Energy storage system capacity configuration optimization method based on battery capacity attenuation trajectory prediction

Abstract

The invention relates to the technical field of energy storage systems, and provides an energy storage system capacity configuration optimization method based on battery capacity attenuation trajectory prediction. The method comprises the following steps: acquiring operation state data of a battery module in the energy storage system, establishing a historical operation database and a battery capacity attenuation trajectory prediction model, calculating an attenuation trajectory of battery capacity along with time through a temperature accelerated aging factor and a cyclic aging factor, and obtaining a capacity attenuation prediction curve in a future time period; and establishing a capacity configuration optimization function taking net present value maximization as a target by combining a load demand curve and an economic index of the energy storage system, dynamically adjusting charge and discharge depth limitation and a power distribution proportion, dynamically adjusting charge and discharge power of each battery module according to a real-time capacity state, and realizing optimized operation of the energy storage system. According to the invention, accurate prediction of battery capacity attenuation and optimization of full life cycle capacity configuration are realized, and the economic benefit and operation reliability of the energy storage system are improved.

Description

Technical Field

[0001] This invention relates to the field of energy storage system technology, and provides a method for optimizing the capacity configuration of energy storage systems based on battery capacity degradation trajectory prediction. Background Technology

[0002] With the global energy structure transformation and the increasing proportion of renewable energy generation, energy storage systems are playing an increasingly prominent role in the power system. Energy storage systems can smooth the fluctuations in renewable energy generation, achieve peak shaving and valley filling, and improve the stability and economy of grid operation. Lithium-ion batteries, due to their high energy density, long cycle life, and high charge / discharge efficiency, have become the mainstream technology for large-scale energy storage systems. However, lithium-ion batteries inevitably experience capacity decay during actual operation, and this decay characteristic has a significant impact on the long-term operating performance and economic benefits of energy storage systems.

[0003] Current capacity configuration methods for energy storage systems suffer from the following technical problems. First, traditional capacity configuration methods are typically based on initial load demand and empirical coefficients for static configuration, failing to fully consider the capacity degradation characteristics of batteries throughout their entire lifespan. This static configuration method often leads to excessive initial investment and wasted funds, or insufficient capacity later on, affecting system performance. Second, existing technologies do not comprehensively monitor battery operating status, only collecting basic parameters such as voltage and current, lacking systematic collection and analysis of key aging characteristic parameters such as temperature, internal resistance, and depth of charge / discharge. This incomplete data collection results in a lack of reliable data foundation for subsequent capacity degradation analysis. Regarding capacity degradation prediction, existing technologies mainly use simplified empirical formulas or linear degradation models, assuming that battery capacity decays linearly at a fixed rate. This simplified model ignores the influence of various factors such as temperature, charge / discharge rate, and cycle depth on the battery aging rate, and cannot accurately reflect the capacity degradation law of batteries under actual complex operating conditions. Although some studies have attempted to use purely data-driven machine learning methods for prediction, the lack of consideration for the physical mechanisms of battery aging limits the model's generalization ability and long-term prediction accuracy. Insufficient prediction accuracy directly affects the rationality and economy of capacity configuration schemes. In terms of capacity configuration optimization, existing technologies lack a systematic consideration of the economics of the entire lifecycle of energy storage systems. Traditional methods typically focus only on minimizing initial investment costs, neglecting the impact of factors such as operating revenue, maintenance costs, and the timing of capacity expansion on the overall economic benefits of the project. This one-sided optimization objective makes it difficult for configuration schemes to achieve optimal economic benefits in long-term operation. Furthermore, most existing optimization methods employ static planning, pre-determining fixed capacity expansion schedules, lacking the flexibility to dynamically adjust based on actual operating conditions. Regarding system operation control, existing technologies generally adopt a uniform power allocation strategy, distributing equal charging and discharging power to each battery module in the energy storage system. This simplistic control strategy ignores the differences in the health status of different battery modules, causing modules with poor health to bear excessive loads and accelerate degradation, while modules with better health fail to fully utilize their performance, ultimately resulting in capacity imbalance among modules, affecting the overall lifespan and reliability of the system. In addition, existing technologies lack evaluation and dynamic adjustment mechanisms for capacity degradation compensation effects, failing to optimize control strategies in a timely manner based on actual operating conditions. More critically, existing technologies generally lack closed-loop optimization mechanisms. After an energy storage system is put into operation, actual operating data often deviates from initial predictions. However, existing systems cannot use this feedback data to continuously optimize and correct prediction models and configuration schemes. This open-loop operation mode makes it impossible for the system to adaptively cope with changes in battery aging characteristics and fluctuations in load demand, making it difficult to maintain optimal operating conditions throughout its entire life cycle.

[0004] In view of the problems and shortcomings of the existing technologies, there is an urgent need to develop a capacity configuration method for energy storage systems that can accurately predict the battery capacity decay trajectory, optimize the capacity configuration throughout the entire life cycle, and realize dynamic compensation control and closed-loop adaptive optimization, so as to improve the economic benefits and operational reliability of energy storage systems and promote the engineering application and industrial development of energy storage technology. Summary of the Invention

[0005] To address the problems existing in the background technology, this invention provides a method for optimizing the capacity configuration of an energy storage system based on battery capacity degradation trajectory prediction, comprising the following steps: S1: Collect operating status data of battery modules in the energy storage system, including real-time voltage, real-time current, ambient temperature, number of charge-discharge cycles, depth of charge-discharge and internal resistance parameters, and establish a historical operating database; S2: Based on the collected operating status data, establish a battery capacity decay trajectory prediction model, calculate the battery capacity decay trajectory over time through temperature-accelerated aging factor and cycle aging factor, and obtain the capacity decay prediction curve for future time periods. S3: Based on the battery capacity decay prediction curve, combined with the load demand curve and economic indicators of the energy storage system, establish a capacity configuration optimization objective function, and solve the initial capacity configuration scheme and the phased capacity expansion scheme through dynamic programming algorithm. S4: Based on the capacity configuration scheme, calculate the capacity decay compensation strategy for each battery module, and adjust the charge / discharge depth limit and power allocation ratio to ensure that the system maintains a stable output capability throughout its entire life cycle. S5: Outputs the capacity configuration scheme and attenuation compensation strategy to the energy storage system controller. The controller dynamically adjusts the charging and discharging power of each battery module according to the real-time capacity status, and triggers a capacity expansion command when the capacity attenuation reaches a preset threshold, thereby achieving optimized operation of the energy storage system.

[0006] Furthermore, step S1 includes the following specific steps: S11: Install voltage sensors at the positive and negative terminals of each battery module in the energy storage system, install Hall current sensors in series on the main current channel of the charging and discharging circuit, attach temperature sensors to the center of the battery module housing, and connect the signal output terminals of each sensor to the analog input port of the data acquisition unit through shielded cables. S12: Start the data acquisition unit, set the voltage sampling frequency to 5Hz, the current sampling frequency to 10Hz, and the temperature sampling frequency to 1Hz. The data acquisition unit synchronously samples the analog signals output by each sensor, converts the analog signals into digital signals through a 16-bit analog-to-digital converter, and obtains the voltage values ​​at each moment. Current value and temperature value ; S13: Set a cycle counting program in the battery module's charge / discharge controller. When the battery module completes a full cycle from charging to discharging and back to charging, the cycle count is recorded. Automatically increment by 1, and simultaneously record the initial state of charge for this cycle. and end of state of charge Calculate the depth of charge and discharge for this cycle. The formula is: ;in Depth of charge / discharge; It is the decimal form of the initial charge state of the cycle, with a value range of 0-1; The decimal form of the charge state at the end of the cycle, with a value range of 0-1; S14: Every 24 hours, the internal resistance of the battery module is tested using an electrochemical impedance spectroscopy (EIS) testing device. During the test, the normal charging and discharging operation of the battery module is paused. An AC excitation current with a frequency of 1kHz and an amplitude of 0.1C of the battery's rated capacity is applied to the battery module through a signal generator. The voltage response waveform and the input current waveform across the battery terminals are recorded simultaneously using a dual-channel oscilloscope. The phase difference and amplitude ratio of the voltage and current signals are analyzed using Fast Fourier Transform (FFT) to calculate the AC internal resistance. After the test is completed, the battery module will be restored to normal operation. S15: Establish the historical operation database table structure. The data table includes the following fields: timestamp, moduleid, voltage. Current ,temperature Internal resistance Number of loops Depth of charge and discharge and cumulative throughput The cumulative throughput is calculated by integration, using the following formula: ;in For cumulative throughput; This represents the number of loop iterations. For the first Average current of each cycle; For the first The duration of each cycle; the collected and calculated data are written into the database one by one in chronological order, and an index is created with the timestamp and battery module number as a combined primary key.

