Energy storage system optimization control method integrating aging evaluation and photovoltaic absorption

Abstract

The invention discloses an energy storage system optimization control method fusing aging evaluation and photovoltaic consumption, and the method comprises the steps: firstly building a load prediction model and a photovoltaic output prediction model through employing a long-short-term memory network, and outputting a load prediction value and a photovoltaic output prediction value in a future time period; then establishing a semi-empirical battery aging evaluation model, and further quantifying the aging cost in the charging and discharging process; establishing an objective function and setting constraint conditions by taking peak clipping and valley filling income, photovoltaic consumption income and energy storage aging cost as optimization objectives; the multi-objective optimization decision model is solved based on a particle swarm optimization algorithm, and an optimal energy storage charging and discharging strategy is obtained; and finally, controlling the operation of the energy storage system according to the optimal energy storage charging and discharging strategy. According to the method, through dynamic fusion of photovoltaic output prediction, battery aging evaluation and demand constraint optimization, collaborative optimization of photovoltaic consumption, peak load shifting and demand control is realized.

Description

Technical Field

[0001] This invention relates to the field of power energy storage technology, specifically an optimized control method for energy storage systems that integrates aging assessment and photovoltaic absorption. Background Technology

[0002] Against the backdrop of the accelerating global energy transition, the power system is undergoing unprecedented changes. With the large-scale integration of renewable energy sources such as photovoltaics and wind power, especially the explosive growth of distributed photovoltaics, the operating characteristics of the power system have undergone profound changes, exhibiting "dual high" characteristics (high proportion of renewable energy and high proportion of power electronic equipment). Photovoltaic output is significantly intermittent and fluctuating, and due to grid capacity limitations, grid connection policy constraints, or insufficient absorption capacity, a large number of distributed photovoltaics cannot be connected to the grid in a timely manner, making "difficulty in photovoltaic absorption" a core pain point in the industry—if excess photovoltaic power cannot be absorbed locally in a timely manner, it will be directly wasted, resulting in low resource utilization.

[0003] This trend has brought unprecedented development opportunities to the energy storage sector, making energy storage systems a key hub connecting renewable energy sources such as photovoltaics with user-side electricity demand. Under the two-part tariff mechanism, energy storage systems face multiple complex challenges when implementing peak-shaving and valley-filling strategies to obtain electricity price differences: On the one hand, the random fluctuations and uncertainties of user loads may lead to frequent charging and discharging of energy storage systems, thereby triggering maximum demand exceedances and generating additional demand charges; on the other hand, during frequent charging and discharging, battery aging issues become increasingly prominent, which not only affects the long-term economics of the system but may also reduce system reliability. More importantly, energy storage systems also need to undertake the task of "real-time absorption" of photovoltaic power. If excess power cannot be absorbed in time during peak photovoltaic output periods, a large amount of photovoltaic power will be wasted, further reducing the overall economic efficiency of the system.

[0004] Therefore, there is an urgent need to develop an energy storage system optimization control method that integrates aging assessment and photovoltaic consumption, and an intelligent optimization control method that can deeply integrate photovoltaic consumption, demand control and battery aging assessment. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and propose an energy storage system optimization control method that integrates aging assessment and photovoltaic consumption. By dynamically integrating photovoltaic output prediction, battery aging assessment and demand constraint optimization, it achieves synergistic optimization of photovoltaic consumption, peak shaving and valley filling and demand control.

[0006] To achieve the above objectives, the technical solution specifically adopted by the present invention is as follows:

[0007] An optimization control method for an energy storage system that integrates aging assessment and photovoltaic power consumption includes the following steps:

[0008] Step 1: Establish a load forecasting model and a photovoltaic output forecasting model using a long short-term memory network. Use historical load data, ambient temperature, and date type as inputs to the load forecasting model and output the load forecast for future periods. Use historical photovoltaic output data, irradiance, and cloud cover data as inputs to the photovoltaic output forecasting model and output the photovoltaic output forecast for future periods.

[0009] Step 2: Based on the real-time data of state of charge, charge / discharge rate and battery temperature collected by the battery management system, establish a semi-empirical battery aging assessment model to quantify the aging cost during the charge and discharge process.

