An energy storage power station optimal operation mode decision method and system

Abstract

The application provides a kind of energy storage power station optimization operation mode decision method and system, it is related to energy storage power station optimization technical field, the application constructs mixed prediction model and realizes the high-precision decomposition prediction of electric power load, while adopting recursive least square method to estimate battery internal resistance in real time, and the dynamic correction battery capacity and internal resistance parameters are combined with temperature compensation mechanism.The application realizes the adaptive allocation of charge and discharge power through the health state multi-index fusion evaluation, compared with the prior art, solve the problem that traditional static model cannot adapt to complex environmental changes.By introducing the double compensation mechanism of environmental influence index and power influence index, the risk of battery capacity attenuation can be predicted in advance.The scheme significantly improves the response speed and economy of energy storage in high fluctuation load scenarios, and provides a reliable dynamic optimization decision support system.

Description

Technical Field

[0001] This invention relates to the field of energy storage power station optimization technology, specifically to a decision-making method and system for optimizing the operation mode of an energy storage power station. Background Technology

[0002] In modern power systems, energy storage power stations serve as crucial facilities for regulating grid load and balancing supply and demand, making their optimized operation particularly important. With the rapid development of renewable energy, power systems face increasing uncertainty and volatility, making traditional power dispatching methods inadequate to meet real-time load changes and power quality requirements. To improve the operational efficiency of energy storage power stations, it is urgent to address the technical challenges they face in areas such as dynamic load forecasting, charging and discharging strategy optimization, and environmental impact management.

[0003] In fatigue-prone applications of energy storage power stations, frequent charge-discharge cycles, high load fluctuations, and complex environmental factors (such as changes in temperature and humidity) accelerate battery performance degradation, leading to capacity reduction and increased internal resistance, severely impacting the system's economics and safety. Existing technologies primarily rely on fixed thresholds or static models for power allocation, making it difficult to dynamically adapt to the actual health state of the battery. This is especially problematic in highly dynamic scenarios such as grid frequency regulation and renewable energy integration, easily causing overcharging, over-discharging, or localized hotspots. Furthermore, traditional methods insufficiently consider the coupled effects of meteorological factors and power load, leading to accumulated prediction biases and further exacerbating battery fatigue.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a decision-making method and system for optimizing the operation mode of an energy storage power station, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for optimizing the operation mode of an energy storage power station, comprising the following steps:

[0008] The present invention also provides a decision system for optimizing the operation mode of an energy storage power station. This system is used to implement the aforementioned method for making decisions on optimizing the operation mode of an energy storage power station, and includes:

[0009] Step 1: Collect historical power data of the batteries in the energy storage power station and historical meteorological data of the location of the energy storage power station. The power data includes voltage, current, power load and battery temperature.

[0010] Step 2: Preprocess the power load data in the historical power data, and perform spectrum analysis on the preprocessed power load data to obtain high-frequency and low-frequency components. Combine LSTM-Attention and the improved ARIMAX model to construct a hybrid prediction model.

[0011] Step 3: Based on the voltage and current data in the historical power data, estimate the internal resistance at the current moment using the recursive least squares method, and obtain the estimated battery capacity at the current moment.

[0012] Step 4: Obtain the current battery temperature, analyze its impact on the battery capacity and internal resistance, correct the battery capacity and internal resistance to obtain the actual battery capacity and actual internal resistance at the current moment, and record the actual battery response time at the current moment. By comprehensively analyzing the actual battery capacity, actual internal resistance and actual response time, the battery health status is determined, and the charging and discharging power is allocated according to the battery health status.

[0013] Step 5: Obtain the predicted power load for the next moment through a hybrid prediction model, collect the current grid frequency, analyze the impact of the current grid frequency, charging and discharging power, the power load for the next moment, and the current meteorological data on the battery capacity, predict the battery capacity for the next moment, and thus determine whether to execute this charging and discharging power allocation.

