changes, achieve global optimization, and improve anti-interference capabilities. During execution, S3 converts the strategy into device-compatible signals, avoiding manual conversion errors and ensuring accurate and timely charging and discharging adjustments. In operation and maintenance management, S4 continuously monitors and provides real-time feedback on anomalies, reducing failure losses. Full-process log recording can also help trace problems and optimize operation and maintenance, reducing costs. Overall, it can improve energy storage efficiency and economic benefits, ensure system reliability and maintainability, and is suitable for various energy storage scenarios.

[0021] The specific steps in step S2 are as follows: Establish a state-space model of the energy storage system, which includes the state variables, control variables, and disturbance variables of the energy storage system; A multi-cycle optimization objective function is constructed, which comprehensively considers the system operating cost, energy storage lifetime loss, and control stability within multiple control cycles. Based on historical data and real-time monitoring data, an intelligent prediction algorithm is used to predict the changes of disturbance variables in the next N control cycles, where N≥2; At each control moment, the optimal control sequence for the next N control cycles is solved using the multi-cycle optimization objective function as the optimization criterion. Only the first control action in the optimal control sequence is executed, and then the process proceeds to the next control moment; At the next control moment, the state space model is updated based on the actual system state. The monitoring data is input into the model to output the control strategy.

[0022] Traditional methods, due to their single-cycle perspective, struggle to balance long-term costs and benefits. In contrast, multi-cycle optimization incorporates long-term operating costs and benefits into a unified framework, planning the charging and discharging rhythm across multiple cycles. This allows for greater energy storage during low-price periods and greater discharging during high-yield periods, achieving full-cycle economic optimization, avoiding the damage of short-term decisions to long-term benefits, and narrowing the gap between actual and theoretically expected benefits. Traditional methods for handling long-term optimization are prone to computational lag due to the increase in optimization dimensions, or sacrifice optimization performance in pursuit of real-time simplification of the model. By rationally dividing the control cycle and optimizing the intelligent prediction and solution algorithms, accurate prediction and fast solution can be achieved in multi-cycle optimization. This ensures the real-time updating of the control strategy without reducing the accuracy of multi-cycle optimization, perfectly balancing the two and enabling the control strategy to still approach the theoretical optimal performance in complex scenarios.

[0023] The state-space model is established by clarifying the physical mechanism boundaries and modeling objectives, defining the dynamic processes and application scenarios under consideration, and selecting variables, including state variables which are selected as the minimum set describing the core dynamics, control variables which are actively regulated inputs, and disturbance variables which are uncontrollable disturbances. Nonlinear state equations are established based on the law of conservation of charge / energy to describe the rate of change of state variables, output equations are defined, and measurable outputs are correlated with variables. Parameters are identified, dynamics and robustness are verified through experiments, and deviations are corrected.

[0024] This application ensures that the model closely matches the real dynamic characteristics of the energy storage system by clearly defining the physical mechanism boundaries and modeling objectives, avoiding the black box bias that may occur in purely data-driven models. The state equations constructed based on conservation laws can accurately reflect the changing laws of state variables, making the model output highly consistent with the dynamics of the actual system, and providing a reliable physical basis for subsequent optimization control.

[0025] The state equation is in the form of:

[0026] in The time derivative of the state variable; It is a state vector; For control vectors; The perturbation vector; It is a nonlinear function; The output equation is defined in the following form:

[0027] in The output vector includes battery terminal voltage and output power.

[0028] The steps for constructing the objective function are as follows: Determine the total optimization duration, divide the control cycle into N discrete control cycles according to the control frequency, clarify the operating constraints of the energy storage system in each cycle, and unify the correspondence between the time scale and the time domain of the variables. Define the mathematical expressions for the three sub-objectives, where the system operating cost expression is: Operating cost = Electricity price × Electricity purchased + Maintenance cost The expression for energy storage lifetime loss is: Lifetime loss = Number of cycles / SOC decay model By using the depth of charge and discharge and the current-related lifetime loss coefficient, it can be converted into equivalent economic cost or lifetime decay. The control stability expression is: Control fluctuation = Energy storage power in adjacent cycles / SOC fluctuation The square of the fluctuation value is used to characterize the control fluctuation; Normalization was used to eliminate the influence of the dimensions of the results; the weights of each sub-objective were determined by the analytic hierarchy process in combination with engineering requirements. Taking the total cumulative sum of the weighted sub-objectives of each period as the optimization objective, the objective function is obtained, and its expression is: Overall goal = Σ (cycle k operating cost × weight 1 + cycle k lifespan loss × weight 2 + cycle k control fluctuation × weight 3).