[0010] Step 3: Construct a multi-objective optimization decision model: With peak shaving and valley filling revenue, photovoltaic consumption revenue and energy storage aging cost as optimization objectives, and with energy storage charging and discharging power constraints, power constraints, charging and discharging state constraints, demand constraints and anti-reverse current constraints as constraints, establish an objective function;

[0011] Step 4: Solve the multi-objective optimization decision model based on the particle swarm optimization algorithm to obtain the optimal energy storage charging and discharging strategy;

[0012] Step 5: Control the operation of the energy storage system according to the optimal energy storage charging and discharging strategy.

[0013] Preferably, the load forecasting model and the photovoltaic output forecasting model introduce a Dropout layer to prevent overfitting; in the photovoltaic output forecasting model, when the photovoltaic output forecast value in the future period exceeds the load forecast value, the energy storage system will participate in the absorption first. If it exceeds its absorption capacity, the excess part will be marked as quantifiable waste and included in the optimization target.

[0014] Preferably, in step 2, the collected electrical state, charge / discharge rate, and battery temperature data need to be preprocessed by outlier removal and linear interpolation.

[0015] As a preferred option, the semi-empirical battery aging assessment model is as follows:

[0016] Q loss (p,Ah)=σ funct (p)·Ah Z ;

[0017] Where Z is the power-law exponent characterizing Ah throughput dependence, σ funct (p) is a nonlinear function of the severity factor;

[0018]

[0019] Where α and β are SOC correction coefficients, Ea is activation energy, η is current coefficient, and I CLet θ be the charging / discharging current, θ be the battery temperature, and Rg be the gas constant.

[0020] Preferably, in step 3, the objective function is:

[0021]

[0022] Where R represents the peak shaving and valley filling revenue from the current moment to the end of the day, and P... disch,i and P ch,i S represents the discharge power and charging power of the energy storage in the i-th scheduling cycle. disch,i and S ch,i The charge / discharge state of the energy storage in the i-th scheduling cycle is represented by α·P, which is a 0-1 variable. pv,i The photovoltaic (PV) grid connection benefit is α, where α is the PV benefit coefficient, and P is the PV grid connection benefit. pv,i For photovoltaic power consumption, β·C age,i Let β represent the energy storage aging cost of the energy storage output in the i-th scheduling cycle of the semi-empirical battery aging assessment model, where β represents the unit aging cost and T represents the duration of a single scheduling cycle.

[0023] Preferably, the photovoltaic absorption capacity P pv,i The calculation method is as follows:

[0024] P PV,i =min(max(0,PV) forecast,i -Load i ), Battery capacity ×(1-SOC i ));

[0025] Among them, PV forecast,i This indicates the predicted photovoltaic output for the current time period. i This represents the current period's load forecast value, Battery capacity State of Charge (SOC) indicates the rated capacity of the energy storage system. i This indicates the real-time energy storage charge status during the current time period.

[0026] Preferably, the objective function is provided with constraints, including energy storage charging and discharging power constraints, energy storage capacity constraints, energy storage charging and discharging state constraints, demand constraints, and anti-reverse current constraints.

[0027] Preferably, the actual demand value of the user must be less than or equal to the target demand value under the demand constraint.

[0028] D actual <Dapply;

[0029] Among them, D actual D represents the user's current actual demand value. apply This indicates the set target demand value.

[0030] Preferably, the anti-reverse current constraint ensures that the energy storage discharge power cannot exceed the current load power at any given time.

[0031]

[0032] Among them, P disch,t P represents the current energy storage discharge power. load,t This indicates the current user load power.

[0033] Preferably, in step 4, the particle swarm optimization algorithm steps are as follows:

[0034] 1) Initialization: Determine the size N of the particle swarm, the dimension D of the position and velocity of each particle, randomly initialize the position and velocity of each particle, and initialize the individual optimal position of each particle as its current position;

[0035] 2) Calculate fitness: For each particle, calculate the corresponding fitness and compare it with the fitness of the individual's best position;

[0036] 3) Update the individual's best position: If the fitness of the current position is better than that of the individual's best position, then set the current position as the individual's best position;

[0037] 4) Update the global best position: Find the particle with the best fitness in the entire particle swarm and set its position as the global best position;

[0038] 5) Update particle velocity and position: For each particle, based on its current velocity and position, as well as its individual and global best positions, update the particle's velocity using the velocity update formula and update its position using the position update formula;

[0039] 6) Repeat steps 2-5 until the number of iterations reaches the preset value, and output the final global best position as the optimal solution to the optimization problem.