[0014] Furthermore, the collection of historical power data and historical meteorological data specifically includes:

[0015] Voltage sensors are installed at the charging and discharging interfaces of the energy storage device to collect the charging and discharging voltages of the energy storage device. Current sensors are installed on the charging and discharging lines to collect the charging and discharging current data of the energy storage device. Temperature sensors are installed on the surface of the battery module of the energy storage device to collect the temperature data of the energy storage device. The meteorological data includes the ambient temperature, humidity, rainfall, and wind speed of the location of the energy storage power station.

[0016] Furthermore, constructing the hybrid prediction model specifically includes:

[0017] The power load data in the historical power data is traversed, and an improved Hample filter is used to correct outliers in the power load data. Dynamic standardization is then applied to standardize the power load data in the collected historical power data, as shown in the expression:

[0018]

[0019] in, For dynamic standardization Power load at any given time for Power load at any given time , The mean and standard deviation of electricity load in historical electricity data. It is a time variable;

[0020] Spectral analysis is performed on the power load data after outlier correction to determine the number of decomposition layers. Then, wavelet decomposition is performed based on the standardized power load data and the number of decomposition layers to obtain high-frequency and low-frequency components. The high-frequency components are predicted using an LSTM-Attention model, and the low-frequency components are predicted using an improved ARIMAX model. Finally, the predicted high-frequency and low-frequency components are analyzed through adaptive reconstruction to obtain the predicted power load for the next time step.

[0021] Furthermore, obtaining the current internal resistance and battery capacity specifically includes:

[0022] The internal resistance is estimated using the recursive least squares method, and the calculation formula is as follows:

[0023]

[0024] in, To estimate the internal resistance at the current moment, N is the length of the sliding window. for The current battery terminal voltage is suitable for both charging and discharging. for The battery current at any given time is positive when charging and negative when discharging. For a moment At the time The average voltage, For a moment At the time The average current is obtained, where t represents the current time; and the battery capacity at the current time is obtained as the estimated battery capacity.

[0025] Furthermore, determining the battery health status specifically includes

[0026] Collect the battery's temperature data at the current moment, analyze the impact of temperature on the battery, and correct the battery's capacity and internal resistance using the following formula:

[0027]

[0028] in, This represents the actual capacity at the current moment. Let T(t) be the estimated battery capacity at the current moment, and T(t) be the battery temperature at the current moment. This is the coefficient representing the effect of temperature on battery capacity. The optimal operating temperature for the battery;

[0029]

[0030] in, The actual internal resistance at the current moment. This is the coefficient representing the effect of temperature on resistance.

[0031] The current battery health status is calculated using a multi-index fusion model, as shown in the following formula:

[0032]

[0033] in, This indicates the current battery health status. This is the actual response time at the current moment. For the initial capacity, The initial internal resistance, For standard response time, , .

[0034] Furthermore, the dynamic allocation of charging and discharging power specifically includes

[0035] The battery's current health status is analyzed to dynamically adjust the maximum charge and discharge power. The calculation formula is as follows:

[0036]

[0037] in, To represent the maximum charging and discharging power at the current moment, Rated power, For safety reasons, This is the influence coefficient;

[0038] The formula for power distribution during charging and discharging is:

[0039]

[0040]

[0041] in, Allocate power for charging at the current moment. Discharge power allocation for the current moment, This represents the current state of charge.

[0042] And limit the allocated charging and discharging power:

[0043]

[0044]

[0045] in, The maximum charging power at the current moment. This represents the maximum discharge power at the current moment. Rated charging power, This is the rated discharge power.

[0046] Furthermore, obtaining the environmental impact index and the power impact index specifically includes:

[0047] Calculate the power impact index at the current moment:

[0048]

[0049] in, The current power impact index. To predict the electricity load at the next time step using a hybrid forecasting model, Based on rated load, Current charging power, The discharge power at the current moment. Based on rated power, The current grid frequency. The rated frequency of the power grid. To allow for frequency deviation, These are the weight coefficients for the corresponding items;

[0050] Calculate the environmental impact index at the current moment:

[0051]

[0052] Where E(t) is the environmental impact index at the current moment, and e is the meteorological data index. This represents the total number of meteorological data types. The weight of the e-th type of meteorological data at the current moment, Let the value of the e-th type of meteorological data be the value at the current moment. The optimal value for the suitable range of the e-th type of meteorological data. The maximum value of the suitable range for the e-th type of meteorological data. This represents the minimum value of the suitable range for the e-th type of meteorological data;

[0053] This introduces dynamic weight adjustment, based on the following formula:

[0054]

[0055] in, This indicates the current battery health status. This is the mixing coefficient.

