A method for on-line evaluation of state of health of optical storage and charging energy storage battery and life optimization

Abstract

The application discloses a kind of light storage energy storage battery health state online evaluation and life optimization method, comprising: collecting battery operating data and extracting working condition characteristics;Three-layer fusion architecture of extended Kalman filter state of charge estimation layer, recursive least square capacity identification layer with forgetting factor and Gaussian process regression residual correction layer are used, and battery health state SOH is estimated online;Based on real-time SOH and working condition, single-step capacity attenuation ratio is calculated, and equivalent replacement cost is converted, to realize life loss monetization;With the minimum sum of grid power purchase cost and battery loss cost as the optimization goal, introduce the dynamic state of charge boundary based on real-time SOH and charge-discharge rate constraint, build rolling optimization scheduling model;Solve optimal charge-discharge power sequence and closed-loop feedback.The application deeply fuses SOH evaluation and scheduling, improves SOH estimation accuracy, and optimizes system economy and battery life through cost quantification and dynamic safety boundary coordination.

Description

Technical Field

[0001] This invention belongs to the field of new energy and battery management system technology, specifically relating to an online assessment of the health status of energy storage batteries and an optimized scheduling method considering lifespan loss, applied to an integrated photovoltaic-storage-charging system. Background Technology

[0002] With the increasing popularity of electric vehicles, integrated photovoltaic-storage-charging (PV-S-C) power stations have become an important part of urban energy infrastructure. In PV-S-C systems, the energy storage battery, as the core buffer unit, faces frequent and irregular charging and discharging conditions, and its State of Health (SOH) directly affects the system's operational safety and economy.

[0003] In existing technologies, the health status assessment of energy storage batteries and system energy dispatch are usually disconnected, resulting in three main drawbacks: 1. Delayed and Inaccurate Health Status Assessment: The industry commonly uses offline capacity testing or a single ampere-hour integration method, which cannot reflect the battery degradation status online in real time, and it is also difficult to eliminate the cumulative error of ampere-hour integration. Although there are online estimation schemes that combine equivalent circuits with Kalman filtering, pure mechanistic models have inherent modeling biases under complex dynamic conditions, and the accuracy of health status estimation cannot meet the needs of fine-grained scheduling.

[0004] 2. Mismatch between scheduling strategy and actual battery capacity: Current energy management systems often treat batteries as ideal models, setting only fixed state-of-charge constraints and completely ignoring the dynamic limitations of real-time battery health on charging and discharging capabilities. This defect can lead to the system issuing power commands to aging batteries that exceed safety boundaries, causing overcharging and over-discharging problems, creating a vicious cycle where "scheduling commands accelerate battery aging".

[0005] 3. Battery life loss cannot be quantified economically: Existing scheduling objectives often focus solely on short-term peak-valley arbitrage and other economic benefits, treating battery life loss as an unquantifiable external cost. Due to the lack of a means to quantify physical degradation into economic costs, scheduling strategies cannot scientifically balance operational benefits with the remaining value of battery life, ultimately increasing the total operating cost over the system's entire lifecycle.

[0006] Therefore, the industry urgently needs a technical method that can accurately assess battery health status online and feed the assessment results back to the scheduling strategy in real time in the form of economic quantification, so as to achieve synergistic optimization of system operating efficiency and battery life. Summary of the Invention

[0007] I. Technical problems to be solved This invention aims to solve the core technical problems of low accuracy and strong lag in battery health status assessment in existing photovoltaic-storage-charging systems, and the disconnect between SOH assessment results and energy dispatch strategies, which leads to dispatch commands exceeding the safe operating boundaries of aging batteries, accelerating battery degradation, and resulting in high operating costs throughout the system's life cycle.

[0008] II. Technical Solution To address the aforementioned technical challenges, this invention proposes an online health status assessment and lifetime optimization method for photovoltaic-energy storage batteries. The overall concept of this method is to construct a complete technical closed loop encompassing "state perception - degradation effect quantification - scheduling decision optimization": the real-time health status of the battery is perceived online through a model that fuses mechanisms and data; the physical degradation effect of the battery is converted into an equivalent monetized cost; and economic scheduling decisions are constrained by the monetized loss cost and a dynamic safety boundary based on state of charge (SOH). The online health status assessment stage employs a three-layer progressive fusion architecture: an EKF state of charge estimation layer, an RLS capacity identification layer, and a GPR residual correction layer. This architecture ensures the interpretability of the mechanism while improving estimation accuracy under complex operating conditions.