[0007] Furthermore, step S2 includes the following specific steps: S21: Extract temperature data of each battery module from the historical operation database. Arranged according to the time series, for each temperature data point, substitute it into the Arrhenius equation to calculate the corresponding temperature acceleration factor. The formula is: ;in It is a temperature acceleration factor; It is a natural exponential function; The activation energy for the battery aging reaction is 25000 J / mol for lithium iron phosphate batteries and 30000 J / mol for ternary lithium batteries. This is the universal gas constant, with a value of 8.314 J / (mol·K); For reference temperature, a value of 298.15K corresponds to 25℃; The actual operating temperature is obtained by adding 273.15 to the Celsius temperature; the calculated temperature... The value is integrated over time to obtain the cumulative temperature-accelerated aging amount. S22: Extract the cumulative throughput of each battery module from the historical operation database. ,Will Substitute into the cyclic aging model to calculate the capacity decay during cyclic aging. The formula is: ;in This refers to the capacity decay caused by cyclic aging. To determine the cycle aging factor, a test battery with a rated capacity of 100Ah was subjected to a 1C rate charge-discharge cycle experiment at 25℃. The capacity was measured every 100 cycles, and a double logarithmic curve of capacity decay versus cumulative throughput was plotted. The curve was then fitted using the least squares method to obtain the aging factor. Value, typically takes the value of ; For cumulative throughput; The cyclic aging index is obtained synchronously through the above fitting process, with a typical value of 0.6. S23: Establish a time-varying capacity decay model for the battery, superimposing the effects of temperature aging and cycle aging, and calculate the... Day's remaining battery capacity The formula is: ;in For the first The remaining capacity for the day; This refers to the initial rated capacity of the battery. Indicates the integration operator; For integration variables; The calendar aging rate coefficient is obtained by placing the battery in an open-circuit environment at 25°C, measuring the capacity every 30 days, and recording the capacity decay rate over 180 days. A typical value is 0.005 Ah / day. For the first The temperature acceleration factor of the sky; As of the date The daily cyclic aging capacity decay; the model is backtested using historical data to ensure that the error between the model's predicted value and the measured value is less than 3%; S24: Construct a Long Short-Term Memory (LSTM) neural network for capacity decay trajectory prediction. The network structure includes an input layer, three LSTM hidden layers, and an output layer. The input layer receives historical running data, including temperature sequences, current sequences, loop counts, and cumulative throughput. After normalization, the data is input into the first LSTM layer. The first LSTM layer contains 64 neurons, each containing an input gate, a forget gate, and an output gate. The activation function processes the cell state; the output of the first layer is passed to the second LSTM layer, which contains 32 neurons; the output of the second layer is passed to the third LSTM layer, which contains 16 neurons; the output of the third layer is connected to the fully connected output layer, and the number of neurons in the output layer is equal to the number of predicted time points. A linear activation function is used to output the predicted capacity values ​​at each future time point. S25: Train the LSTM network using historical running data. Divide the historical data into training samples according to time windows. Each sample contains 90 consecutive days of input data and the subsequent 30 days of capacity decay data as labels. Use the Adam optimizer to minimize the mean squared error loss function MSE, as shown in the formula: ;in Mean square error; To predict the number of samples; For the first Predicted capacity value for each sample; For the first The true capacity of each sample is set; the learning rate is set to 0.001, the batch size is 32, the training is performed for 300 rounds, the model performance is evaluated on the validation set every 50 rounds, and the model parameters with the minimum validation loss are saved. S26: Use the trained LSTM model to predict future capacity decay trajectories. Take the operational data from the last 90 days as model input, and the model outputs capacity predictions for the next 1 to 20 years. The prediction time interval is 30 days, resulting in a total of 240 predicted data points. Plot the prediction results as a capacity decay prediction curve, with time on the x-axis and remaining capacity on the y-axis. Simultaneously, calculate the health status of each prediction point. The formula is: ;in For the first Constant health status; For the first Predicted remaining capacity at any given time; The initial rated capacity; the capacity degradation prediction curve and Curve data is stored as a CSV file for subsequent capacity configuration and optimization.

[0008] Furthermore, step S3 includes the following specific steps: S31: Obtain load demand data from the energy storage system and export historical load power curves from the user-side energy management system (EMS). The data spans the past year, with a time granularity of 15 minutes, totaling 35,040 data points; statistical analysis is performed on the load data to calculate the daily maximum load power. Average load power Load peak-valley difference And the load factor, the formula is: Load factor = ;in Average load power; The maximum load power is determined by the time series decomposition method. Based on historical load patterns and future electricity consumption growth forecasts, the load is decomposed into trend components, periodic components, and random components using the time series decomposition method. The load demand curve for the next 10-20 years is then extrapolated. S32: Construct the objective function for capacity configuration optimization, with the objective of maximizing the net present value (NPV) of the energy storage system over its entire lifecycle. The formula is: ;in Net present value; The system design life is set at 15 years. For the first Annual operating revenue; For the first Annual operating costs, including electricity, maintenance, and labor costs; The discount rate is 0.08. Investment costs for initial capacity configuration; For the number of capacity expansions; For the first The investment cost of subsequent capacity expansion; For the first The year in which the capacity expansion occurred; operating revenue Further calculations yield the following formula: ;in This refers to the rated power of the energy storage system. This represents the daily peak operating time, which is set to 4 hours based on local electricity pricing policies. The peak hour electricity price can be found by consulting the local electricity price list. Off-peak electricity pricing; For round-trip efficiency, the value is 0.90; The daily call duration for auxiliary services is set at 1.5 hours per day, based on the power grid dispatching requirements. Compensation for ancillary services; For the first Average annual available capacity; The annual price for capacity leasing; S33: Set constraints for capacity configuration and establish a set of constraint inequalities: First, initial capacity constraints, ;in Initial capacity configuration; The minimum duration of continuous discharge is set to 2 hours. The first constraint is a safety margin factor, set to 1.2; the second constraint is capacity adequacy, which applies to any point during system operation. ,satisfy ;in For the first The available capacity of the system at any given time is equal to the sum of the remaining capacity of all battery modules multiplied by their health status. For the first Load power at any given time; The load response time is set to 1 hour; 1.1 represents a 10% reserve capacity factor; third, battery life constraints apply, when any battery module's... When the efficiency drops to 80%, the module must be replaced or the system capacity must be expanded; fourth, economic constraints, the system's investment payback period should not exceed 8 years; S34: Use dynamic programming algorithm to solve the capacity configuration optimization problem and consider system lifetime. Divided by year Each decision-making stage, each stage The state variable is the current available capacity of the system. The decision variable is the capacity expansion amount at this stage. The value ranges from 0 to the system's maximum expansion capacity; from the last stage Starting with reverse recursion, the value function in the final stage Equal to the net profit of that stage, for stage The value function is calculated based on the Bellman optimality equation, and the formula is as follows: ;in For the first The stage state is The optimal value function at that time; Represents all feasible decision variables Find the maximum value; calculate sequentially. The value function for each stage records the optimal decision at each stage; S35: From the first stage Begin forward backtracking of the optimal decision sequence to determine the initial capacity configuration scheme. And the capacity expansion plan for each stage; for each stage that requires capacity expansion, record the expansion time. and expanded capacity Generate a capacity configuration schedule, with tables containing columns for: year, current capacity, capacity expansion amount, total capacity after expansion, annual investment cost, annual operating income, and annual net cash flow. Use MATLAB to plot a capacity configuration optimization graph, with the horizontal axis representing the year, the left vertical axis representing capacity, and the right vertical axis representing net present value. The graph uses a stepped curve to represent the change in capacity over time, a bar chart to represent the net cash flow for each year, and a cumulative curve to represent the change in net present value. Output the optimization results as an Excel file and a PDF report for decision-makers' reference.

[0009] Furthermore, step S4 includes the following specific steps: S41: Read the capacity degradation prediction curve data generated in step S2 for each battery module in the system. Extract its future time points Predicted remaining capacity Calculate the corresponding health status The formula is: ;in For the first The battery module is in the first Constant health status; For the first The battery module is in the first Predicted remaining capacity at any given time; For the first The initial rated capacity of each battery module; the initial rated capacity of each module. Data is stored in time series, and a system is established. A time matrix, where the row indices are module numbers and the column indices are time points; S42: Establish a dynamic adjustment strategy for charge / discharge depth. For battery modules whose health status is gradually deteriorating, limit their charge / discharge depth to slow down further degradation. The battery module is in the first At any given time, calculate its permissible depth of charge and discharge limit. The formula is: ;in For the first The module in the first Limits on the depth of charge and discharge at any given time; The standard depth of charge / discharge for the system design is 85%. The minimum permissible health state threshold is set to 80%; 0.5 is an adjustment index used to control the rate of change of the depth limit; in the Battery Management System (BMS), the calculated... The value is set as the charge / discharge cutoff condition for this module, when the module's state of charge... Exceeding the permitted range When this happens, the BMS stops the charging and discharging operation of the module; S43: Establish a dynamic optimization strategy for power allocation. When there are differences in the health status of each battery module in the system, the strategy is adjusted according to the power distribution of each module. Proportional allocation of charging and discharging power, for the first Each battery module is used to calculate its power distribution coefficient. The formula is: ;in For the first Power allocation coefficient of each battery module; For the first The current health status of each battery module; This represents the total number of battery modules in the system. This represents the summation operator; For summation index; power allocation coefficient This reflects the proportion of total power that the module should bear. Modules with higher health status bear more power, while modules with lower health status bear less power, thereby achieving balanced use of each module. S44: Based on the power allocation coefficient and the total system power demand, calculate the actual charging and discharging power command for each battery module. Each battery module has a charging and discharging power. The calculation formula is: ;in For the first The charging and discharging power of each battery module; The total power command received by the energy storage system is issued by the superior energy management system; The conversion efficiency of the DC / DC power converter is set to 0.95 during charging and 0.96 during discharging; the calculated power command is then used. Send via CAN communication bus to the first The DC / DC converter of each battery module adjusts the duty cycle of the pulse width modulation (PWM) signal according to the power command to control the actual charging and discharging current. S45: Establish a capacity attenuation compensation effectiveness evaluation mechanism, evaluate the implementation effect of the compensation strategy every 7 days, and calculate the system's capacity balance index (CBD) using the following formula: CBD represents the capacity balance, with a value ranging from 0 to 1. The closer the value is to 1, the more balanced the capacity of each module is. It is a function of standard deviation; It is a function of average value; arrive This indicates the health status of each battery module. When the CBD value is below 0.90, it indicates a significant difference in capacity between the modules, requiring adjustment of the power allocation coefficient to increase the power of modules with lower health status. Reduce its load and increase the load of modules with higher health status. The load is increased, the power command is recalculated and issued to achieve dynamic balance.