[0040] Preferably, the particle position update method in the particle swarm optimization algorithm is as follows:

[0041]

[0042] in, This represents the position of particle i at time t+1. This represents the position of particle i at time t. Let represent the velocity of particle i at time t.

[0043] Preferably, the particle velocity update method in the particle swarm optimization algorithm is as follows:

[0044]

[0045] in, Let w represent the velocity of particle i at time t+1, w be the inertia weight, c1 and c2 be the learning factors, and r1 and r2 be random numbers between 0 and 1. pbest i represents the historical best position of particle i, and gbest is the historical best position of the entire swarm.

[0046] Preferably, in step 5, the optimal energy storage charging and discharging strategy is a rolling optimization strategy, in which steps 1 to 4 are re-executed every scheduling cycle to achieve real-time dynamic optimization control.

[0047] This invention has the following characteristics and beneficial effects:

[0048] This method, based on a semi-empirical battery aging model, incorporates aging costs into the objective function and optimizes energy storage charging and discharging strategies in real time: During peak photovoltaic output periods, it prioritizes absorbing excess electricity for local consumption while mitigating the risk of exceeding maximum demand limits; during periods of load fluctuation, it intelligently adjusts energy storage charging and discharging, extending battery life through adaptive strategies while meeting anti-reverse current requirements, thereby improving overall lifecycle returns. This method effectively resolves the triple contradiction of "efficient photovoltaic resource consumption, demand cost control, and battery life protection," achieving a balance between the economic viability and sustainability of energy storage systems.

[0049] It can accurately predict photovoltaic output and load fluctuations, dynamically optimize energy storage charging and discharging strategies, meet users' electricity needs, effectively avoid the risk of exceeding maximum demand limits, reduce additional demand electricity costs, and maximize the economic efficiency throughout the entire life cycle by integrating dynamic assessment of battery aging. This ensures system reliability while extending battery life, avoiding capacity decay and long-term revenue loss caused by overcharging and discharging, and ultimately achieving the triple goals of "efficient photovoltaic resource utilization - economical operation of energy storage system - continuous optimization of user electricity costs", providing core support for the sustainable development of energy storage systems. Attached Figure Description

[0050] Figure 1 This is a flowchart illustrating an energy storage system optimization control method that integrates aging assessment and photovoltaic absorption, according to an embodiment of the present invention.

[0051] Figure 2 This is a schematic diagram of a typical daily photovoltaic-load-energy storage coordinated operation curve according to an embodiment of the present invention. Detailed Implementation

[0052] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0053] An optimized control method for energy storage systems that integrates aging assessment and photovoltaic power consumption, such as Figure 1 As shown, it includes the following steps:

[0054] Step 1: Construct a multi-source data fusion and dynamic prediction model: A Long Short-Term Memory (LSTM) network is used to establish a load forecasting model and a photovoltaic (PV) output forecasting model, respectively. Inputs include historical load data, historical PV output data, ambient temperature, date type, irradiance, and cloud cover data. Outputs are the predicted load and PV output values ​​for future periods.

[0055] Specifically, this invention employs a Long Short-Term Memory (LSTM) network to construct a dual prediction model for the load and photovoltaic (PV) output of an energy storage system, achieving accurate predictions by fusing multi-source time-series data. LSTM, with its unique gating mechanism, effectively captures periodic fluctuations in load data (such as diurnal patterns and weekly cycles) and external influencing factors (such as weather changes and date types), while accurately modeling the intermittent characteristics of PV output (affected by dynamic factors such as irradiance and cloud cover). The prediction system is based on historical load curves, overlaying key features such as ambient temperature and date type to construct the input sequence; PV prediction integrates historical output, irradiance, and cloud cover data, specifically incorporating the core constraint of "PV power not being able to connect to the grid and needing local consumption" into the prediction process—when the predicted output exceeds the current absorption capacity of the energy storage system, it automatically triggers a priority adjustment for consumption, marking the excess as quantifiable waste and incorporating it into subsequent optimization objectives. The model adopts a two-layer LSTM structure design, using a Dropout layer to suppress overfitting and combining time-series cross-validation to ensure prediction accuracy.

[0056] Furthermore, Long Short-Term Memory (LSTM) is a time-centric recurrent neural network, mainly composed of forget gates, memory gates, and output gates.