[0009] This method includes the following five execution steps: Step S1: Data Acquisition and Feature Extraction This step provides basic data support for the entire process, completing the collection of raw operational data, calculation of basic physical quantities, and data-driven feature extraction.

[0010] With a fixed sampling period Real-time acquisition of operating data of the energy storage battery during the operation of the photovoltaic-energy storage-charging system, including battery terminal voltage. Loop current Battery body temperature The symbols for current are uniformly defined: discharge current is positive and charging current is negative.

[0011] Calculate the instantaneous power of the battery based on the collected voltage and current data:

[0012] Based on the collected time-series operational data, feature vectors required for the data-driven model are extracted. ; Eigenvector The constituent elements include at least two types of operating condition features: the root mean square value of current and the average value of battery temperature, which provide input features for subsequent Gaussian process regression residual correction.

[0013] Step S2: Online estimation of health status based on fusion model This step is the core state perception stage, which adopts a three-layer fusion estimation architecture of "EKF state of charge estimation layer + RLS capacity identification layer + GPR residual correction layer". Based on the first-order Thevenin equivalent circuit, it achieves high-precision online SOH estimation through three-layer progressive calculation.

[0014] The execution sequence follows a recursive iterative logic: at time k, the capacity estimate obtained from the RLS identification at the previous time is first called and substituted into the EKF state equation, and then the optimal estimate of the state of charge at the current time is output after extended Kalman filtering. Then The temporal difference is used as the observation value and fed into the RLS capacity identification layer to update the capacity estimate at the current time, eliminating the logical black box caused by the unclear parameter calling order; finally, GPR residual compensation is superimposed to output a high-precision real-time battery health state estimate. This process is implemented in four steps: S2.1 Establishing a first-order Thevenin equivalent circuit mechanism model Step S2.1 provides a unified mechanistic model foundation for the subsequent three-layer estimation. The first-order Thevenin equivalent circuit of the battery is constructed, and the discretized state equations and output equations are as follows:

[0015] in: : Battery state of charge at time k; : The current maximum available battery capacity at time k; during online recursive execution, the EKF state prediction at time k uses the capacity estimate obtained from the RLS identification at the previous time. Substitute into the calculation; Battery polarization voltage at time k; : Battery internal resistance in ohms; Battery polarization internal resistance; Battery polarization capacitor; The nonlinear mapping function between open-circuit voltage and state of charge can be achieved by table lookup or polynomial fitting. : Battery charge / discharge coulombic efficiency.

[0016] S2.2 Charge State Closed-Loop Correction Based on Extended Kalman Filter (EKF) (Corresponding to EKF Charge State Estimation Layer) Using the difference between the measured battery terminal voltage and the terminal voltage output by the equivalent circuit model as the observation error, the extended Kalman filter algorithm is used to perform real-time closed-loop iterative correction on the state-of-charge time series values ​​to obtain the optimal state of charge. This provides high-precision observational input for subsequent capacity identification.

[0017] S2.3 Available capacity identification based on recursive least squares (RLS) with forgetting factor (corresponding to RLS capacity identification layer) Step S2.3 Based on the SOC timing results output by EKF, identify the current maximum usable capacity of the battery online. The estimated SOH value output by the mechanistic model is obtained.

[0018] Starting from the SOC ampere-hour integral recursive relationship in S2.1, the target identification quantity is... Linearization parameter reconstruction: Since the capacity is located in the denominator of the SOC recursive formula, it cannot be solved directly using least squares. Therefore, let... This transforms the nonlinear identification problem into a standard linear regression problem.

[0019] By rearranging the terms of the SOC recursive formula, we can obtain:

[0020] Reconstructed into the standard form of linear regression:

[0021] The above equation is the standard matrix form of linear regression, and the physical meanings of each variable in the equation are as follows: Observations That is, the time-series change in the state of charge output by the EKF; Regression vector ; , which is the reconstructed parameter to be identified, and is the reciprocal of the available battery capacity.

[0022] The solution is obtained iteratively using a recursive least squares algorithm with a forgetting factor, and the iterative formula is as follows:

[0023] in: Forgetting factor, with a value range of: This is used to assign higher weight to recent operating data, thereby improving the algorithm's ability to track slow battery capacity drift. : The recursive gain matrix at time k; : The covariance matrix of parameter estimation error at time k; : Identity matrix.

[0024] The parameter estimates at time k are obtained by iterative solution. Then, the estimated available battery capacity at time k is obtained. The health status estimate obtained solely from the mechanistic model is calculated based on the capacity ratio:

[0025] in: This refers to the rated capacity of a single battery cell / battery pack.