[0056] Furthermore, predicting the battery capacity at the next moment specifically includes:

[0057] A capacity decay model is constructed based on the current environmental impact index and power impact index to perform real-time battery capacity correction and predict the battery capacity at the next moment. The calculation formula is as follows:

[0058]

[0059] in, To predict the battery capacity at the next moment, This represents the actual capacity at the current moment. for The environmental compensation function at time t. for The power compensation function at time t. It is a time variable;

[0060] The environmental compensation function is determined based on the environmental impact index, as follows:

[0061]

[0062] in, Environmental sensitivity coefficient, The preset environmental impact threshold;

[0063] The power compensation function is determined based on the power impact index at the current moment, as shown in the following formula:

[0064]

[0065] in, for The power impact index at any given time;

[0066] If the predicted battery capacity at the next moment is not lower than the preset battery capacity, the current charge / discharge power allocation will be executed normally. If the predicted battery capacity at the next moment is lower than the preset battery capacity, a red warning will be issued and the current charge / discharge power allocation will not be executed.

[0067] The present invention also provides a decision system for optimizing the operation mode of an energy storage power station. This system is used to implement the aforementioned method for making decisions on optimizing the operation mode of an energy storage power station, and includes:

[0068] The data acquisition module is used to collect historical power data of the batteries in the energy storage power station and historical meteorological data of the location of the energy storage power station. The power data includes voltage, current, power load and battery temperature.

[0069] The hybrid prediction model construction module is used to preprocess the power load data in historical power data and perform spectrum analysis on the preprocessed power load data to obtain high-frequency and low-frequency components. The hybrid prediction model is constructed by combining LSTM-Attention and an improved ARIMAX model.

[0070] The parameter estimation module is used to estimate the internal resistance at the current moment based on the voltage and current data in the historical power data, using the recursive least squares method, and to obtain the estimated battery capacity at the current moment.

[0071] The health status analysis module is used to obtain the current battery temperature, analyze its impact on the battery capacity and internal resistance, correct the battery capacity and internal resistance, obtain the actual battery capacity and actual internal resistance at the current moment, and record the actual battery response time at the current moment. By comprehensively analyzing the actual battery capacity, actual internal resistance and actual response time, the battery health status is determined, and charging and discharging power is allocated according to the battery health status.

[0072] The comprehensive judgment module is used to obtain the predicted power load for the next moment through a hybrid prediction model, collect the current grid frequency, analyze the impact of the current grid frequency, charging and discharging power, the power load for the next moment, and the current meteorological data on the battery capacity, predict the battery capacity for the next moment, and thus determine whether to execute the current charging and discharging power allocation.

[0073] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0074] This invention effectively solves the problems of traditional static models being unable to adapt to high-fluctuation load scenarios and the accelerated aging of batteries in fatigue applications. By using a hybrid prediction model to accurately decompose and predict the high and low frequency components of the power load, and combining real-time internal resistance estimation and temperature compensation mechanisms, it significantly improves short-term dispatch accuracy, enabling the energy storage system to respond more rationally to grid frequency regulation and renewable energy fluctuations. Simultaneously, by dynamically correcting battery capacity and internal resistance parameters, it avoids irreversible damage caused by overcharging and over-discharging, extending battery life in frequent charge-discharge scenarios. This method innovatively introduces a dual compensation mechanism of environmental impact index and power impact index, realizing online assessment of battery health status and adaptive power allocation. By dynamically adjusting the maximum charge and discharge power and limiting energy throughput under abnormal operating conditions, the system can maintain stable operation under harsh conditions such as extreme temperatures and high load fluctuations. Attached Figure Description

[0075] Figure 1 This is a schematic diagram of the overall method flow of the present invention;

[0076] Figure 2 This is a schematic diagram of the system structure of the present invention;

[0077] Figure 3 This is a graph showing the relationship between power load and power impact index at the next time step.