[0057] The forgetting gate determines which information is forgotten:

[0058] f t =σ(w f ×[h t-1 ,x t ]+b f (1)

[0059] f t This is the output of the forget gate, x t It is the input data at the current moment, h t-1 It is the output from the previous moment, w f Represents weight, b f σ represents the bias, and σ represents the Sigmoid activation function.

[0060] Memory gates determine which information is retained:

[0061]

[0062] Among them, i t The output of the sigmoid function, which uses a memoization gate, determines which values ​​are updated. (C) t ′ is the output of the activation function Tanh at the current time, w i and w c b represents the weight. i and b c Indicates bias.

[0063] The output gate generates the result and hidden state at the current time step:

[0064]

[0065] Among them, C t It is the current state of the memory cells, C t-1 It refers to the state of the memory cells in the previous moment, O t It is the output of the sigmoid activation function of the output gate, h t It is the output at the current moment.

[0066] LSTM uses these three basic gating units to selectively forget or remember relevant time series information to prevent gradient vanishing and exploding, thus better capturing long-term dependencies in time series data and making it suitable for nonlinear data prediction.

[0067] The prediction model acquires relevant data, preprocesses the data according to the required fine granularity, and then inputs it into the LSTM recurrent neural network to train the model, thereby achieving accurate prediction of short-term load and photovoltaic power.

[0068] Step 2: Construct a semi-empirical battery aging assessment model: Based on the real-time state of charge (SOC), charge-discharge rate (C-rate), and battery temperature data collected by the battery management system (BMS), a semi-empirical battery aging assessment model is established to quantify the aging cost during the charge-discharge process.

[0069] Specifically, key operational data of the energy storage system, including the current state of charge (SOC), charge-discharge rate (C-rate), and battery operating temperature, are acquired in real time through the battery management system (BMS). Outlier removal and linear interpolation methods form the data foundation for dynamic aging assessment. Based on industry-standard battery aging characteristics, a semi-empirical battery aging assessment model is constructed. This model achieves accurate quantification of aging costs during the charge-discharge process by integrating the battery's physical aging mechanism with actual operational experience data. The core idea of ​​the model is that battery aging is affected by multiple factors such as SOC, C-rate, and temperature. A semi-empirical formula is used to establish a dynamic correlation between these factors and the aging rate, avoiding the limitations of traditional static models.

[0070] The dynamic evaluation model for battery aging is shown in equation (4) below:

[0071]

[0072] Where Qbatt(0) is the initial capacity of the battery (unit: Ah), Qbatt(p,Ah) is the "benchmark" of capacity loss, the aging parameter p corresponds to the remaining capacity under the state (unit: Ah), p can be the number of cycles, aging time, or an equivalent factor of the comprehensive aging scenario, and Qbatt(0)-Qbatt(p,Ah) is the absolute capacity loss (Ah).

[0073] The aging formula can be further rewritten based on relevant electrochemical formulas as follows:

[0074] Q loss (p,Ah)=σ funct (p)·Ah Z (5)

[0075] Where Z is the power-law exponent characterizing Ah throughput dependence, σ funct (p) is a nonlinear function of the severity factor, mainly representing the synergistic effect of multiple factors such as SOC, C-rate, and temperature on battery aging, as shown in Equation 6 below:

[0076]

[0077] Where α and β are SOC correction coefficients, Ea is the activation energy (characterizing the effect of temperature on the aging rate), η is the current coefficient (reflecting the sensitivity of high-rate charge-discharge to aging), and I... C Let θ be the charging / discharging current, θ be the battery temperature, and Rg be the gas constant.

[0078] To ensure the model accurately matches actual battery aging characteristics, this invention employs a data-driven dynamic parameter tuning method. The system constructs a multi-dimensional aging feature library based on historical operating data continuously collected by the BMS, and uses a nonlinear least squares fitting algorithm to match the model output with actual aging data in real time. This ensures a high degree of consistency between the aging assessment and actual operating conditions, providing reliable input for subsequent optimization strategies.

[0079] Step 3: Multi-objective optimization decision model

[0080] 1) Objective function

[0081] The life-cycle optimization objective function constructed in this invention strictly follows the energy flow logic (PV → load → energy storage), achieving dynamic coordination between peak shaving and valley filling, PV absorption, and energy storage aging. The objective function is designed as follows:

[0082]

[0083] Where R represents the peak shaving and valley filling revenue from the current moment to the end of the day, and P... disch,i and P ch,i S represents the discharge power and charging power of the energy storage in the i-th scheduling cycle. disch,i and S ch,i The charge / discharge state of the energy storage in the i-th scheduling cycle is represented by α·P, which is a 0-1 variable. pv,i The photovoltaic (PV) grid connection benefit is α, where α is the PV benefit coefficient, and P is the PV grid connection benefit. pv,i For photovoltaic power consumption, β·C age,i β represents the energy storage aging cost, T represents the unit aging cost, and T represents the duration of a single scheduling cycle.