[0026] S2.4 SOH Residual Compensation Correction Based on Gaussian Process Regression (GPR) (Corresponding GPR Residual Correction Layer) Pure mechanistic models are inherently prone to modeling errors due to fluctuations in operating conditions and parameter drift. Therefore, this layer uses a data-driven approach to adaptively compensate for the estimation residuals of mechanistic SOH, thereby further improving the estimation accuracy.

[0027] First, define the output of the mechanism model. Residuals from the actual battery health status:

[0028] in: The residual represents the actual health state of the battery at time k, and is essentially the overall estimation deviation of the mechanism model under the current operating conditions.

[0029] The working condition feature vector extracted in step S1 As input, the pre-trained Gaussian process regression model is used to predict the residual estimate at the current time step:

[0030] By superimposing the mechanism SOH estimate with the GPR prediction residual, a high-precision real-time health status final estimate after correction using fused operating condition data is obtained:

[0031] The output of this step It serves as the sole real-time input parameter for subsequent lifetime loss quantification and dynamic scheduling constraints.

[0032] Step S3: Establish a model for quantifying and monetizing battery life loss. Step S3 provides a lifetime loss prediction cost for the scheduling optimization stage, complementing the online SOH identification function in step S2: Step S2 identifies the actual health status of the battery online based on measured operating data, providing real-time boundaries for dynamic safety constraints; Step S3 predicts the capacity decay and equivalent cost caused by single-step operation based on aging mechanisms, providing loss cost input for optimizing the objective function.

[0033] Based on the real-time output of step S2 By establishing a quantitative mapping model between the physical degradation of the battery and its equivalent economic cost in accordance with the real-time operating conditions of the battery, the battery aging loss, which cannot be directly quantified, is transformed into an economic indicator that can be calculated in conjunction with the cost of purchasing electricity from the grid.

[0034] S3.1 Calculation of single-step charge / discharge capacity decay ratio The battery energy throughput within a single sampling period is defined. To avoid the physical failure of the exponentiation operation of the degradation model due to the negative power sign during charging, the absolute value of the instantaneous power is uniformly used to calculate the total throughput energy. Both charging and discharging aging losses are accumulated in a positive direction.

[0035] The rated energy of a battery is calculated from its rated capacity and nominal voltage:

[0036] In the formula This refers to the battery's nominal voltage.

[0037] A formula for calculating the single-step capacity decay ratio is established based on energy throughput, depth of discharge, and temperature aging effects. in: , : Aging model fitting coefficients, which can be obtained by combining battery accelerated aging tests with least squares fitting calibration; : The maximum cumulative discharge depth corresponding to the most recent complete charge-discharge cycle up to the current k steps; its recursive rule is: based on the most recent full charge state, the cumulative discharge depth increases positively as the SOC decreases during the discharge process, and decreases correspondingly as the SOC increases during the charging process. The baseline value is updated after a single complete charge-discharge cycle; during the rolling optimization prediction period, the cumulative discharge depth at the corresponding time can be updated synchronously following the SOC prediction trajectory. Activation energy of battery electrochemical reaction; Universal gas constant; Reference temperature for battery aging model calibration; : Measured battery temperature at time k.

[0038] S3.2 Monetary Conversion of Physical Quantities of Lifespan Loss Converting the single-step capacity decay rate into an equivalent replacement economic cost monetizes aging losses:

[0039] in: Total cost of replacing the entire energy storage battery pack; : Battery health status threshold for battery retirement; It will be included as a cost item in the scheduling optimization objective function, alongside the cost of electricity purchase.

[0040] Step S4: Construct a single-objective rolling optimization scheduling model that considers lifetime loss. This step is based on the Model Predictive Control (MPC) framework, in conjunction with the data sampling period. Consistent scheduling cycles, building future continuity A rolling optimization model with several time steps (rolling optimization prediction time-domain step size); the optimization objective is to minimize the total system operating cost (single optimization objective, including two types of cost items: grid power purchase cost and battery life equivalent loss cost), while also introducing real-time... The dynamic safety operation constraints eliminate the shortcomings of traditional fixed constraints that do not match the battery aging state.

[0041] S4.1 Optimize the objective function With the goal of minimizing the total cumulative operating cost over the next H steps, no additional manual weighting coefficients are required. The electricity purchase cost and battery depreciation cost are measured in the same dimensions and can be directly added together for comparison.