[0029] The intelligent prediction algorithm acquires historical data and constructs feature engineering based on the historical data. It uses the ARIMA model for prediction, trains and optimizes the ARIMA model, fuses the prediction results of individual models, inputs real-time data into the optimized model, and outputs the predicted values ​​of disturbance variables for the next N control cycles. It compares the actual disturbance value with the predicted value in each cycle, calculates the error and feeds it back. If the error exceeds the threshold, it continuously optimizes the model parameters.

[0030] This application constructs feature engineering (such as time period and lag term features) based on historical data, which can fully explore the changing patterns of disturbance variables (such as load and temperature) and provide high-quality input for the ARIMA model. Through model training optimization and multi-result fusion, it can make up for the limitations of a single ARIMA model in nonlinear scenarios, reduce prediction bias, and accurately predict disturbances for the next N periods. This can help energy storage systems plan charging and discharging strategies in advance, avoid passive control due to sudden disturbances, and improve the system's ability to cope with uncertainty.

[0031] At each control moment, the system's current state, future N-period disturbance predictions, and constraints are integrated. The multi-period optimization objective function is transformed into a mathematical programming problem. The optimal control sequence is solved based on the LP algorithm. The first control action is executed and the command is issued. During execution, anomalies are monitored and temporary compensations are provided. At the next control moment, the actual operating data of the previous period is collected, verified, and the state variables are updated. Kalman filtering is used to correct errors in the state transition equation. The time-varying parameters of internal resistance are re-identified using the least squares method, and the model structure is adjusted to match system changes. The new monitoring data and the re-predicted disturbance values ​​are input, and the optimization process is repeated to solve for a new optimal sequence, combined with the updated state-space model.

[0032] The historical data in step S1 includes charging and discharging power, charging and discharging capacity, charging and discharging duration, state of charge, environmental data, and time log data; data preprocessing includes data cleaning, data format standardization, and normalization.

[0033] The core parameters selected in this application—charge and discharge power, energy capacity, and duration—directly reflect the energy interaction process of energy storage; the state of charge reflects the state of the energy storage device; environmental data is correlated with external interference; and time logs record the temporal patterns.

[0034] Step S3 includes analyzing the control strategy, extracting the charging and discharging mode and core power parameters, verifying compliance against the equipment's rated values, and automatically reducing the rating if the limit is exceeded; matching the equipment's communication protocol to encapsulate the instructions, converting the digital instructions into physical signals with appropriate levels through D / A conversion, and then outputting them for execution; in step S4, abnormal feedback is achieved by establishing quantified abnormal judgment rules to avoid misjudgment or omission, and feedback is provided through pop-up windows, warning sounds, and alarm sounds.

[0035] This application analyzes the strategy to extract core parameters and verify compliance, which can avoid issuing instructions that exceed the equipment's rated value, prevent equipment overload damage, and ensure control continuity when the strategy exceeds the limit. Matching protocol encapsulation instructions and D / A conversion to adapt levels can solve the problems of communication format differences and signal incompatibility among multiple devices, ensuring that digital instructions are converted into physical signals without distortion, avoiding execution errors caused by signal deviation, and ensuring that charging and discharging adjustments strictly conform to the optimization strategy objectives.

[0036] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for optimizing charge and discharge control of an energy storage system based on intelligent prediction, characterized in that, The control steps are as follows: S1. Obtain historical charging and discharging data of the energy storage system, preprocess the historical data, and then upload and store it. S2. Construct a state-space model and optimize the objective function, obtain historical data and real-time monitoring data, use intelligent algorithms to predict the disturbance variables for at least two control cycles in the future, solve the optimal control sequence for the next N cycles based on the multi-cycle optimization objective function, and input the monitoring data into the model output based on the actual system state update model to obtain the energy storage system charging and discharging optimization control strategy. S3. Convert the output control strategy into a device-compatible physical control signal to adjust the charging and discharging process; S4. Continuously monitor the charging and discharging status, provide timely feedback on any abnormalities, write the control process into the log, and perform subsequent access monitoring.

2. The method for optimizing charge and discharge control of an energy storage system based on intelligent prediction according to claim 1, characterized in that, The specific steps in step S2 are as follows: Establish a state-space model of the energy storage system, which includes the state variables, control variables, and disturbance variables of the energy storage system; A multi-cycle optimization objective function is constructed, which comprehensively considers the system operating cost, energy storage lifetime loss, and control stability within multiple control cycles. Based on historical data and real-time monitoring data, an intelligent prediction algorithm is used to predict the changes of disturbance variables in the next N control cycles, where N≥2; At each control moment, the optimal control sequence for the next N control cycles is solved using the multi-cycle optimization objective function as the optimization criterion. Only the first control action in the optimal control sequence is executed, and then the process proceeds to the next control moment; At the next control moment, the state space model is updated based on the actual system state. The monitoring data is input into the model to output the control strategy.