[0084] Photovoltaic (PV) power output prioritizes meeting load demand; only when PV output exceeds load is the excess absorbed by energy storage. The specific formula for PV absorption capacity is:

[0085] P PV,i =min(max(0,PV) forecast,i -Load i ), Battery capacity ×(1-SOC i (8)

[0086] Among them, PV forecast,i This indicates the predicted photovoltaic output for the current time period. i This represents the current period's load forecast value, Battery capacity State of Charge (SOC) indicates the rated capacity of the energy storage system. i This indicates the real-time energy storage charge status during the current time period.

[0087] The specific expression for the energy storage aging cost is as follows:

[0088]

[0089] 2) Constraints

[0090] The corresponding constraints include conventional energy storage charging and discharging power constraints and energy storage capacity constraints.

[0091] Energy storage charging and discharging power constraints: the charging and discharging power of energy storage at any given time cannot exceed the maximum energy storage power.

[0092]

[0093] Among them, P disch and P ch P represents the discharge power and charging power of energy storage. es,max This indicates the maximum value of the energy storage's own charging and discharging power.

[0094] Energy storage capacity is constrained; the State of Charge (SOC) of the stored energy at any given time cannot exceed the set upper and lower limits to prevent overcharging or over-discharging.

[0095]

[0096] Among them, E es,t State of Charge (SOC) indicates the current charge state of the stored energy. min and SOC max These represent the upper and lower limits of the charge state set for energy storage, respectively.

[0097] Energy storage charge / discharge state constraints: Energy storage can only be in one of the charging or discharging states at any given time.

[0098]

[0099] To achieve effective demand control and backflow prevention, specific constraints need to be added: demand constraints and backflow prevention constraints.

[0100] Demand constraint: The user's actual demand must be less than or equal to the target demand.

[0101] D actual <Dapply (13)

[0102] Among them, D actual D represents the user's current actual demand value. apply This indicates the set target demand value.

[0103] Anti-reverse current constraint: the energy storage discharge power cannot exceed the current load power at any time, and the net margin is greater than 0.

[0104]

[0105] Among them, P disch,t P represents the current energy storage discharge power. load,t This indicates the current user load power.

[0106] Understandably, only when the net surplus is greater than 0 does it mean there is excess photovoltaic power that needs to be absorbed by energy storage; otherwise, the charging and discharging of energy storage will still be carried out according to peak shaving and valley filling. Energy storage aging factors also need to be considered.

[0107] Step 4: Solve the multi-objective optimization decision model using the Particle Swarm Optimization (PSO) algorithm to obtain the optimal energy storage charging and discharging strategy. PSO is a heuristic swarm intelligence algorithm that finds the optimal solution by simulating the cooperation and competition among individuals in a flock of birds or a school of fish. The PSO algorithm is simple and easy to implement, possesses good global optimization ability, good convergence and robustness, and is well-suited for multimodal functions and high-dimensional optimization problems. In the PSO algorithm, each solution in the search space is called a particle. Each particle has a position and velocity, and updates its position and velocity based on its own experience and the experience of the group to find the optimal solution. The basic idea of ​​the PSO algorithm is to continuously adjust the position and velocity of particles to search for the optimal solution, thereby simulating information sharing and cooperation among individuals in a group.

[0108] The particle position update in the PSO algorithm is shown in Equation 15:

[0109]

[0110] in, This represents the position of particle i at time t+1. This represents the position of particle i at time t. Let represent the velocity of particle i at time t.

[0111] The particle velocity update formula is shown in Equation 16:

[0112]

[0113] in, Let w represent the velocity of particle i at time t+1, w be the inertia weight, c1 and c2 be the learning factors, and r1 and r2 be random numbers between 0 and 1. pbest i represents the historical best position of particle i, and gbest is the historical best position of the entire swarm.

[0114] The overall steps of the PSO optimization algorithm are as follows:

[0115] 1) Initialization: Determine the size N of the particle swarm, the dimension D of the position and velocity of each particle, randomly initialize the position and velocity of each particle, and initialize the individual optimal position of each particle as its current position.