[0042] in: Time-of-use electricity price at time k+j; The power exchanged between the energy storage system and the grid at time k+j is positive for electricity purchase and negative for electricity sales. : Equivalent loss cost of battery life at time k+j.

[0043] All constraints of model S4.2 1. System power balance constraints

[0044] Photovoltaic power output; This refers to the energy storage charging and discharging power (discharging is positive, charging is negative). This represents the total load power of the charging piles within the station.

[0045] 2. Dynamic state-of-charge boundary constraints based on real-time state of charge (SOH) As the battery ages, the SOC window can be shrunk accordingly to avoid overcharging and over-discharging risks in aging batteries. The quantitative constraint formula is as follows:

[0046] in: , : Fixed upper and lower limits of state of charge for brand new batteries at the factory; , : Dynamic narrowing coefficient of the charge interval, characterizing the extent to which the SOH decay shrinks the available SOC window.

[0047] 3. Dynamic charge / discharge rate constraint based on real-time SOH As battery health deteriorates, the maximum allowable charge and discharge power of the battery is linearly reduced to prevent accelerated degradation of aging batteries under high current conditions.

[0048] in: Rated charge and discharge power of the energy storage converter; , The power derating linear fitting parameters are the slope and intercept terms, which are used to characterize the linear influence of SOH decay on the maximum charge and discharge power.

[0049] Step S5: Output the optimal scheduling strategy and form a complete closed-loop feedback. The rolling optimization model established in step S4 is numerically solved to obtain the optimal charging and discharging power control sequence for energy storage in the next H steps. The system issues this power command to control the energy storage converter (PCS) to perform charging and discharging operations.

[0050] After a single scheduling cycle is completed, the system collects the new voltage, current, and temperature time-series data of the battery after this scheduling operation in real time, and directly transmits them back to the data acquisition stage in step S1. This restarts a new round of the entire process of "data acquisition → fusion of SOH online assessment → monetization and quantification of life loss → optimized scheduling with dynamic safety constraints", forming a continuous and self-iterative closed-loop control link, and continuously adjusting the energy storage operation strategy according to the real-time aging status of the battery.

[0051] III. Beneficial Effects This invention constructs a complete technical closed loop of "online SOH assessment of fusion model → monetization conversion of battery physical degradation → economic scheduling based on SOH dynamic boundary". There is a strong progressive dependency between each step, which solves the defect of SOH assessment and scheduling strategy being separated in the prior art.

[0052] This invention constructs a layered fusion estimation architecture consisting of an "EKF state of charge estimation layer + RLS capacity identification layer + GPR residual correction layer". It relies on the mechanistic model to ensure physical interpretability and uses data-driven GPR to perform condition-adaptive compensation for the residuals of the mechanistic model. This overcomes the shortcomings of pure mechanistic model estimation drift and the poor generalization of pure data-driven models, and provides high-precision input for subsequent stages.

[0053] This invention transforms the abstract physical degradation of battery capacity into an equivalent replacement cost consistent with the power grid purchase cost, and directly superimposes these costs to form the scheduling objective function. The scheduling model can automatically balance the arbitrage benefits of discharge with the cost of battery aging, making the optimal economic decision over the entire life cycle. This solves the problem of traditional scheduling that only pursues short-term gains and excessively consumes battery life, and requires no manual weight adjustment.

[0054] Based on the real-time updated SOH estimate, this invention synchronously and dynamically shrinks the available SOC range and linearly reduces the maximum allowable charge and discharge power. From the scheduling command level, it actively avoids dangerous operating ranges that are prone to aggravating battery degradation, solves the problem that traditional fixed constraints cannot adapt to the continuous aging of batteries, and significantly improves the long-term operating safety of the system. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating the overall execution process of the method of the present invention. Figure 2 This is a logical structure diagram of the SOH online estimation hierarchical fusion module of the present invention; Figure 3 This is a schematic diagram of the rolling optimization scheduling closed-loop control principle of the present invention, which takes into account battery life loss. Detailed Implementation

[0056] This implementation method relies on the physical system of an integrated photovoltaic, energy storage and charging power station to complete the entire process. The entire system includes a photovoltaic array, an energy storage battery pack (lithium iron phosphate energy storage battery), an energy storage converter PCS, a charging pile load, a grid access unit, a battery management system (BMS), an edge controller, and a data acquisition module. The edge controller has a built-in program that implements the complete method logic of this invention.