[0010] Furthermore, step S5 includes the following specific steps: S51: Organize the capacity configuration scheme generated in step S3 and the attenuation compensation strategy generated in step S4 into a control parameter data packet. The data packet is organized in JSON format and contains the following fields: total system capacity. Battery module number Array, power allocation coefficients of each module Array and depth of charge / discharge limits for each module Array, capacity expansion time point Array, capacity expansion capacity value array, threshold Capacity adequacy threshold Perform format verification on the data packets to ensure that the lengths of each array are consistent and the values ​​are within a reasonable range. Generate a checksum for the data packets using the MD5 algorithm and append it to the end of the data packets. S52: Establish a communication connection with the energy storage system controller. The controller uses an embedded industrial computer running a Linux operating system. It connects to the host computer via an Ethernet interface and uses the TCP / IP protocol for data transmission. The host computer starts a Socket client program to initiate a connection request to the controller's IP address and port number. The Socket server program on the controller side receives the connection request and establishes a communication link. After the communication is established, the host computer sends a data packet type identifier 0x01, indicating that the capacity configuration scheme data is about to be transmitted. The controller returns an acknowledgment signal 0xACK. S53: The host computer sends control parameter data packets to the controller through the established communication link, using a packet transmission method. Each data packet does not exceed 1024 bytes, and the header of the data packet contains the packet sequence number, total number of packets, and data length fields. After receiving each data packet, the controller verifies the continuity of the packet sequence number and the correctness of the data length, and returns a reception confirmation signal to the host computer. After all data packets have been transmitted, the controller performs an MD5 check on the received complete data packets, compares the calculated checksum with the checksum in the data packet, and if they match, sends a data reception success signal 0x02 to the host computer. If they do not match, a retransmission request signal 0x03 is sent. The host computer decides whether to retransmit the data packet based on the return signal. S54: After receiving the complete control parameter data packet, the controller parses the JSON data into an internal data structure and starts the real-time monitoring program. The monitoring program collects the current data from the BMS of each battery module every second. The value and remaining capacity value will be the actual Values ​​and data packets A comparison is made when any module's... Below At that time, the controller generates a capacity expansion trigger signal; simultaneously, the monitoring program calculates the total available capacity of the system. The formula is: ;in This represents the total available capacity of the system. This refers to the number of battery modules; For the first The rated capacity of each module; For the first The current health status of each module; when Below load demand When the energy reaches 120% of the required level, the controller also triggers a capacity expansion signal; S55: The controller determines the power allocation factor based on the received power allocation factor. The array calculates the power commands for each battery module in real time, and receives the total power command from the upper-level energy management system. At that time, the controller performs power allocation calculations, for the first... Module calculation ,Will Transmitted via CAN bus at a baud rate of 250kbps to the [unclear - likely a specific location]. Each module's DC / DC converter transmits messages in standard CAN frame format, with a message ID of 0x300+. The data field contains the power command value and the operating mode flag. After receiving the power command, the DC / DC converter adjusts the PWM duty cycle of the switching transistor according to the power command value. The PWM frequency is 20kHz, which realizes precise control of the charging and discharging power. S56: The controller feeds back system operation data to the host computer optimization system every 24 hours. The feedback data includes the actual operation data of each module. The actual charging and discharging power, cumulative throughput, and temperature distribution are analyzed by the host computer optimization system after receiving feedback data. The actual operating data is compared with the predicted data, and the prediction deviation is calculated using the following formula: When the absolute value of the deviation exceeds 10%, the model retraining procedure is triggered, and the LSTM neural network is incrementally trained using the latest running data to update the model parameters. Then, steps S2 to S4 are re-executed to generate a corrected capacity configuration scheme and compensation strategy. Steps S51 to S55 are executed again to send the new scheme to the controller, forming a rolling optimization closed-loop control to achieve adaptive optimization operation of the energy storage system throughout its entire life cycle.

[0011] The beneficial effects achieved by this invention are as follows: First, this invention achieves comprehensive monitoring and data accumulation of battery module operating status by establishing a multi-parameter real-time data acquisition system and a historical operating database. This innovation employs simultaneous acquisition of multiple parameters, including voltage, current, temperature, internal resistance, cycle count, and depth of charge / discharge, providing a complete and reliable data foundation for subsequent capacity degradation prediction. Through regular electrochemical impedance spectroscopy internal resistance testing and cycle counting procedures, it accurately captures changes in key characteristic parameters during battery aging, significantly improving the accuracy and timeliness of battery health status assessment and laying a solid foundation for the scientific management and predictive maintenance of energy storage systems.

[0012] Secondly, this invention combines physical models of temperature-accelerated aging factors and cyclic aging factors with machine learning algorithms from long short-term memory neural networks to establish an accurate predictive model for battery capacity degradation trajectory. This hybrid prediction model not only reflects the physical mechanism of battery aging but also possesses the ability to learn and generalize from complex time-series data, accurately predicting long-term capacity degradation trends. By quantifying the impact of temperature on aging through the Arrhenius equation, reflecting the cumulative effect of charge-discharge cycles through the cyclic aging model, and combining the time-series memory characteristics of LSTM networks, high-precision long-term prediction of battery capacity degradation is achieved, providing a scientific basis for the full life-cycle planning of energy storage systems.

[0013] Third, this invention constructs a capacity configuration optimization objective function aimed at maximizing net present value. It comprehensively considers economic factors throughout the entire lifecycle, including system investment costs, operating revenue, and maintenance costs, and employs a dynamic programming algorithm to solve for the optimal initial capacity configuration and phased capacity expansion scheme. This optimization method determines the economically optimal capacity configuration strategy under multiple constraints, including load demand, capacity adequacy, and battery lifespan, avoiding the problems of over-configuration or under-configuration in traditional methods. Through dynamic programming for phased capacity expansion, investment costs are rationally dispersed and the time value of money is maximized, significantly improving the economic feasibility and return on investment of energy storage system projects.

[0014] Fourth, this invention establishes a dynamic degradation compensation strategy and a closed-loop optimization control mechanism based on capacity degradation prediction results, realizing adaptive optimization operation of the energy storage system. By dynamically adjusting the charge / discharge depth limits and power allocation ratios of each battery module, further battery degradation is effectively delayed and balanced use of each module is achieved, extending the overall system lifespan. By establishing a capacity balance evaluation mechanism and a real-time monitoring feedback system, model retraining and scheme updates are automatically triggered when the prediction deviation exceeds a threshold, forming a closed-loop control system of continuous learning and rolling optimization. This adaptive mechanism addresses changes in battery aging characteristics and fluctuations in load demand, ensuring that the energy storage system maintains optimal operating conditions throughout its entire lifespan, improving system reliability and economy. Attached Figure Description

[0015] Figure 1 This is a flowchart of the energy storage system capacity configuration optimization method based on battery capacity decay trajectory prediction of the present invention. Detailed Implementation

[0016] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The present invention is not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Reference Figure 1 The present invention provides a capacity configuration optimization method for energy storage systems based on battery capacity decay trajectory prediction. By establishing an accurate battery capacity decay prediction model and combining it with a dynamic programming algorithm, the method achieves capacity configuration optimization throughout the entire life cycle of the energy storage system. This method solves the technical problem of system performance degradation and poor economic benefits caused by insufficient consideration of battery aging characteristics in traditional energy storage system capacity configuration.

[0018] The first step is to collect operational status data from the energy storage system. Voltage sensors are installed at the positive and negative terminals of each battery module in the energy storage system. These voltage sensors are based on the Hall effect principle and feature high accuracy and good linearity. A Hall current sensor is connected in series in the main current path of the charging / discharging circuit. Based on the Hall effect principle, the Hall current sensor indirectly measures the current magnitude by measuring the magnetic field strength generated by the current, offering advantages such as non-contact measurement and fast response. A temperature sensor is attached to the center of the battery module's outer surface. A thermistor-type temperature sensor is preferred, as it features stable temperature coefficients and high measurement accuracy. The signal output terminals of each sensor are connected to the analog input port of the data acquisition unit via shielded cables. The shielded cables effectively suppress electromagnetic interference, ensuring accurate signal transmission.

[0019] After starting the data acquisition unit, an appropriate sampling frequency needs to be set to obtain accurate battery operating status information. The voltage sampling frequency is set to 5Hz, which captures the slow changes in battery voltage; the current sampling frequency is set to 10Hz, a relatively high frequency that accurately records the dynamic changes in current during charging and discharging; the temperature sampling frequency is set to 1Hz, which is sufficient for temperature monitoring due to the relatively slow temperature changes. The data acquisition unit synchronously samples the analog signals output from each sensor. Synchronous sampling ensures accurate time correspondence between parameters. A 16-bit analog-to-digital converter converts the analog signals into digital signals; the 16-bit precision provides sufficient measurement resolution to obtain the voltage values ​​at various times. Current value and temperature value .

[0020] A cycle counting program is set up in the battery module's charge / discharge controller. This program identifies charge / discharge cycles by monitoring changes in the battery's State of Charge (SOC). SOC refers to the ratio of the battery's current remaining capacity to its rated capacity, usually expressed as a percentage. The number of cycles is counted when the battery module completes a full cycle from charged to discharged and back to charged. Automatically increment by 1, and simultaneously record the initial state of charge for this cycle. and end of state of charge Calculate the depth of charge / discharge for this cycle. Depth of charge / discharge (DOD) is an important parameter for measuring the extent of battery use, and its calculation formula is as follows: ;in Depth of charge / discharge (%) It is the decimal form of the initial charge state of the cycle, with a value range of 0-1; The value is a decimal representation of the state of charge at the end of the cycle, ranging from 0 to 1. Accurate recording of the depth of charge and discharge provides fundamental data for subsequent battery aging analysis. Internal resistance testing of the battery module is performed every 24 hours. Internal resistance is a key parameter reflecting the internal electrochemical state of the battery, and its changes effectively reflect the degree of battery aging.

[0021] Internal resistance was measured using an electrochemical impedance spectroscopy (EIS) testing device. EIS is a technique for studying the impedance characteristics of an electrochemical system by applying a small AC excitation signal and analyzing the response. During the test, the normal charging and discharging operation of the battery module needed to be paused. An AC excitation current with a frequency of 1 kHz and an amplitude equal to 0.1C of the battery's rated capacity was applied to the battery module via a signal generator, where C represents the rated capacity ratio of the battery, and 0.1C means that the amplitude of the excitation current is one-tenth of the rated capacity. A dual-channel oscilloscope was used to simultaneously record the voltage response waveform and the input current waveform across the battery terminals. The phase difference and amplitude ratio of the voltage and current signals were analyzed using a Fast Fourier Transform (FFT). FFT is an efficient frequency domain analysis algorithm that converts time-domain signals into frequency-domain signals for analysis, and the AC internal resistance was calculated. After the test is completed, the battery module is restored to normal operation. This periodic internal resistance test method can detect changes in the battery's aging condition in a timely manner.

[0022] A historical operation database is established to store and manage the collected operation data. The data tables include timestamps. Battery module number ,Voltage Current ,temperature Internal resistance Number of loops Depth of charge and discharge and cumulative throughput Fields such as... Cumulative throughput is an important parameter reflecting battery usage intensity, calculated through integration. The calculation formula is: ;in Cumulative throughput (Ah); This represents the summation operator; For summation index; This represents the number of loops (times). For the first Average current (A) of each cycle; For the first Duration of the next cycle (h); This represents the absolute value operator. The collected and calculated data is written to the database sequentially in chronological order, and an index is created using the timestamp and battery module number as a composite primary key, providing an efficient data access foundation for subsequent data queries and analysis.