[0116] 2) Calculate fitness: For each particle, calculate the corresponding fitness and compare it with the fitness of the individual's best position.

[0117] 3) Update the individual's best position: If the fitness of the current position is better than that of the individual's best position, then set the current position as the individual's best position.

[0118] 4) Update the global best position: Find the particle with the best fitness in the entire particle swarm and set its position as the global best position.

[0119] 5) Update particle velocity and position: For each particle, update the particle velocity using the velocity update formula and update the particle position using the position update formula, based on the current velocity and position, as well as the individual and global best positions.

[0120] 6) Repeat steps 2-5 until the number of iterations reaches the preset value, and output the final global best position as the optimal solution to the optimization problem.

[0121] Step 5: Control the operation of the energy storage system according to the optimal energy storage charging and discharging strategy.

[0122] The overall optimization control of this invention constructs a closed-loop dynamic optimization system. Starting with load-priority photovoltaic consumption, it constrains the energy storage strategy in real time through a dynamic aging cost model, and finally outputs the optimal charging and discharging plan that meets safety constraints.

[0123] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

[0124] To verify the effectiveness of the control method described in this invention, an empirical analysis was conducted using a commercial and industrial park as the application object. This park is equipped with a photovoltaic power generation system and a 200kWh energy storage system (rated charging and discharging power 100kW). The proposed synergistic optimization control strategy combining energy storage aging and photovoltaic absorption is employed. A typical day was selected for testing. Through high-precision short-term prediction, the system predicted a relative photovoltaic surplus window between 12:00 and 13:00. Based on this prediction information, the optimized controller of this invention dynamically adjusts the energy storage charging strategy in advance: reducing the total charging amount during low electricity price periods, effectively absorbing the photovoltaic surplus; and simultaneously, by introducing an aging cost term, automatically balancing the charging rate, effectively reducing the cycle aging rate. Specifically, as follows... Figure 2 As shown:

[0125] Using the entire lifecycle of the energy storage system as the evaluation period, under the same configuration and operational boundary conditions, compared with the traditional "peak shaving and valley filling" control strategy, this solution not only improves the local photovoltaic consumption rate but also effectively controls the charging and discharging behavior of energy storage, significantly delaying the battery aging process. Therefore, while ensuring the economic efficiency of system operation, it further enhances the overall profitability and return on investment of energy storage assets throughout their entire lifecycle.

[0126] This case demonstrates that by integrating photovoltaic-load forecasting with energy storage aging models, the present invention enables the system to have "proactive absorption capacity," allowing it to proactively schedule energy storage resources before photovoltaic overflow occurs, taking into account economy, absorption efficiency, and equipment lifespan, and has significant engineering practical value.

[0127] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing and controlling an energy storage system that integrates aging assessment and photovoltaic power generation, characterized in that, Includes the following steps: Step 1: Establish a load forecasting model and a photovoltaic output forecasting model using a long short-term memory network. Use historical load data, ambient temperature, and date type as inputs to the load forecasting model and output the load forecast for future periods. Use historical photovoltaic output data, irradiance, and cloud cover data as inputs to the photovoltaic output forecasting model and output the photovoltaic output forecast for future periods. Step 2: Based on the real-time data of state of charge, charge / discharge rate and battery temperature collected by the battery management system, establish a semi-empirical battery aging assessment model to quantify the aging cost during the charge and discharge process. Step 3: Construct a multi-objective optimization decision model: With peak shaving and valley filling revenue, photovoltaic consumption revenue and energy storage aging cost as optimization objectives, and with energy storage charging and discharging power constraints, power constraints, charging and discharging state constraints, demand constraints and anti-reverse current constraints as constraints, establish an objective function; Step 4: Solve the multi-objective optimization decision model based on the particle swarm optimization algorithm to obtain the optimal energy storage charging and discharging strategy; Step 5: Control the operation of the energy storage system according to the optimal energy storage charging and discharging strategy.

2. The method according to claim 1, characterized in that, In step 1, the load forecasting model and the photovoltaic output forecasting model introduce a Dropout layer to prevent overfitting; in the photovoltaic output forecasting model, when the photovoltaic output forecast value in the future period exceeds the load forecast value, the energy storage system will take priority in absorbing the load. If the load exceeds its absorption capacity, the excess portion will be marked as quantifiable waste and included in the optimization target.