[0057] This embodiment uses a lithium iron phosphate energy storage battery pack with a rated capacity of [missing information]. nominal voltage Total cost of full battery pack replacement Yuan, battery scrapping threshold System unified sampling and scheduling cycle Rolling optimization prediction time domain Step (corresponding to a 2-hour scheduling window); the basic parameters of the battery aging model, the Thevenin equivalent circuit parameters, and the parameters of the EKF, RLS, and GPR models are all calibrated offline through battery accelerated aging tests and static OCV calibration tests, and are pre-stored in the edge controller, relying only on real-time running data for online iterative calculation.

[0058] I. Implementation Step S1: Data Acquisition and Feature Extraction 1. Real-time acquisition of raw operating data Edge controller with Real-time operating data of the energy storage battery is collected and uploaded to the BMS at fixed intervals: battery terminal voltage. Loop current Average temperature of the battery body Strictly adhere to the current sign rule: assign positive values ​​to battery discharge current and negative values ​​to charging current. Synchronously collect power data from the entire station: real-time photovoltaic output. Total load power of charging piles Power grid interaction This provides basic data for subsequent power balance constraints.

[0059] 2. Instantaneous power calculation The controller's built-in computing module calculates the battery's instantaneous power at time k in real time. This power synchronization is used for subsequent single-step energy throughput. calculate.

[0060] 3. Extraction of working condition feature vectors Based on continuous time-series voltage, current, and temperature data, feature vectors are extracted for each cycle. The vector contains two core operating condition features: the root mean square value of the current and the average battery temperature within the current window. These features are used as input features for the pre-trained GPR model and stored in the controller cache, waiting to be called by the S2.4 residual correction stage.

[0061] II. Implementation Step S2: Online Health Status Estimation Based on a Three-Layer Fusion Model This step uses the first-order Thevenin equivalent circuit as a unified mechanism basis, and follows a time-series progressive iteration from the EKF state-of-charge estimation layer to the RLS capacity identification layer to the GPR residual correction layer, eliminating logic black boxes between each layer and achieving high precision in a layered and step-by-step manner. Online output.

[0062] Discretization of the first-order Thevenin equivalent circuit mechanism model in S2.1 The controller's discrete state equations and output equations are fixed, and offline calibration parameters are used: ohmic internal resistance. Polarization internal resistance Polarized capacitors Charge and discharge coulombic efficiency Open circuit voltage A polynomial fitting function is embedded into the program, and the corresponding open-circuit voltage is output upon inputting the State of Charge (SOC). During each iterative calculation, the EKF state prediction stage consistently calls the RLS (Recognition and Scale) function from the previous time step to identify the output capacity. Substituting into the SOC recursive formula, the complete discrete equation system is as follows: S2.2 Charge State Closed-Loop Correction Based on Extended Kalman Filter (EKF) 1. The difference between the output voltage of the Thevenin model and the measured voltage of the BMS is used as the observation error; 2. Run the extended Kalman filter iterative process to perform real-time closed-loop correction of the SOC time series value; 3. Output the optimal estimate It is stored in the cache as the sole observation input for capacity identification of the lower-level RLS.

[0063] S2.3 Available Capacity Identification Based on Recursive Least Squares (RLS) with Forgetting Factor 1. Linear parameter reconstruction: Addressing the capacity-related aspects of the denominator in the SOC recursive formula. The resulting nonlinearity issues make the parameters to be identified... This transforms the nonlinear identification into a standard linear regression form. 2. Constructing Observations and Regression Vectors: Observations Regression vector ; 3. Solidify the RLS iterative formula, and take the forgetting factor as... Give higher weight to recent operating conditions and update the recursive gain in real time. Error covariance matrix Parameters to be identified ; 4. Capacity conversion and mechanism SOH calculation: from Given the current available capacity, calculate the pure mechanistic health status using the capacity ratio:

[0064] Output Transferred to the GPR correction layer.

[0065] S2.4 SOH Residual Compensation Correction Based on Gaussian Process Regression (GPR) 1. Residual definition: The estimation bias of the mechanistic model. ; 2. Model Invocation: The working condition feature vector extracted by S1 is used... Input an offline pre-trained Gaussian process regression model and output the predicted residual values. ; 3. Fusion Correction Calculation: By superimposing the estimated value of the mechanism and the predicted residual, the final high-precision real-time health status is output.

[0066] Should As a real-time input parameter for S3 lifetime loss quantification and S4 dynamic scheduling constraints, it is updated every 5 minutes.