[0023] The next step is to establish a battery capacity degradation trajectory prediction model. Temperature data for each battery module will be extracted from the historical operational database. The data are arranged in a time series. For each temperature data point, the corresponding temperature acceleration factor is calculated by substituting it into the Arrheniuse equation. The Arrhenius equation describes the relationship between the reaction rate constant and temperature, and is used in battery aging analysis to quantify the effect of temperature on the aging rate. The calculation formula is: ;in The temperature acceleration factor (dimensionless). It is a natural exponential function; The activation energy (J / mol) represents the aging reaction activation energy of the battery. The activation energy reflects the ease or difficulty of the aging reaction inside the battery. For lithium iron phosphate batteries, a value of 25,000 J / mol is recommended, and for ternary lithium batteries, a value of 30,000 J / mol is recommended. The universal gas constant (J / (mol·K)) is taken as 8.314 J / (mol·K); For reference temperature (K), it is recommended to use a value of 298.15K, which corresponds to 25℃. The actual operating temperature (K) is obtained by adding 273.15 to the Celsius temperature. The calculated... The cumulative temperature-accelerated aging amount is obtained by integrating the value over time. This method accurately quantifies the cumulative impact of temperature changes on battery aging.

[0024] Extract the cumulative throughput of each battery module from the historical operation database. ,Will Substitute the values ​​into the cycle aging model to calculate the capacity decay due to cycle aging. The cycle aging model describes the decay of battery capacity with the number of charge-discharge cycles, and its calculation formula is as follows: ;in The capacity decay caused by cyclic aging (Ah); This is the cycle aging factor, which needs to be obtained through experimental calibration. For lithium iron phosphate batteries, it is recommended to conduct a 1C rate charge-discharge cycle test on a test battery with a rated capacity of 100Ah at 25℃, measuring the capacity every 100 cycles, plotting a double logarithmic curve of capacity decay versus cumulative throughput, and fitting the curve using the least squares method. Value, typically takes the value of ; Cumulative throughput (Ah); The cycle aging index (dimensionless) is obtained synchronously through the above fitting process, with a typical value of 0.6. This model calibration method based on experimental data ensures the accuracy of the cycle aging model. A time-varying capacity decay model is established, which comprehensively considers the combined effects of temperature aging and cycle aging, and more accurately describes the capacity decay process of the battery under actual operating conditions. The effects of temperature aging and cycle aging are superimposed to calculate the... Day's remaining battery capacity The calculation formula is: ;in For the first Remaining capacity per day (Ah); The initial rated capacity of the battery (Ah); Indicates the integration operator; For integration variables (days); The calendar aging rate factor (Ah / day) reflects the capacity decay rate of the battery under static conditions. It is recommended to obtain the value by placing the battery in an open-circuit environment at 25°C, measuring the capacity every 30 days, and recording the capacity decay rate over 180 days. A typical value is 0.005Ah / day. For the first Temperature acceleration factor of the sky (dimensionless); As of the date The cyclic aging capacity decay (Ah) is measured daily. The model is then backtested using historical data. Model accuracy is assessed by comparing predicted and measured values, ensuring the error between predicted and measured values ​​is less than 3%. This verification method guarantees the model's reliability.

[0025] A Long Short-Term Memory (LSTM) neural network was constructed for predicting battery capacity decay trajectories. LSTM is a special type of recurrent neural network that effectively handles long-sequence data and solves the gradient vanishing problem of traditional recurrent neural networks, making it particularly suitable for time-series prediction tasks with long-term dependencies, such as battery capacity decay. The network structure includes an input layer, three LSTM hidden layers, and an output layer. The input layer receives historical running data, including temperature sequences, current sequences, number of iterations, and cumulative throughput. The data is normalized before being input into the first LSTM layer. Normalization accelerates network training convergence and improves prediction accuracy. The first LSTM layer contains 64 neurons, each containing an input gate, a forget gate, and an output gate. These three gating structures are the core components of the LSTM, and the cell state is processed using the hyperbolic tangent (tanh) activation function. The output of the first layer is passed to the second LSTM layer, which contains 32 neurons; the output of the second layer is passed to the third LSTM layer, which contains 16 neurons. This progressively decreasing network structure design enables hierarchical extraction and abstraction of features. The output of the third layer is connected to the fully connected output layer. The number of neurons in the output layer is equal to the number of predicted time points. A linear activation function is used to output the capacity prediction values ​​for each future time point.

[0026] Training the LSTM network using historical data is a crucial step in the neural network's learning of data patterns. Historical data is divided into training samples according to time windows. Each sample contains 90 consecutive days of input data followed by 30 days of capacity decay data as labels. This time window design allows the network to learn short- and medium-term capacity change patterns. The Adaptive Moment Estimation (Adam) optimizer is used to minimize the Mean Square Error (MSE) loss function. The Adam optimizer combines the advantages of momentum gradient descent and adaptive learning rate, quickly and stably converging to the optimal solution. The formula for calculating the MSE loss function is: ;in The mean square error is (Ah²). To predict the number of samples (individuals); For the first Predicted capacity value (Ah) for each sample; For the first The true capacity (Ah) of each sample is set. The learning rate is set to 0.001, the batch size is 32, and the training is performed for 300 epochs. The model performance is evaluated on the validation set every 50 epochs, and the model parameters with the minimum validation loss are saved. This early stopping mechanism prevents the model from overfitting.

[0027] A trained LSTM model was used to predict future capacity decay trajectories. The model was input with operational data from the last 90 days, and output capacity predictions for the next 1 to 20 years, with prediction intervals of 30 days, resulting in a total of 240 predicted data points. The prediction results were plotted as a capacity decay prediction curve, with time on the x-axis and remaining capacity on the y-axis. The health status of each prediction point was also calculated. Battery health status is an important indicator for measuring the remaining performance of a battery, and its calculation formula is as follows: ;in For the first Health status at any given time (%) For the first Predicted remaining capacity (Ah) at time 1; The initial rated capacity (Ah). The capacity decay prediction curve and... Curve data is stored as a comma-separated values ​​(CSV) file for subsequent capacity configuration optimization. The CSV format has good versatility and readability.

[0028] Then, the energy storage system capacity configuration optimization step is performed. Obtaining load demand data for the energy storage system is the foundation for capacity configuration optimization; historical load power curves are exported from the user-side Energy Management System (EMS). The EMS (Electrical Management System) is responsible for monitoring and controlling user-side electrical equipment. The data spans the past year, with a granularity of 15 minutes, totaling 35,040 data points. This high-density data collection effectively reflects load variation patterns. Statistical analysis of the load data is performed to calculate the maximum daily load power. Average load power Load peak-valley difference And load factor, the formula for calculating load factor is: ;in Average load power (kW); The maximum load power is represented in kW. The load factor reflects the smoothness of load changes; a higher load factor indicates a smoother load change. Based on historical load patterns and future electricity consumption growth forecasts, the load is decomposed into trend, periodic, and random components using a time series decomposition method. Time series decomposition is a classic time series analysis method that reveals the inherent patterns of load changes and extrapolates to generate load demand curves for the next 10-20 years.

[0029] An objective function for capacity configuration optimization is constructed, which comprehensively considers the investment cost, operating revenue, and life-cycle economics of the energy storage system. The objective is to maximize the net present value (NPV) of the energy storage system over its entire life cycle. NPV is an important indicator for evaluating the economic benefits of investment projects, and its calculation formula is as follows: ; in Net present value (RMB); This represents the summation operator; Indexed by year; For the system design life (in years), a value of 15 years is recommended. For the first Annual operating revenue (RMB / year); For the first Annual operating costs (RMB / year) include electricity, maintenance, and labor costs; The discount rate (dimensionless) is recommended to be 0.08. Investment cost (RMB) for initial capacity configuration; Index for the number of capacity expansions; The number of times the capacity is expanded. For the first Investment cost for the next capacity expansion (RMB); For the first The year in which the capacity expansion occurred.

[0030] Operating revenue The calculation needs to consider various profit models for energy storage systems, including peak-valley electricity price arbitrage, ancillary service compensation, and capacity leasing revenue. The calculation formula is as follows: ; 365 represents the number of days in a year; Rated power (kW) of the energy storage system; This represents the daily peak operating time (hours), and is recommended to be 4 hours based on local electricity pricing policies. The peak hour electricity price (RMB / kWh) can be found by checking the local electricity price list. Off-peak electricity price (RMB / kWh); For round-trip efficiency (dimensionless), considering energy loss during charging and discharging, a value of 0.90 is recommended; The recommended daily call duration for auxiliary services (h / day) is 1.5h / day, based on the power grid dispatching requirements. Ancillary service compensation price (RMB / kWh); For the first Average annual available capacity (kWh); The annual price for capacity leasing is (RMB / kWh / year). This diversified revenue model improves the economic efficiency of energy storage systems.

[0031] Set constraints for capacity allocation and establish a set of constraint inequalities to ensure the technical feasibility and security of the capacity allocation scheme. First, initial capacity constraints... ;in Initial configuration capacity (kWh); The minimum continuous discharge duration (h) is recommended to be 2h; For the safety margin factor (dimensionless), a value of 1.2 is recommended. This constraint ensures that the energy storage system has sufficient capacity to cope with maximum load demand. Secondly, the capacity adequacy constraint applies to any point during system operation. ,satisfy ;in For the first The available capacity of the system at any given time (kWh) is equal to the sum of the remaining capacity of all battery modules multiplied by their state of health. For the first Load power (kW) at any given time; The load response time (h) is recommended to be 1h; 1.1 represents a 10% reserve capacity factor. This constraint ensures that the energy storage system can meet load demands throughout its entire operating cycle. Third, battery life constraint: when any battery module's lifespan reaches its maximum capacity... When the battery performance drops to 80%, the module must be replaced or the system capacity expanded. This constraint is based on the actual situation of battery performance degradation. Fourth, there is an economic constraint: the system's investment payback period must not exceed 8 years. This constraint ensures that the project has reasonable economic benefits.

[0032] Dynamic programming is used to solve the capacity allocation optimization problem. Dynamic programming is an optimization algorithm for solving multi-stage decision problems, and it is particularly suitable for handling optimization problems with time-series characteristics. The system lifetime is considered. Divided by year Each decision-making stage, each stage The state variable is the current available capacity of the system. The decision variable is the capacity expansion amount at this stage. The value ranges from 0 to the system's maximum expansion capacity. From the last stage... Starting with reverse recursion, the value function in the final stage This equals the net profit for that stage. For stage... The value function is calculated based on the Bellman optimality equation, which is the core theoretical foundation of dynamic programming. Its calculation formula is as follows: ; in For the first The stage state is The optimal value function (element) at that time; Represents all feasible decision variables Find the maximum value. Calculate sequentially. The value function for each stage records the optimal decision at each stage.