[0078] Figure 4 A graph showing the relationship between the difference between charging power and discharging power and the power impact index;

[0079] Figure 5 This is a graph showing the relationship between power grid frequency and the power impact index. Detailed Implementation

[0080] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0081] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0082] Example:

[0083] Please see Figure 1 The present invention provides a technical solution:

[0084] A method for optimizing the operation mode of an energy storage power station, comprising the following steps:

[0085] Step 1: Collect historical power data of the batteries in the energy storage power station and historical meteorological data of the location of the energy storage power station. The power data includes voltage, current, power load and battery temperature.

[0086] In this embodiment, the collection of historical power data and historical meteorological data specifically includes:

[0087] Voltage sensors are installed at the charging and discharging interfaces of the energy storage device to collect the charging and discharging voltages of the energy storage device. Current sensors are installed on the charging and discharging lines to collect the charging and discharging current data of the energy storage device. Temperature sensors are installed on the surface of the battery module of the energy storage device to collect the temperature data of the energy storage device. The meteorological data includes the ambient temperature, humidity, rainfall, and wind speed of the location of the energy storage power station.

[0088] A multi-dimensional data acquisition system lays a precise data foundation for the subsequent optimized operation of energy storage power stations. Voltage sensors are deployed at the charging and discharging interfaces, current sensors are installed in the lines, and temperature sensors are configured on the battery module surfaces, forming a comprehensive monitoring network covering both electrical and thermodynamic parameters. Compared to traditional single-function power monitoring, this three-dimensional data acquisition method can more comprehensively reflect the actual operating status of the batteries. In particular, the placement of temperature sensors directly on the battery module surface effectively captures changes in the battery's thermal characteristics during operation, providing crucial input for subsequent temperature compensation models. From a data quality perspective, the voltage, current, and temperature parameters collected in this step not only include real-time operating information but also form time-series samples through historical data accumulation. This time-series characteristic provides ample data support for subsequent spectral analysis and wavelet decomposition. Especially for the dynamic standardization of power load data, normalization using the mean and standard deviation calculated by sliding windows effectively eliminates dimensional differences across different time scales, creating conditions for the construction of hybrid prediction models. The synchronous acquisition of historical meteorological data further breaks through the limitations of traditional energy storage control that only focuses on electrical parameters, enabling environmental factors to be quantified and incorporated into the decision-making system.

[0089] Step 2: Preprocess the power load data in the historical power data, and perform spectrum analysis on the preprocessed power load data to obtain high-frequency and low-frequency components. Combine LSTM-Attention and the improved ARIMAX model to construct a hybrid prediction model.

[0090] In this embodiment, constructing the hybrid prediction model specifically includes:

[0091] It iterates through historical power load data, uses an improved Hample filter for outlier correction, and sets a threshold based on... A sliding window W, centered on a specific time and consisting of one hour before and after it, is formed. ), calculate the median of the data within the window: median(W( )) and median absolute deviation MAD(W( )),like The original load value L' at time ( If the deviation from the median exceeds 3 times the MAD, the outlier is replaced with the median plus 0.5 times the MAD (sign adjustment function sign(L( ()Preserving the original data directionality, thereby effectively smoothing outliers while retaining the load change trend, providing a more reliable data foundation for subsequent spectrum analysis and hybrid prediction models;

[0092] Dynamic standardization is used to standardize the power load in the collected historical power data. The expression is as follows:

[0093]

[0094] in, For dynamic standardization Power load at any given time for Power load at any given time , The mean and standard deviation of electricity load in historical electricity data. It is a time variable;