[0067] III. Implementation Step S3: Establish a quantitative and monetization model for battery life loss. S3.1 Real-time calculation of single-step capacity decay ratio 1. Single-cycle energy throughput calculation: To avoid the failure of the degradation model due to negative charging power, the absolute value of the power is used for calculation:

[0068] 2. Rated energy is pre-calculated offline and stored in the controller: ; 3. Cumulative depth of discharge Real-time recursion: Based on the battery's fully charged state, the depth of discharge is accumulated synchronously as the State of Charge (SOC) decreases during the discharge process, and deducted during the charging process. The baseline value is updated after a single complete charge-discharge cycle. Rolling optimization updates the prediction at each moment in the prediction time domain, following the SOC prediction trajectory. ; 4. Single-step attenuation ratio calculation: Controller-based attenuation formula, aging coefficient ,activation energy Gas constant Reference temperature Calibrated by accelerated aging test, with real-time temperature substituted. Complete the attenuation ratio calculation:

[0069] S3.2 Implementation of Monetization Conversion of Life Loss Convert the physical attenuation into an equivalent loss cost with dimensions consistent with electricity prices, and solidify the conversion formula:

[0070] The equivalent replacement cost for battery aging in a single cycle is directly fed into the S4 scheduling objective function to participate in cost optimization, without the need for manual setting of weight coefficients, thus achieving equivalent accounting of electricity purchase cost and battery loss cost.

[0071] IV. Implementation Step S4: Construct a single-objective rolling optimization scheduling model that considers lifetime loss. Based on the Model Predictive Control (MPC) framework, For the scheduling period, predict the time domain. In the next step, the edge controller calls the linear programming solver to complete the rolling optimization solution.

[0072] 1. Optimize the implementation of the objective function The sole optimization objective is to minimize the total operating cost over the next H steps. The objective function has no additional weight terms, and the two types of costs are directly summed.

[0073] Time-of-use electricity price in the formula Send it to the controller one day in advance. The equivalent loss cost at each future time point is predicted based on the S3 attenuation model.

[0074] 2. All constraints are implemented in a fixed and standardized manner. (1) System power balance constraints

[0075] Energy storage power Discharging is positive and charging is negative, consistent with the sign rules for current.

[0076] (2) Based on the dynamic SOC boundary constraint of real-time SOH, the lower limit of the new battery is determined. Offline calibration, narrowing factor Fixed, real-time reading per cycle Dynamically shrink the available SOC window:

[0077] The higher the degree of aging, the narrower the SOC operating range, thus preventing overcharging and over-discharging.

[0078] (3) Rated power of energy storage converter constrained by dynamic charge / discharge rate based on real-time SOH Linear depreciation parameters Pre-stored, the maximum allowable energy storage capacity is linearly adjusted based on real-time SOH:

[0079] Aging batteries automatically limit high-current charging and discharging to slow down the rate of degradation.

[0080] V. Implementation Step S5: Output the optimal scheduling strategy and form a complete closed-loop feedback. 1. Solve the S4 rolling optimization model and output the optimal charging and discharging power control sequence for energy storage in the next H steps. ; 2. The edge controller sends the optimal power command for the first time step to the PCS to control the energy storage battery to perform charging and discharging actions; 3. Single scheduling cycle After the execution is completed, the BMS re-collects the battery's new voltage, current, and temperature timing data and sends it back to the S1 data acquisition stage. 4. Automatically restart a new round of the entire process of "data acquisition → three-layer fusion SOH online assessment → life loss monetization and quantification → rolling optimization scheduling with dynamic safety constraints", forming an uninterrupted self-iterative closed-loop control link, and continuously adjusting the energy storage operation strategy according to the battery aging status.

[0081] VI. Verification of Implementation Results A comparative experiment was conducted at the same photovoltaic-storage-charging station using the complete implementation method described above. The control group adopted the traditional scheme (offline capacity calibration SOH, fixed SOC and power constraints, scheduling only optimized for peak-valley arbitrage benefits), while the scheme of this invention served as the experimental group. It was run continuously for 6 months to verify the four core inventive technological advantages: 1. SOH Assessment and Scheduling: End-to-End Closed-Loop Collaborative Effect The control group underwent a 7-day offline SOH test, and the evaluation results could not be fed into the scheduling system in real time, as the scheduling system always treated the battery as an ideal model; this implementation method updates every 5 minutes. The system achieves real-time synchronization with the strong coupling of loss calculation and scheduling constraints, completely resolving the problem of the separation between evaluation and scheduling. No over-limit charging or discharging commands for aged batteries were issued during the test.