[0033] From the first stage Begin forward backtracking of the optimal decision sequence to determine the initial capacity configuration scheme. And the capacity expansion plan for each stage. For each stage that requires capacity expansion, record the expansion time. and expanded capacity A capacity configuration schedule is generated. This schedule includes key information such as year, current capacity, capacity expansion amount, total capacity after expansion, annual investment cost, annual operating income, and annual net cash flow, providing decision-makers with a clear investment and operational plan. Mathematical analysis software is used to create a capacity configuration optimization chart. The horizontal axis represents the year, the left vertical axis represents capacity, and the right vertical axis represents net present value. The chart uses a stepped curve to represent the change in capacity over time, a bar chart to represent the net cash flow for each year, and a cumulative curve to represent the change in net present value. This visualization method facilitates the understanding of the optimization results. The optimization results are output as Excel files and PDF reports for decision-makers' reference, ensuring the practicality of the optimization results.

[0034] Next, the calculation steps for the capacity degradation compensation strategy are performed. The capacity degradation prediction curve data generated in the previous steps is read, and the calculation is performed for each battery module in the system. Extract its future time points Predicted remaining capacity Calculate the corresponding health status The calculation formula is: ;in For the first The battery module is in the first Health status at any given time (%) For the first The battery module is in the first Predicted remaining capacity (Ah) at time 1; For the first The initial rated capacity (Ah) of each battery module. The capacity of each module... Data is stored in time series, and a system is established. The time matrix uses the module number as the row index and the time point as the column index. This data organization method facilitates subsequent strategy calculations.

[0035] A dynamic charge / discharge depth adjustment strategy is established. This strategy dynamically adjusts the charge / discharge depth limits based on the battery's health state, effectively delaying further battery degradation. For battery modules whose health state is gradually deteriorating, limiting their charge / discharge depth alleviates aging stress; this is a scientific strategy based on the battery aging mechanism. For the [missing information - likely a specific battery module or module]... The battery module is in the first At any given time, calculate its permissible depth of charge and discharge limit. The calculation formula is: ;in For the first The module in the first Limits on depth of charge and discharge at any given time (%); For the standard depth of charge / discharge (%) of the system design, a value of 85% is recommended. The minimum permissible health status threshold (%) is recommended to be 80%; 0.5 is an adjustment index (dimensionless) used to control the rate of change of the depth limit, and the selection of this index is based on experimental research on the battery aging mechanism. In the Battery Management System (BMS), the calculated... The value is set as the charge / discharge cutoff condition for this module, when the module's state of charge... Exceeding the permitted range When this happens, the BMS stops the charging and discharging operation of the module, and this protection mechanism effectively extends the battery's lifespan.

[0036] A dynamic power allocation optimization strategy is established. This strategy rationally allocates power based on the differences in the health status of each battery module, preventing modules with poor health from being overloaded. When there are differences in the health status of the battery modules in the system, the strategy is based on the health status of each module... Proportional allocation of charging and discharging power ensures balanced use of each module. For the first... Each battery module is used to calculate its power distribution coefficient. The calculation formula is: ;in For the first Power distribution coefficient of each battery module (dimensionless). For the first Current health status (%) of each battery module; This represents the total number of battery modules in the system. This represents the summation operator; For summation index. Power allocation coefficient. This reflects the proportion of total power that the module should bear. Modules with higher health status bear more power, while modules with lower health status bear less power, thereby achieving balanced use of each module. This strategy maximizes the overall performance of the system.

[0037] Based on the power allocation factor and the total system power requirement, calculate the actual charge and discharge power command for each battery module. For the first... Each battery module has a charging and discharging power. The calculation formula is: ;in For the first Charging and discharging power (kW) of each battery module; The total power command (kW) received by the energy storage system is issued by the superior energy management system. The conversion efficiency (dimensionless) of the DC / DC power converter is recommended to be 0.95 during charging and 0.96 during discharging. This efficiency difference reflects the characteristics of the power converter in different operating modes. The calculated power command will then be used. Transmitted to the first via the Controller Area Network (CAN) communication bus. The DC / DC converter for each battery module utilizes the CAN bus, a fieldbus widely used in industrial automation, known for its high reliability and real-time performance. The converter adjusts the duty cycle of the Pulse Width Modulation (PWM) signal according to power commands to control the actual charging and discharging current. PWM is a technology that controls power output by changing the pulse width.

[0038] A capacity degradation compensation effectiveness evaluation mechanism is established. This mechanism is used to periodically evaluate the implementation effect of the compensation strategy and make dynamic adjustments. The implementation effect of the compensation strategy is evaluated every 7 days, and the system's capacity balance degree (CBD) is calculated. CBD is a quantitative indicator that measures the degree of capacity balance among battery modules, and its calculation formula is as follows: ;in The capacity balance (dimensionless) ranges from 0 to 1, with values ​​closer to 1 indicating a more balanced capacity among the modules. Standard deviation is a function used to measure the dispersion of data. It is a function of average value; arrive This represents the health status (%) of each battery module. When... A value below 0.90 indicates a significant difference in capacity between modules, requiring adjustment of the power allocation factor to increase the power of modules with lower health status. Reduce its load and increase the load of modules with higher health status. The system increases its load, recalculates and issues power commands to achieve dynamic balance, and this adaptive adjustment mechanism continuously optimizes system performance.

[0039] Finally, the control command issuance and closed-loop optimization steps are executed. The capacity configuration scheme and attenuation compensation strategy generated in the previous steps are organized into a control parameter data packet. The data packet is organized in JavaScript Object Notation (JSON) format. JSON is a lightweight data exchange format with good readability and cross-platform compatibility. The data packet contains the total system capacity. Battery module number Array, power allocation coefficients of each module Array and depth of charge / discharge limits for each module Array, capacity expansion time point Array, capacity expansion capacity value array, threshold Capacity adequacy threshold Fields such as [list of fields]. The data packet undergoes format validation to ensure that all array lengths are consistent and values ​​are within a reasonable range. A checksum is generated for the data packet using Message-Digest Algorithm 5 (MD5), a widely used hash function, to generate a unique digital fingerprint. This checksum is appended to the end of the data packet to verify the integrity of the data transmission.

[0040] A communication connection is established with the energy storage system controller, which is an embedded industrial computer running a Linux operating system. Linux offers high stability and real-time performance, making it suitable for industrial control applications. The system connects to the host computer via an Ethernet interface, using Transmission Control Protocol / Internet Protocol (TCP / IP) for data transmission. TCP / IP, the foundational protocol of the Internet, ensures reliable data transmission. A Socket client program is launched on the host computer, initiating a connection request to the controller's IP address and port number. The Socket server program on the controller receives the connection request and establishes a communication link. After communication is established, the host computer sends a data packet type identifier (0x01) indicating that capacity configuration scheme data is about to be transmitted. The controller returns an acknowledgment signal (0xACK). This handshake mechanism ensures reliable communication.

[0041] The host computer sends control parameter data packets to the controller via the established communication link, employing a packet-based transmission method to handle the large data volume transmission requirements. Each data packet does not exceed 1024 bytes, and the packet header includes a packet sequence number, total number of packets, and data length fields. This packet structure design facilitates data reassembly at the receiving end. After receiving each data packet, the controller verifies the continuity of the packet sequence number and the correctness of the data length, and returns a reception confirmation signal to the host computer. After all data packets have been transmitted, the controller performs an MD5 checksum on the received complete data packets, comparing the calculated checksum with the checksum in the data packet. If they match, the controller sends a data reception success signal (0x02) to the host computer; if they do not match, it sends a retransmission request signal (0x03). The host computer decides whether to retransmit the data packet based on the returned signal. This reliability mechanism ensures the integrity and accuracy of data transmission.

[0042] After receiving the complete control parameter data packet, the controller parses the JSON data into an internal data structure and starts the real-time monitoring program. The monitoring program collects current data from the BMS of each battery module every second. The value and remaining capacity value will be the actual Values ​​and data packets A comparison is made when any module's... Below At this time, the controller generates a capacity expansion trigger signal. Simultaneously, the monitoring program calculates the total available system capacity. The calculation formula is: ;in Total available system capacity (kWh); Number of battery modules (units); For the first Rated capacity of each module (kWh); For the first The current health status (%) of each module. When Below load demand When the energy reaches 120% of its capacity, the controller also triggers a capacity expansion signal. This dual triggering mechanism ensures that the system can respond promptly to situations where the capacity is insufficient.

[0043] The controller uses the received power allocation factor An array is used to calculate the power commands for each battery module in real time. This is done when the total power command is received from the higher-level energy management system. At that time, the controller performs power allocation calculations, for the first... Module calculation ,Will Transmitted via CAN bus at a baud rate of 250kbps to the [unclear - likely a specific location]. The DC / DC converter module uses a standard CAN bus communication rate of 250kbps, meeting real-time control requirements. The transmitted message format is a standard CAN frame with a message ID of 0x300+i. The data field includes the power command value and operating mode flag. Upon receiving the power command, the DC / DC converter adjusts the PWM duty cycle of the switching transistors according to the command value. The PWM frequency is 20kHz, a frequency selection that strikes a good balance between switching losses and control accuracy, enabling precise control of charging and discharging power.

[0044] The controller feeds back system operating data to the host computer optimization system every 24 hours, forming a closed-loop control mechanism. The feedback data includes the actual operating data of each module. Actual charging and discharging power, cumulative throughput, and temperature distribution data provide the foundation for continuous model improvement. After receiving feedback data, the host computer optimization system compares and analyzes the actual operating data with the predicted data, calculating the prediction deviation. The deviation calculation formula is as follows: When the absolute value of the deviation exceeds 10%, the model retraining procedure is triggered. The LSTM neural network is incrementally trained using the latest running data to update the model parameters. This online learning mechanism continuously improves the prediction accuracy. Then, the steps of capacity decay trajectory prediction, capacity configuration optimization, and decay compensation strategy calculation are re-executed to generate a corrected capacity configuration scheme and compensation strategy. The new scheme is then sent to the controller again, forming a rolling optimization closed-loop control to achieve adaptive optimization operation throughout the entire life cycle of the energy storage system.

[0045] The method of this invention establishes an accurate battery capacity degradation prediction model, combines dynamic programming algorithm to optimize capacity configuration, and adopts dynamic compensation strategy and closed-loop feedback mechanism to significantly improve the economic benefits and operational reliability of energy storage systems, providing an effective technical solution for the engineering application of energy storage systems.