3. The method according to claim 1, characterized in that, In step 2, the collected electrical state, charge / discharge rate, and battery temperature data need to be preprocessed by outlier removal and linear interpolation.

4. The method according to claim 1, characterized in that, The semi-empirical battery aging assessment model: Q loss (p,Ah)=σ funct (p)·Ah Z ; Where Z is the power-law exponent characterizing Ah throughput dependence, σ funct (p) is a nonlinear function of the severity factor; Where α and β are SOC correction coefficients, Ea is activation energy, η is current coefficient, and I C Let θ be the charging / discharging current, θ be the battery temperature, and Rg be the gas constant.

5. The method according to claim 1, characterized in that, In step 3, the objective function is: Where R represents the peak shaving and valley filling revenue from the current moment to the end of the day, and P... disch,i and P ch,i S represents the discharge power and charging power of the energy storage in the i-th scheduling cycle. disch,i and S ch,i The charge / discharge state of the energy storage in the i-th scheduling cycle is represented by α·P, which is a 0-1 variable. pv,i The photovoltaic (PV) grid connection benefit is α, where α is the PV benefit coefficient, and P is the PV grid connection benefit. pv,i For photovoltaic power consumption, β·C age,i Let β represent the energy storage aging cost of the energy storage output in the i-th scheduling cycle of the semi-empirical battery aging assessment model, where β represents the unit aging cost and T represents the duration of a single scheduling cycle.

6. The method according to claim 5, characterized in that, The photovoltaic absorption capacity P pv,i The calculation method is as follows: P PV,i =min(max(0,PV forecast,i -Load i ),Battery capacity ×(1-SOC i )); Among them, PV forecast,i This indicates the predicted photovoltaic output for the current time period. i This represents the current period's load forecast value, Battery capacity State of Charge (SOC) indicates the rated capacity of the energy storage system. i This indicates the real-time energy storage charge status during the current time period.

7. The method according to claim 5, characterized in that, The objective function is subject to constraints, including energy storage charging and discharging power constraints, energy storage capacity constraints, energy storage charging and discharging state constraints, demand constraints, and anti-reverse current constraints.

8. The method according to claim 7, characterized in that, The demand constraint stipulates that the user's actual demand value must be less than or equal to the target demand value. D actual <Dapply; Among them, D actual D represents the user's current actual demand value. apply This indicates the set target demand value.

9. The method according to claim 7, characterized in that, The anti-reverse current constraint ensures that the energy storage discharge power cannot exceed the current load power at any given time. Among them, P disch,t P represents the current energy storage discharge power. load,t This indicates the current user load power.

10. The method according to any one of claims 1-9, characterized in that, In step 4, the particle swarm optimization algorithm steps are as follows: 1) Initialization: Determine the size N of the particle swarm, the dimension D of the position and velocity of each particle, randomly initialize the position and velocity of each particle, and initialize the individual optimal position of each particle as its current position; 2) Calculate fitness: For each particle, calculate the corresponding fitness and compare it with the fitness of the individual's best position; 3) Update the individual's best position: If the fitness of the current position is better than that of the individual's best position, then set the current position as the individual's best position; 4) Update the global best position: Find the particle with the best fitness in the entire particle swarm and set its position as the global best position; 5) Update particle velocity and position: For each particle, based on its current velocity and position, as well as its individual and global best positions, update the particle's velocity using the velocity update formula and update its position using the position update formula; 6) Repeat steps 2-5 until the number of iterations reaches the preset value, and output the final global best position as the optimal solution to the optimization problem.

11. The method according to claim 10, characterized in that, The particle position update method in the particle swarm optimization algorithm is as follows: in, This represents the position of particle i at time t+1. This represents the position of particle i at time t. Let represent the velocity of particle i at time t.

12. The method according to claim 10, characterized in that, The particle velocity update method in the particle swarm optimization algorithm is as follows: in, Let w represent the velocity of particle i at time t+1, w be the inertia weight, c1 and c2 be the learning factors, and r1 and r2 be random numbers between 0 and 1. pbest i represents the historical best position of particle i, and gbest is the historical best position of the entire swarm.

13. The method according to claim 1, characterized in that, In step 5, the optimal energy storage charging and discharging strategy is a rolling optimization strategy. Steps 1 to 4 are re-executed every scheduling cycle to achieve real-time dynamic optimization control.

Images