[0082] 2. Improved SOH estimation accuracy of the three-layer fusion architecture The control group used a single RLS mechanism for estimation, with a maximum SOH estimation error of 4.2% under peak charge and discharge conditions. This implementation relies on a three-layer progressive fusion architecture of EKF-RLS-GPR, and uses GPR to complete adaptive compensation of operating condition residuals. Under the same operating conditions, the maximum SOH estimation error is only 0.9%. It has the advantages of mechanism interpretability and data-driven high precision, providing reliable input for loss quantification and dynamic constraints.

[0083] 3. Optimizing economic benefits through monetization of lifespan depreciation. The control group's scheduling only pursued the lowest short-term electricity purchase cost and frequently engaged in deep charge-discharge arbitrage, resulting in a 2.7% battery capacity degradation over 6 months. This solution converts aging losses into equivalent replacement costs for unified optimization, and the controller automatically balances arbitrage gains with battery losses. During the same period, the battery capacity degradation was only 1.1%, and the estimated total operating cost of the system over its entire life cycle decreased by 16.3%. No manual adjustment of weights was required, and the economic decision-making was objective and unbiased.

[0084] 4. SOH dynamic safety boundary delays aging effect The control group used a fixed SOC window and a fixed power limit throughout the process, and still performed 1C high-power charging and discharging when the battery SOH dropped to 0.88. This implementation method shrinks the SOC range synchronously with the SOH decay and linearly reduces the charging and discharging power. The lower the SOH, the stricter the operating constraints, actively avoiding dangerous conditions that accelerate aging, significantly improving the long-term operating safety of the battery, and effectively extending the replacement cycle of the energy storage battery.

[0085] The entire solution of this invention can be implemented using existing BMS, PCS, and edge computing hardware, without the need for additional dedicated testing equipment, resulting in low implementation costs and strong adaptability.

Claims

1. A method for online assessment of the health status and lifespan optimization of photovoltaic-energy storage batteries, characterized in that, Includes the following steps: Step S1: Using a fixed sampling period Real-time acquisition of operating data of the energy storage battery, including at least battery terminal voltage, circuit current and battery body temperature; calculation of instantaneous battery power based on the acquired voltage and current data; Extracting feature vectors required for data-driven models based on time-series operational data; Step S2: Based on a fusion of mechanistic models and data-driven approaches, the health status of the energy storage battery is estimated online, and the real-time health status estimate is output. ; Step S3: Based on the real-time health status estimate output in step S2 Based on the real-time operating conditions of the battery, a quantitative mapping model is established from the degree of physical degradation of the battery to the equivalent economic cost, and the equivalent loss cost of a single-step battery life is output. ; Step S4: Based on the model predictive control framework, in conjunction with the data sampling period Consistent scheduling cycles, building future continuity A rolling optimization model with time steps is used, with the optimization objective being to minimize the total system operating cost, which includes the grid power purchase cost and the equivalent battery life loss cost. Constraints include at least system power balance constraints and constraints based on the real-time health status estimate. Dynamic state of charge boundary constraints and based on the real-time health state estimate Dynamic charge / discharge rate constraints; Step S5: Numerically solve the rolling optimization model established in Step S4 to obtain the future... The optimal charging and discharging power control sequence for energy storage is obtained, and the power command is issued to control the energy storage converter to perform charging and discharging actions. After the execution of a single scheduling cycle is completed, the battery operation data is collected again and sent back to step S1 to form an uninterrupted closed-loop control link.

2. The method for online health status assessment and lifespan optimization of photovoltaic-energy storage batteries according to claim 1, characterized in that, In step S2, a three-layer progressive fusion architecture of "EKF state of charge estimation layer + RLS capacity identification layer + GPR residual correction layer" is adopted for online health state estimation. Specifically, this includes: establishing a first-order Thevenin equivalent circuit mechanism model, and performing closed-loop correction of the state of charge based on the extended Kalman filter algorithm to obtain the optimal estimated value of the state of charge. Based on the recursive least squares algorithm with a forgetting factor, The time difference is used as the observed value to identify the current maximum usable capacity of the battery online, and the health state estimate output by the mechanism model is obtained. Using the feature vector as input, the pre-trained Gaussian process regression model is invoked to estimate the residual. ,Will and The data is superimposed and output as the final estimated health status after correction of the fused operating condition data. .

3. The method for online health status assessment and lifespan optimization of photovoltaic-energy storage batteries according to claim 2, characterized in that, In the RLS capacity identification layer, starting from the SOC ampere-hour integral recursive relationship of the first-order Thevenin equivalent circuit model, let the parameters to be identified... The nonlinear identification problem is transformed into a standard linear regression problem; observations are constructed. Regression vector ,in To improve the battery's charge-discharge coulombic efficiency; a forgetting factor is employed. The parameter estimates are obtained by iteratively solving the problem using the recursive least squares algorithm. The estimated usable battery capacity is obtained by conversion. Based on capacity ratio Computer model health status estimate , This refers to the battery's rated capacity.