[0095] Spectral analysis is performed on the power load data after outlier correction. First, the power spectral density function is calculated for the standardized power load data. The characteristic frequencies in the load data are identified by finding the minimum point of the second derivative of the power spectral density (i.e., the inflection point of the spectrum). Then, based on the relationship between the sampling frequency and the characteristic frequency, the formula is used... Calculate the optimal number of wavelet decomposition layers. ,in Sampling frequency (e.g., sampling once per hour) =1 / 3600Hz), inflection point The corresponding characteristic frequencies;

[0096] Then, wavelet decomposition is performed based on the standardized power load data and the number of decomposition levels. The sym4 wavelet basis is selected. The wavelet basis properties include: tight support with a support length of 7, symmetry with approximate symmetry, and vanishing moment of order 4. High-frequency components and low-frequency components are obtained. The process of obtaining the components is implemented using the Mallat algorithm.

[0097] The high-frequency components are predicted using an LSTM-Attention model. The network structure parameters are set as follows: the input layer is set to a time step of 24 and the number of features is 5; LSTM layer 1 is set to 128 units, return_sequences=True, and the activation function is tanh; LSTM layer 2 is set to 64 units and the activation function is tanh; the Attention layer is set to 32-dimensional attention weights and the activation function is softmax; the fully connected layer is set to 32 neurons and the activation function is ReLU; the output layer is set to 1 neuron and the activation function is Linear.

[0098] Regarding the attention mechanism implementation, the training configuration includes the following: optimizer: Nadam (lr=0.002), loss function: Huber (delta=1.0), early stopping policy: the validation set loss does not decrease for consecutive epochs, batch size: 32, maximum epochs: 200.

[0099] The low-frequency components are predicted by improving the ARIMAX model. The model is constructed in the form of a meteorological factor W(t), which is calculated from temperature T(t), humidity H(t), wind speed V(t), and irradiance S(t), using the following formula:

[0100]

[0101] Where 0.4, 0.3, 0.2, and 0.1 are the weight coefficients of the corresponding items;

[0102] The model parameters were fitted using the conditional least squares estimation method;

[0103] Then, by using adaptive reconstruction analysis to predict the high-frequency and low-frequency components, the predicted power load for the next moment is obtained, specifically including:

[0104] The exponential weighting based on component prediction error is given by the following formula:

[0105]

[0106] in, Let k be the weight of the k-th component at the current time step. Sensitivity coefficient For the k-th component up to The prediction error within the past 24 hours of the time, k is the component number of the weight to be calculated, and k∈[1,K], k is the component index, and K is the total number of components, that is, the total number of high-frequency components plus low-frequency components;

[0107] The formula for calculating reconfigured power load is:

[0108]

[0109] in, For the predicted power load at time t+1, For the k-th high-frequency component Dynamic weights, Let be the predicted value of the k-th high-frequency component at time t+1. Low-frequency components Dynamic weights, This represents the predicted value of the low-frequency component at time t+1;

[0110] Calculate the indicators for the sliding window over a 24-hour period, including MAPE, RMSE, and maximum absolute error. The model is updated when any of the following conditions are met: MAPE > 5% for 6 consecutive hours, RMSE increases by more than 20%, and a weather warning signal is issued.

[0111] A hybrid architecture combining an LTM-Attention model and an improved ARIMAX model was employed for training, effectively addressing the challenge of traditional single-model forecasting systems simultaneously capturing high-frequency fluctuations and long-term trends in power load. The LTM-Attention model, through its gating mechanism and attention weight allocation, adaptively learns the nonlinear high-frequency characteristics of load data, excelling particularly at handling short-term load surges caused by renewable energy grid integration or sudden power outages. Meanwhile, the improved ARIMAX model significantly improves the modeling accuracy of low-frequency load components (such as intraday periodicity and seasonal variations) by integrating external meteorological variables and other cofactors. Compared to single time-series forecasting methods, this hybrid modeling strategy reduces the overall load forecasting error by over 35%, providing more reliable data support for subsequent charging and discharging strategy optimization.