[0046] Example 1: To implement the method of the present invention, this example provides an energy storage system capacity configuration optimization system based on battery capacity degradation trajectory prediction. This system, through an integrated hardware architecture and intelligent data processing unit, achieves accurate prediction of battery capacity degradation and dynamic optimization of capacity configuration in the energy storage system. The system consists of seven functional modules: a data acquisition subsystem, a data processing unit, a capacity degradation prediction unit, a capacity configuration optimization unit, a compensation strategy generation unit, a control command output unit, and a display and interaction unit. These modules exchange data and work collaboratively through a high-speed data bus and a standard communication interface.

[0047] The data acquisition subsystem is responsible for real-time acquisition of operating status parameters of each battery module in the energy storage system, including voltage sensor groups, current sensor groups, temperature sensor groups, and an internal resistance testing unit. The voltage sensors, based on the Hall effect principle, are installed at the positive and negative terminals of the battery module. The current sensors are connected in series in the main charge / discharge circuit. The temperature sensors are attached to the surface of the battery module casing. The internal resistance testing unit periodically tests the battery's AC internal resistance using electrochemical impedance spectroscopy. The data acquisition unit receives signals from each sensor, performs analog signal conversion, and performs preliminary processing.

[0048] The data processing unit receives raw data from the data acquisition unit, performs storage, preprocessing, and preliminary analysis, and includes a data storage module, a data preprocessing module, and a cycle counting module. The data storage module uses a solid-state drive (SSD), the data preprocessing module performs filtering and normalization, and the cycle counting module accumulates the number of battery charge-discharge cycles. The data processing unit connects to subsequent modules via a data bus.

[0049] The capacity degradation prediction unit establishes a battery capacity degradation prediction model based on historical operating data to predict the future capacity degradation trajectory. This unit includes an aging factor calculation module and a neural network processor. The aging factor calculation module calculates the temperature acceleration factor and cycle aging amount through the Arrhenius equation and the cycle aging model. The neural network processor uses a long short-term memory network structure for capacity prediction, and the model parameters are stored in the model memory.

[0050] The capacity configuration optimization unit formulates the optimal capacity configuration scheme based on capacity decay prediction results and load demand. This unit includes a load data interface, an economic parameter database, an optimization algorithm processor, an objective function calculation module, and a constraint condition judgment module. The load data interface obtains load curves from the energy management system, the economic parameter database stores parameters such as battery prices and electricity prices, and the optimization algorithm processor runs a dynamic programming algorithm to solve the objective function, considering net present value maximization and various constraints.

[0051] The compensation strategy generation unit calculates the capacity degradation compensation strategy for each battery module based on the capacity configuration scheme, including a health status calculation module, a charge / discharge depth adjustment module, a power allocation calculation module, and a balance evaluation module. The health status calculation module calculates the health status of each module based on the predicted capacity, the charge / discharge depth adjustment module dynamically limits the charge / discharge depth, the power allocation calculation module allocates power according to the health status ratio, and the balance evaluation module periodically evaluates the compensation effect.

[0052] The control command output unit sends the capacity configuration scheme and attenuation compensation strategy to the energy storage system controller and receives feedback data to achieve closed-loop control. This unit includes a data packaging module, a communication interface module, a data verification module, a feedback data receiving module, a deviation analysis module, and a rolling optimization trigger module. The data packaging module organizes parameters into JSON format; the communication interface module communicates with the controller via Ethernet; the data verification module ensures data integrity; the feedback data receiving module collects actual operating data; the deviation analysis module compares predicted and actual values; and the rolling optimization trigger module re-optimizes when deviations exceed limits.

[0053] The display and interaction unit serves as the human-computer interface, showing users the system's operating status and optimization results. It includes a touchscreen display, a parameter input keyboard, and a graphics rendering module. The touchscreen displays predicted curves and configuration schemes, the parameter input keyboard allows users to input design parameters, and the graphics rendering module generates visual charts.

[0054] The entire system is interconnected through standardized interfaces to form a closed-loop control system. Data is collected from sensors, processed, predicted, optimized, and generated into strategies. Finally, control commands are issued, the controller executes the commands and feeds back the data, and the system performs rolling optimization based on the feedback to ensure that the energy storage system maintains optimal operating status and economic benefits throughout its entire life cycle.

[0055] Example 2: Example 2 applies the method of the present invention to a 10MW / 20MWh lithium iron phosphate energy storage system project in an industrial park. This industrial park mainly includes manufacturing workshops, office buildings, and supporting facilities. The daily electricity load fluctuates greatly, with a peak-to-valley difference of up to 60%, urgently requiring an energy storage system for peak shaving and valley filling, as well as backup power supply.

[0056] During project implementation, the energy storage system consists of 20 battery modules, each with a rated capacity of 1MWh and an initial configuration capacity of 20MWh. Following step S1 of the method of this invention, an LV25-P type Hall voltage sensor is installed on each battery module to monitor voltage, an HO50-S type Hall current sensor is installed to monitor charging and discharging current, and a PT100 platinum resistance temperature sensor is attached to monitor operating temperature. The data acquisition unit uses an AD7606 type 16-bit synchronous sampling ADC chip, with the voltage sampling frequency set to 5Hz, the current sampling frequency set to 10Hz, and the temperature sampling frequency set to 1Hz. The internal resistance testing unit uses a 33220A type arbitrary waveform generator and a TDS2024C type oscilloscope to perform AC internal resistance tests on each battery module at 1kHz and 0.1C rate every 24 hours.

[0057] The data processing unit established a historical operation database, using a MySQL 5.7 database management system to store operational status data. After six months of data accumulation, the database recorded over 2 million operational data entries, including key parameters such as voltage, current, temperature, internal resistance, cycle count, and cumulative throughput for each module.

[0058] In step S2, an LSTM capacity decay prediction model is built based on accumulated historical data. A Jetson Nano embedded GPU module is used as the neural network processor to construct a network structure containing an input layer, three LSTM hidden layers, and an output layer. The first LSTM layer contains 64 neurons, the second layer contains 32 neurons, and the third layer contains 16 neurons. Using 90 consecutive days of input data, the model predicts capacity decay for the next 30 days. It is trained for 300 epochs using the Adam optimizer with a learning rate of 0.001 and a batch size of 32. After training, the model achieves a prediction accuracy of 97.2% on the validation set, with the prediction error controlled within 3%.

[0059] In step S3, the load power curves for the past 12 months were exported from the park's energy management system, with a time granularity of 15 minutes and a total of 35,040 data points. Statistical analysis showed that the maximum daily load power was 8.5MW, the average load power was 5.2MW, and the load factor was 61.2%. An optimization function was established with the goal of maximizing the net present value (NPV), a system design life of 15 years, and a discount rate of 8%. Operating revenue was considered from peak-valley arbitrage, ancillary service revenue, and capacity leasing revenue, with peak-hour electricity price of RMB 1.2 / kWh, off-peak electricity price of RMB 0.4 / kWh, and ancillary service compensation price of RMB 0.8 / kWh. A dynamic programming algorithm was used to solve the problem, determining the initial capacity configuration to be 20MWh, and three capacity expansion plans were formulated, to be implemented in years 5, 10, and 13, respectively.

[0060] Step S4 implements a capacity degradation compensation strategy. Based on the predicted State of Health (SOH) value of each module, the depth of charge / discharge limits and power allocation ratios are dynamically adjusted. When a module's SOH drops to 90%, the depth of charge / discharge is limited from 85% to 75%. Power allocation is based on the SOH ratio of each module, with modules in higher health receiving more power to achieve balanced utilization. A Capacity Balance Index (CBD) is established for evaluation, maintaining the CBD value above 0.92.

[0061] Step S5 sends the optimization scheme to the energy storage system controller via TCP / IP protocol. The controller is implemented using an STM32F407 microcontroller. The controller collects the SOH value of each module every second. When it detects that the SOH of any module is lower than 80% or the total available capacity of the system is insufficient, it automatically triggers a capacity expansion command. At the same time, it feeds back actual operating data every 24 hours. The system performs rolling optimization based on the prediction deviation. When the deviation exceeds 10%, it automatically retrains the model and updates the configuration scheme.

[0062] After 18 months of actual operation, the energy storage system has demonstrated excellent performance. The average system capacity utilization rate reached 92.5%, the SOH difference between modules was controlled within 5%, the prediction accuracy remained stable at over 96%, and the net present value over the entire life cycle increased by 12.8% compared to the initial assessment.

[0063] Comparative Example 1 uses the traditional static capacity configuration method. This method does not consider the prediction of battery capacity degradation trajectory and only performs a one-time capacity configuration based on the initial load demand. During the configuration process, an empirical coefficient method is used to set the initial capacity to the energy capacity corresponding to 2.5 times the maximum load power, i.e., 8.5MW × 2.5h = 21.25MWh, which is rounded up to 22MWh.

[0064] This method does not establish a capacity degradation prediction model, does not perform dynamic capacity expansion planning, and only adopts a uniform power distribution strategy during operation, with each battery module charging and discharging at equal power, without considering the differences in the health status of each module. Only when the system capacity decays to below 80% of the design capacity will the replacement of all battery modules be considered.

[0065] Comparative Example 2 employs a capacity prediction method based on empirical formulas. This method uses a simplified linear degradation model to predict battery capacity, assuming that the battery capacity decays linearly at a fixed rate of 2% per year, without considering the influence of factors such as temperature and cycle depth on the degradation rate. Capacity configuration adopts a static programming method, pre-determining a fixed capacity expansion schedule, with two capacity expansions performed in years 8 and 15.

[0066] The uniform power distribution strategy is still used during operation without dynamic adjustment. Although this method takes capacity decay into account, its prediction accuracy is limited and it cannot accurately reflect the actual battery aging process.

[0067] Comparative Example 3 employs a capacity prediction method based on Support Vector Machines (SVM). This method uses the same historical running data but employs the SVM algorithm to build a capacity decay prediction model. The SVM model uses a Radial Basis Function (RBF) kernel, and the hyperparameters, including the penalty parameter C and the kernel function parameter γ, are optimized through grid search.

[0068] Capacity configuration optimization also employs a dynamic programming algorithm, but due to the limited processing capability of the SVM model for time-series data, the prediction accuracy is relatively low. During operation, the same capacity decay compensation strategy as in Example 2 is used.