4. The method for online health status assessment and lifespan optimization of photovoltaic-energy storage batteries according to claim 2, characterized in that, The feature vector includes at least two types of operating condition features: root mean square current value and average battery temperature value; in the GPR residual correction layer, the residual estimated by the mechanism model is defined as... ,in For batteries Based on the real-time health status, the GPR model uses the aforementioned feature vector as input and outputs residual estimates. The final health status estimate is .

5. The method for online health status assessment and lifespan optimization of photovoltaic-energy storage batteries according to claim 1, characterized in that, In step S3, the formula for calculating the single-step capacity decay ratio is: in For single-cycle energy throughput, This refers to the instantaneous power of the battery. For the battery's rated energy, For the present The maximum cumulative depth of discharge corresponding to the most recent complete charge-discharge cycle. , These are the fitting coefficients for the aging model. The activation energy of the battery's electrochemical reaction. This is the universal gas constant. To calibrate the reference temperature for the battery aging model, for Real-time measured battery temperature; The formula for monetizing lifetime loss is: in The total cost of fully replacing the energy storage battery pack. This is the health status threshold for batteries that are no longer suitable for disposal.

6. The method for online health status assessment and lifespan optimization of photovoltaic-energy storage batteries according to claim 1, characterized in that, In step S4, based on the real-time health status estimate The dynamic charge state boundary constraints are: in , These are the fixed upper and lower limits of the state of charge for brand new batteries when they leave the factory. , is the dynamic narrowing coefficient of the charge interval, which characterizes the extent to which the SOH decay shrinks the available SOC window.

7. The method for online health status assessment and lifespan optimization of photovoltaic-energy storage batteries according to claim 1, characterized in that, In step S4, based on the real-time health status estimate The dynamic charge / discharge rate constraint is: in For energy storage charging and discharging power, The rated charge and discharge power of the energy storage converter, , The parameters are the linear fitting parameters for power derating, and the terms are the slope and intercept, respectively, used to characterize the linear influence of SOH decay on the maximum charge and discharge power.

8. The method for online health status assessment and lifespan optimization of photovoltaic-energy storage batteries according to claim 5, characterized in that, The cumulative discharge depth The recursive rule is as follows: taking the most recent full charge state as the baseline, the discharge process accumulates the discharge depth positively as the SOC decreases, and the charging process deducts the corresponding depth as the SOC increases. The baseline value is updated after a single complete charge and discharge cycle. During the rolling optimization prediction period, the cumulative discharge depth at the corresponding time is updated synchronously following the SOC prediction trajectory.

9. The method for online health status assessment and lifespan optimization of photovoltaic-energy storage batteries according to claim 1, characterized in that, The objective function for optimization in step S4 is: in For the first Time-of-use electricity pricing for the power grid For the first The power exchange between the energy storage system and the grid is real-time; purchasing electricity is a positive value, and selling electricity is a negative value. For the first The equivalent cost of battery life loss at any given moment.

10. A system for online health status assessment and lifespan optimization of photovoltaic-energy storage and charging batteries, characterized in that, The system is configured to perform the method as described in any one of claims 1 to 9, the system comprising: The data acquisition module is used to perform data acquisition at a fixed sampling period. Real-time acquisition of energy storage battery operation data and total station power data, wherein the operation data includes at least the battery terminal voltage. Loop current and battery body temperature The total power data of the station includes at least the real-time output of photovoltaic power. Total load power of charging piles Power interaction with the power grid ; The online health status estimation module is used to perform the fusion estimation in step S2 of claim 1 and output a real-time health status estimate. ; The lifespan loss quantization module is used to perform the quantization mapping in step S3 of claim 1 and output the equivalent loss cost of battery life in a single step. ; The rolling optimization scheduling module is used to perform rolling optimization model solving based on the model predictive control framework in step S4 of claim 1, and output the optimal charging and discharging power control sequence for energy storage. The closed-loop control module is used to send power commands to the energy storage converter to perform charging and discharging actions, and to send the battery operation data collected after each scheduling cycle back to the data acquisition module, forming an uninterrupted closed-loop control link. The system is implemented based on the existing battery management system, energy storage converter, and edge controller hardware of the photovoltaic-storage-charging integrated power station.

Images