3. The method for optimizing charge and discharge control of an energy storage system based on intelligent prediction according to claim 2, characterized in that, The state-space model is established by clarifying the physical mechanism boundaries and modeling objectives, defining the dynamic processes and application scenarios under consideration, and selecting variables, including state variables which are selected as the minimum set describing the core dynamics, control variables which are actively regulated inputs, and disturbance variables which are uncontrollable disturbances. Nonlinear state equations are established based on the law of conservation of charge / energy to describe the rate of change of state variables, output equations are defined, and measurable outputs are correlated with variables. Parameters are identified, dynamics and robustness are verified through experiments, and deviations are corrected.

4. The method for optimizing charge and discharge control of an energy storage system based on intelligent prediction according to claim 3, characterized in that, The state equation is in the form of: in The time derivative of the state variable; It is a state vector; For control vectors; The perturbation vector; It is a nonlinear function; The output equation is defined in the following form: in The output vector includes battery terminal voltage and output power.

5. The method for optimizing charge and discharge control of an energy storage system based on intelligent prediction according to claim 2, characterized in that, The steps for constructing the objective function are as follows: Determine the total optimization duration, divide the control cycle into N discrete control cycles according to the control frequency, clarify the operating constraints of the energy storage system in each cycle, and unify the correspondence between the time scale and the time domain of the variables. Define the mathematical expressions for the three sub-objectives, where the system operating cost expression is: Operating cost = Electricity price × Electricity purchased + Maintenance cost The expression for energy storage lifetime loss is: Lifetime loss = Number of cycles / SOC decay model By using the depth of charge and discharge and the current-related lifetime loss coefficient, it can be converted into equivalent economic cost or lifetime decay. The control stability expression is: Control fluctuation = Energy storage power in adjacent cycles / SOC fluctuation The square of the fluctuation value is used to characterize the control fluctuation; Normalization was used to eliminate the influence of the dimensions of the results; the weights of each sub-objective were determined by the analytic hierarchy process in combination with engineering requirements. Taking the total cumulative sum of the weighted sub-objectives of each period as the optimization objective, the objective function is obtained, and its expression is: Overall goal = Σ (cycle k operating cost × weight 1 + cycle k life loss × weight 2 + cycle k control fluctuation × weight 3).

6. The method for optimizing charge and discharge control of an energy storage system based on intelligent prediction according to claim 2, characterized in that: The intelligent prediction algorithm acquires historical data and constructs feature engineering based on the historical data. It uses the ARIMA model for prediction, trains and optimizes the ARIMA model, fuses the prediction results of individual models, inputs real-time data into the optimized model, and outputs the predicted values ​​of disturbance variables for the next N control cycles. It compares the actual disturbance value with the predicted value in each cycle, calculates the error and feeds it back. If the error exceeds the threshold, it continuously optimizes the model parameters.

7. The method for optimizing charge and discharge control of an energy storage system based on intelligent prediction according to claim 2, characterized in that: At each control moment, the current state of the system, the prediction of disturbances in the next N periods, and the constraints are integrated. The multi-period optimization objective function is transformed into a mathematical programming problem. The optimal control sequence is solved based on the LP algorithm. The first control action is executed and the instruction is issued. During the execution, anomalies are monitored and temporary compensation is provided. Entering the next control moment, the actual operating data of the previous cycle is collected, the state variables are updated after verification, the error of the state transition equation is corrected by Kalman filtering, the time-varying parameters of internal resistance are re-identified by the least squares method, and the model structure is adjusted to match the system changes; the monitoring data of the new moment and the re-predicted disturbance value are input, and the optimization process is repeated to solve for the new optimal sequence in combination with the updated state space model.

8. The method for optimizing charge and discharge control of an energy storage system based on intelligent prediction according to claim 1, characterized in that: The historical data in step S1 includes charging and discharging power, charging and discharging capacity, charging and discharging duration, state of charge, environmental data, and time log data; data preprocessing includes data cleaning, data format standardization, and normalization.

9. The method for optimizing charge and discharge control of an energy storage system based on intelligent prediction according to claim 1, characterized in that: Step S3 includes analyzing the control strategy, extracting the charging and discharging modes and core power parameters, verifying compliance against the equipment's rated values, and automatically reducing the rating if the limit is exceeded; matching the equipment's communication protocol to encapsulate the instructions, converting the digital instructions into physical signals with compatible levels through D / A conversion, and then outputting them for execution.

10. The method for optimizing charge and discharge control of an energy storage system based on intelligent prediction according to claim 1, characterized in that: In step S4, anomaly feedback is achieved by establishing quantified anomaly judgment rules to avoid misjudgment or omission, and feedback is provided through pop-up windows, warning sounds, and alarm sounds.

Images