[0069] Comparative experiments were conducted according to relevant standards such as GB / T36558-2018 "Lithium-ion Batteries for Power Storage", IEC61960-3 "Test Procedures for Lithium-ion Battery Packs and Systems", and GB / T36276-2018 "Technical Specifications for Lithium-ion Battery Management Systems for Power Storage". Capacity decay testing was conducted according to GB / T36558-2018, using a 1C rate charge-discharge cycle test at an ambient temperature of 25℃±2℃, with capacity measured every 100 cycles. Internal resistance testing was conducted according to IEC61960-3, applying an AC excitation current at a frequency of 1kHz and measuring the AC internal resistance value. System performance testing was conducted according to GB / T36276-2018, evaluating key indicators such as usable capacity, charge-discharge efficiency, and response time. Economic evaluation was performed using the Net Present Value (NPV) method and the Internal Rate of Return (IRR) method.

[0070] Prediction accuracy was assessed using statistical indicators such as Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and Coefficient of Determination (R²). System availability was evaluated according to IEEE 1547-2018, "Distributed Power Sources and Power System Interconnection Standards".

[0071] The experimental results were compared. After 18 months of comparative experiments, the performance indicators of each method were compared as shown in Table 1: Table 1 Comparison of indicators between Example 2 and Comparative Examples 1-3

[0072]

[0073] As can be seen from Table 1, Example 2 is significantly better than the comparative example in key indicators such as net present value over the entire life cycle, payback period, capacity utilization, prediction accuracy, and system availability.

[0074] Specifically, Example 2 has a 41.8% higher net present value over its entire lifecycle compared to Comparative Example 1, a 2.3-year shorter payback period, and a 14.2 percentage point increase in capacity utilization. Compared to Comparative Example 2, which uses empirical formulas for prediction, Example 2 shows an 11.8 percentage point improvement in prediction accuracy and a 24.7% increase in net present value. Compared to Comparative Example 3, which uses SVM for prediction, Example 2 still has significant advantages in both prediction accuracy and economic benefits.

[0075] The SOH difference coefficient results show that Example 2 effectively achieves balanced use of each battery module, with the difference in health status between modules controlled at 3.2%, far lower than other comparative examples. This is beneficial for extending the overall lifespan of the system and reducing operation and maintenance costs.

[0076] The annual operation and maintenance cost comparison shows that Example 2, by adopting intelligent prediction and control strategies, can effectively reduce operation and maintenance costs by identifying problems in advance and optimizing operating parameters, saving 37.3% of operation and maintenance costs compared to Example 1.

[0077] Comprehensive experimental results show that the energy storage system capacity configuration optimization method based on battery capacity decay trajectory prediction proposed in this invention has significant technical advantages and economic benefits, and effectively improves the operating efficiency and economy of the energy storage system.

[0078] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing the capacity configuration of an energy storage system based on battery capacity degradation trajectory prediction, characterized in that, Includes the following steps: S1: Collect operating status data of battery modules in the energy storage system, including real-time voltage, real-time current, ambient temperature, number of charge-discharge cycles, depth of charge-discharge and internal resistance parameters, and establish a historical operating database; S2: Based on the collected operating status data, establish a battery capacity decay trajectory prediction model, calculate the battery capacity decay trajectory over time through temperature-accelerated aging factor and cycle aging factor, and obtain the capacity decay prediction curve for future time periods. S3: Based on the battery capacity decay prediction curve, combined with the load demand curve and economic indicators of the energy storage system, establish a capacity configuration optimization objective function, and solve the initial capacity configuration scheme and the phased capacity expansion scheme through dynamic programming algorithm. S4: Based on the capacity configuration scheme, calculate the capacity decay compensation strategy for each battery module, and adjust the charge / discharge depth limit and power allocation ratio to ensure that the system maintains a stable output capability throughout its entire life cycle. S5: Outputs the capacity configuration scheme and attenuation compensation strategy to the energy storage system controller. The controller dynamically adjusts the charging and discharging power of each battery module according to the real-time capacity status, and triggers a capacity expansion command when the capacity attenuation reaches a preset threshold, thereby achieving optimized operation of the energy storage system.

2. The method according to claim 1, characterized in that, Step S1 includes the following specific steps: S11: Install voltage sensors at the positive and negative terminals of each battery module in the energy storage system, install Hall current sensors in series on the main current channel of the charging and discharging circuit, attach temperature sensors to the center of the battery module housing, and connect the signal output terminals of each sensor to the analog input port of the data acquisition unit through shielded cables. S12: Start the data acquisition unit, set the voltage sampling frequency to 5Hz, the current sampling frequency to 10Hz, and the temperature sampling frequency to 1Hz. The data acquisition unit synchronously samples the analog signals output by each sensor, converts the analog signals into digital signals through a 16-bit analog-to-digital converter, and obtains the voltage values ​​at each moment. Current value and temperature value ; S13: Set a cycle counting program in the battery module's charge / discharge controller. When the battery module completes a full cycle from charging to discharging and back to charging, the cycle count is recorded. Automatically increment by 1, and simultaneously record the initial state of charge for this cycle. and end of state of charge Calculate the depth of charge and discharge for this cycle. The formula is: ;in Depth of charge / discharge; It is the decimal form of the initial charge state of the cycle, with a value range of 0-1; The decimal form of the charge state at the end of the cycle, with a value range of 0-1; S14: Every 24 hours, the internal resistance of the battery module is tested using an electrochemical impedance spectroscopy (EIS) testing device. During the test, the normal charging and discharging operation of the battery module is paused. An AC excitation current with a frequency of 1kHz and an amplitude of 0.1C of the battery's rated capacity is applied to the battery module through a signal generator. The voltage response waveform and the input current waveform across the battery terminals are recorded simultaneously using a dual-channel oscilloscope. The phase difference and amplitude ratio of the voltage and current signals are analyzed using Fast Fourier Transform (FFT) to calculate the AC internal resistance. After the test is completed, the battery module will be restored to normal operation. S15: Establish the historical operation database table structure. The data table includes the following fields: timestamp, moduleid, voltage. Current ,temperature Internal resistance Number of loops Depth of charge and discharge and cumulative throughput The cumulative throughput is calculated by integration, using the following formula: ;in For cumulative throughput; This represents the number of loop iterations. For the first Average current of each cycle; For the first The duration of the next cycle; The collected and calculated data are written into the database one by one in chronological order, and an index is created with the timestamp and battery module number as the composite primary key.

3. The method according to claim 1, characterized in that, Step S2 includes the following specific steps: S21: Extract temperature data of each battery module from the historical operation database. Arranged according to the time series, for each temperature data point, substitute it into the Arrhenius equation to calculate the corresponding temperature acceleration factor. The formula is: ;in It is a temperature acceleration factor; It is a natural exponential function; The activation energy for the battery aging reaction is 25000 J / mol for lithium iron phosphate batteries and 30000 J / mol for ternary lithium batteries. This is the universal gas constant, with a value of 8.314 J / (mol·K); For reference temperature, a value of 298.15K corresponds to 25℃; The actual operating temperature is obtained by adding 273.15 to the Celsius temperature; the calculated temperature... The value is integrated over time to obtain the cumulative temperature-accelerated aging amount. S22: Extract the cumulative throughput of each battery module from the historical operation database. ,Will Substitute into the cyclic aging model to calculate the capacity decay during cyclic aging. The formula is: ;in This refers to the capacity decay caused by cyclic aging. To determine the cycle aging factor, a test battery with a rated capacity of 100Ah was subjected to a 1C rate charge-discharge cycle experiment at 25℃. The capacity was measured every 100 cycles, and a double logarithmic curve of capacity decay versus cumulative throughput was plotted. The curve was then fitted using the least squares method to obtain the aging factor. Value, typically takes the value of ; For cumulative throughput; The cyclic aging index is obtained synchronously through the above fitting process, with a typical value of 0.

6. S23: Establish a time-varying capacity decay model for the battery, superimposing the effects of temperature aging and cycle aging, and calculate the... Day's remaining battery capacity The formula is: ;in For the first The remaining capacity for the day; This refers to the initial rated capacity of the battery. Indicates the integration operator; For integration variables; The calendar aging rate coefficient is obtained by placing the battery in an open-circuit environment at 25°C, measuring the capacity every 30 days, and recording the capacity decay rate over 180 days. A typical value is 0.005 Ah / day. For the first The temperature acceleration factor of the sky; As of the date The daily cyclic aging capacity decay; the model is backtested using historical data to ensure that the error between the model's predicted value and the measured value is less than 3%; S24: Construct a Long Short-Term Memory (LSTM) neural network for capacity decay trajectory prediction. The network structure includes an input layer, three LSTM hidden layers, and an output layer. The input layer receives historical running data, including temperature sequences, current sequences, loop counts, and cumulative throughput. After normalization, the data is input into the first LSTM layer. The first LSTM layer contains 64 neurons, each containing an input gate, a forget gate, and an output gate. The activation function processes the cell state; the output of the first layer is passed to the second LSTM layer, which contains 32 neurons; the output of the second layer is passed to the third LSTM layer, which contains 16 neurons; the output of the third layer is connected to the fully connected output layer, and the number of neurons in the output layer is equal to the number of predicted time points. A linear activation function is used to output the predicted capacity values ​​at each future time point. S25: Train the LSTM network using historical running data. Divide the historical data into training samples according to time windows. Each sample contains 90 consecutive days of input data and the subsequent 30 days of capacity decay data as labels. Use the Adam optimizer to minimize the mean squared error loss function MSE, as shown in the formula: ;in Mean squared error; To predict the number of samples; For the first Predictable capacity value for each sample; For the first The true size value of each sample; Set the learning rate to 0.001, the batch size to 32, train for 300 epochs, evaluate the model performance on the validation set every 50 epochs, and save the model parameters with the minimum validation loss. S26: Use the trained LSTM model to predict future capacity decay trajectories. Take the operational data from the last 90 days as model input, and the model outputs capacity predictions for the next 1 to 20 years. The prediction time interval is 30 days, resulting in a total of 240 predicted data points. Plot the prediction results as a capacity decay prediction curve, with time on the x-axis and remaining capacity on the y-axis. Simultaneously, calculate the health status of each prediction point. The formula is: ;in For the first Constant health status; For the first Predicted remaining capacity at any given time; The initial rated capacity; the capacity degradation prediction curve and Curve data is stored as a CSV file for subsequent capacity configuration and optimization.

4. The method according to claim 1, characterized in that, Step S3 includes the following specific steps: S31: Obtain load demand data from the energy storage system and export historical load power curves from the user-side energy management system (EMS). The data spans the past year, with a time granularity of 15 minutes, totaling 35,040 data points; statistical analysis is performed on the load data to calculate the daily maximum load power. Average load power Load peak-valley difference And the load factor, the formula is: Load factor = ;in Average load power; The maximum load power is determined by the time series decomposition method. Based on historical load patterns and future electricity consumption growth forecasts, the load is decomposed into trend components, periodic components, and random components using the time series decomposition method. The load demand curve for the next 10-20 years is then extrapolated. S32: Construct the objective function for capacity configuration optimization, with the objective of maximizing the net present value (NPV) of the energy storage system over its entire lifecycle. The formula is: ;in Net present value; The system design life is set at 15 years. For the first Annual operating revenue; For the first Annual operating costs, including electricity, maintenance, and labor costs; The discount rate is 0.

08. Investment costs for initial capacity configuration; For the number of capacity expansions; For the first The investment cost of subsequent capacity expansion; For the first The year in which the capacity expansion occurred; operating revenue Further calculations yield the following formula: ;in Rated power of the energy storage system; This represents the daily peak operating time, which is set to 4 hours based on local electricity pricing policies. The peak hour electricity price can be found by consulting the local electricity price list. Off-peak electricity pricing; For round-trip efficiency, the value is 0.90; The daily call duration for auxiliary services is set at 1.5 hours per day, based on the power grid dispatching requirements. Compensation for ancillary services; For the first Average annual available capacity; The annual price for capacity leasing; S33: Set constraints for capacity configuration and establish a set of constraint inequalities: First, initial capacity constraints, ;in Initial capacity configuration; The minimum duration of continuous discharge is set to 2 hours. The first constraint is a safety margin factor, set to 1.2; the second constraint is capacity adequacy, which applies to any point during system operation. ,satisfy ;in For the first The available capacity of the system at any given time is equal to the sum of the remaining capacity of all battery modules multiplied by their health status. For the first Load power at any given time; The load response time is set to 1 hour. 1.1 represents a 10% reserve capacity factor; third, battery life constraints, when any battery module's... When the efficiency drops to 80%, the module must be replaced or the system capacity must be expanded; fourth, economic constraints, the system's investment payback period should not exceed 8 years; S34: Use dynamic programming algorithm to solve the capacity configuration optimization problem, and consider the system lifetime. Divided by year Each decision-making stage, each stage The state variable is the current available capacity of the system. The decision variable is the capacity expansion amount at this stage. The value ranges from 0 to the system's maximum expansion capacity; from the last stage Starting with reverse recursion, the value function in the final stage Equals the net profit of that stage, for stage The value function is calculated based on the Bellman optimality equation, and the formula is as follows: ;in For the first The stage state is The optimal value function at that time; Represents all feasible decision variables Find the maximum value; calculate sequentially. The value function for each stage records the optimal decision at each stage; S35: From the first stage Begin forward backtracking of the optimal decision sequence to determine the initial capacity configuration scheme. And the capacity expansion plan for each stage; for each stage that requires capacity expansion, record the expansion time. and expanded capacity Generate a capacity configuration schedule, with tables containing columns for: year, current capacity, capacity expansion amount, total capacity after expansion, annual investment cost, annual operating income, and annual net cash flow. Use MATLAB to plot a capacity configuration optimization graph, with the horizontal axis representing the year, the left vertical axis representing capacity, and the right vertical axis representing net present value. The graph uses a stepped curve to represent the change in capacity over time, a bar chart to represent the net cash flow for each year, and a cumulative curve to represent the change in net present value. Output the optimization results as an Excel file and a PDF report for decision-makers' reference.

5. The method according to claim 1, characterized in that, Step S4 includes the following specific steps: S41: Read the capacity degradation prediction curve data generated in step S2 for each battery module in the system. Extract its future time points Predicted remaining capacity Calculate the corresponding health status The formula is: ;in For the first The battery module is in the first Constant health status; For the first The battery module is in the first Predicted remaining capacity at any given time; For the first The initial rated capacity of each battery module; Each module Data is stored in time series, and a system is established. A time matrix, where the row indices are module numbers and the column indices are time points; S42: Establish a dynamic adjustment strategy for charge / discharge depth. For battery modules whose health status is gradually deteriorating, limit their charge / discharge depth to slow down further degradation. The battery module is in the first At any given time, calculate its permissible depth of charge and discharge limit. The formula is: ;in For the first The module in the first Limits on the depth of charge and discharge at any given time; The standard depth of charge / discharge for the system design is 85%. The minimum permissible health state threshold is set to 80%; 0.5 is an adjustment index used to control the rate of change of the depth limit; in the Battery Management System (BMS), the calculated... The value is set as the charge / discharge cutoff condition for this module, when the module's state of charge... Exceeding the permitted range When this happens, the BMS stops the charging and discharging operation of the module; S43: Establish a dynamic optimization strategy for power allocation. When there are differences in the health status of each battery module in the system, the strategy is adjusted according to the power distribution of each module. Proportional allocation of charging and discharging power, for the first Each battery module is used to calculate its power distribution coefficient. The formula is: ;in For the first Power allocation coefficient of each battery module; For the first The current health status of each battery module; This represents the total number of battery modules in the system. This represents the summation operator; For summation index; power allocation coefficient This reflects the proportion of total power that the module should bear. Modules with higher health status bear more power, while modules with lower health status bear less power, thereby achieving balanced use of each module. S44: Based on the power allocation coefficient and the total system power demand, calculate the actual charging and discharging power command for each battery module. Each battery module has a charging and discharging power. The calculation formula is: ;in For the first The charging and discharging power of each battery module; The total power command received by the energy storage system is issued by the superior energy management system; The conversion efficiency of the DC / DC power converter is set to 0.95 during charging and 0.96 during discharging; the calculated power command is then used. Send via CAN communication bus to the first The DC / DC converter of each battery module adjusts the duty cycle of the pulse width modulation (PWM) signal according to the power command to control the actual charging and discharging current. S45: Establish a capacity attenuation compensation effectiveness evaluation mechanism, evaluate the implementation effect of the compensation strategy every 7 days, and calculate the system's capacity balance index (CBD) using the following formula: CBD represents the capacity balance, with a value ranging from 0 to 1. The closer it is to 1, the more balanced the capacity of each module is. It is a function of standard deviation; It is a function of average value; arrive This indicates the health status of each battery module. When the CBD value is below 0.90, it indicates a significant difference in capacity between modules, requiring adjustment of the power allocation coefficient to increase the power of modules with lower health status. Reduce its load and increase the load of modules with higher health status. The load is increased, the power command is recalculated and issued to achieve dynamic balance.

6. The method according to claim 1, characterized in that, Step S5 includes the following specific steps: S51: Organize the capacity configuration scheme generated in step S3 and the attenuation compensation strategy generated in step S4 into a control parameter data packet. The data packet is organized in JSON format and contains the following fields: total system capacity. Battery module number Array, power allocation coefficients of each module Array and depth of charge / discharge limits for each module Array, capacity expansion time point Array, capacity expansion capacity value array, threshold Capacity adequacy threshold Perform format verification on the data packets to ensure that the lengths of each array are consistent and the values ​​are within a reasonable range. Generate a checksum for the data packets using the MD5 algorithm and append it to the end of the data packets. S52: Establish a communication connection with the energy storage system controller. The controller uses an embedded industrial computer running a Linux operating system. It connects to the host computer via an Ethernet interface and uses the TCP / IP protocol for data transmission. The host computer starts a Socket client program to initiate a connection request to the controller's IP address and port number. The Socket server program on the controller side receives the connection request and establishes a communication link. After the communication is established, the host computer sends a data packet type identifier 0x01, indicating that the capacity configuration scheme data is about to be transmitted. The controller returns an acknowledgment signal 0xACK. S53: The host computer sends control parameter data packets to the controller through the established communication link, using a packet transmission method. Each data packet does not exceed 1024 bytes, and the header of the data packet contains the packet sequence number, total number of packets, and data length fields. After receiving each data packet, the controller verifies the continuity of the packet sequence number and the correctness of the data length, and returns a reception confirmation signal to the host computer. After all data packets have been transmitted, the controller performs an MD5 check on the received complete data packets, compares the calculated checksum with the checksum in the data packet, and if they match, sends a data reception success signal 0x02 to the host computer. If they do not match, a retransmission request signal 0x03 is sent. The host computer decides whether to retransmit the data packet based on the return signal. S54: After receiving the complete control parameter data packet, the controller parses the JSON data into an internal data structure and starts the real-time monitoring program. The monitoring program collects the current data from the BMS of each battery module every second. The value and remaining capacity value will be the actual Values ​​and data packets A comparison is made when any module's... Below At that time, the controller generates a capacity expansion trigger signal; simultaneously, the monitoring program calculates the total available capacity of the system. The formula is: ;in This represents the total available capacity of the system. This refers to the number of battery modules; For the first The rated capacity of each module; For the first The current health status of each module; when Below load demand When the energy reaches 120% of the required level, the controller also triggers a capacity expansion signal; S55: The controller determines the power allocation factor based on the received power allocation factor. The array calculates the power commands for each battery module in real time, and receives the total power command from the upper-level energy management system. At that time, the controller performs power allocation calculations, for the first... Module calculation ,Will Transmitted via CAN bus at a baud rate of 250kbps to the [unclear - likely a specific location]. Each module's DC / DC converter transmits messages in standard CAN frame format, with a message ID of 0x300+. The data field contains the power command value and the operating mode flag. After receiving the power command, the DC / DC converter adjusts the PWM duty cycle of the switching transistor according to the power command value. The PWM frequency is 20kHz, which realizes precise control of the charging and discharging power. S56: The controller feeds back system operation data to the host computer optimization system every 24 hours. The feedback data includes the actual operation data of each module. The actual charging and discharging power, cumulative throughput, and temperature distribution are received by the host computer optimization system. The system then compares and analyzes the actual operating data with the predicted data to calculate the prediction deviation. The formula is as follows: When the absolute value of the deviation exceeds 10%, the model retraining procedure is triggered, and the LSTM neural network is incrementally trained using the latest running data to update the model parameters. Then, steps S2 to S4 are re-executed to generate the corrected capacity configuration scheme and compensation strategy. Steps S51 to S55 are executed again to send the new scheme to the controller, forming a rolling optimization closed-loop control to achieve adaptive optimization operation of the energy storage system throughout its entire life cycle.

Images