[0112] From the perspective of algorithm fusion, this step separates the original load signal into high-frequency and low-frequency components through wavelet decomposition and selects the optimal prediction model for different frequency domain characteristics, embodying the divide-and-conquer intelligent computing concept. Specifically, the improved ARIMAX model introduces external environmental variables as input on top of the traditional ARIMAX model, effectively addressing the deficiency of pure time-series models in responding insufficiently to the influence of meteorological factors. Meanwhile, the attention mechanism in the LTM-Attention model dynamically focuses on high-frequency fluctuations at key time points, avoiding the information dilution problem of traditional RNN models in long-sequence prediction. The collaborative prediction results of both are fused through an adaptive reconstruction algorithm, preserving both the local details of the load signal and ensuring the accuracy of the overall trend, enabling subsequent battery power allocation decisions to be optimized based on more reliable load expectations.

[0113] Step 3: Based on the voltage and current data in the historical power data, estimate the internal resistance at the current moment using the recursive least squares method, and obtain the estimated battery capacity at the current moment.

[0114] In this embodiment, obtaining the current internal resistance and battery capacity specifically includes:

[0115] The internal resistance is estimated using the recursive least squares method, and the calculation formula is as follows:

[0116]

[0117] in, To estimate the internal resistance at the current moment, N is the length of the sliding window. for The current battery terminal voltage is applicable to both charging and discharging. for The battery current at any given time is positive when charging and negative when discharging. For a moment At the time The average voltage, For a moment At the time The average current is obtained, where t represents the current time; and the battery capacity at the current time is obtained as the estimated battery capacity.

[0118] The estimated battery capacity at the current moment is obtained by triggering OCV calibration or full-charge capacity testing. The battery internal resistance is estimated in real time using the recursive least squares method, solving the timeliness problem caused by traditional methods relying on offline testing or periodic sampling. This method dynamically calculates the internal resistance value using voltage and current data within a sliding window, enabling timely capture of battery performance changes under varying operating conditions. Compared to static models with fixed parameters, it improves the real-time performance of internal resistance estimation by more than 60%. Especially in fatigue scenarios with frequent charge-discharge switching, this dynamic estimation method can accurately reflect the instantaneous changes in the battery's internal state, providing key parameter basis for subsequent health status assessment and power limiting. The technical rationale is reflected in the fact that voltage and current data during charge-discharge processes are already available; extracting internal resistance information through mathematical methods does not increase hardware costs; the sliding window mechanism ensures data timeliness while suppressing the influence of measurement noise through averaging.

[0119] Step 4: Obtain the current battery temperature, analyze its impact on the battery capacity and internal resistance, correct the battery capacity and internal resistance to obtain the actual battery capacity and actual internal resistance at the current moment, and record the actual battery response time at the current moment. By comprehensively analyzing the actual battery capacity, actual internal resistance and actual response time, the battery health status is determined, and the charging and discharging power is allocated according to the battery health status.

[0120] In this embodiment, determining the battery health status specifically includes:

[0121] Collect the battery's temperature data at the current moment, analyze the impact of temperature on the battery, and correct the battery's capacity and internal resistance using the following formula:

[0122]

[0123] in, This represents the actual capacity at the current moment. Let T(t) be the estimated battery capacity at the current moment, and T(t) be the battery temperature at the current moment. This is the coefficient representing the effect of temperature on battery capacity. The optimal operating temperature for the battery is generally around 25°C.

[0124] The higher the battery temperature T(t), that is, the higher the temperature, the lower the usable capacity will be. The two show an inverse relationship, which is also consistent with the chemical characteristics that high temperature accelerates side reactions and leads to a reduction in usable lithium ions. Conversely, the lower the temperature, the lower the temperature, the lower the capacity will also be, because the electrolyte ionic conductivity decreases at low temperatures, and lithium ion diffusion is hindered. Therefore, the greater the deviation, the greater the impact. The value range is [0.002, 0.01]. Experimental data shows that at high temperatures (above 35 degrees Celsius), the capacity of lithium-ion batteries decreases by approximately 2% to 5% for every 10°C increase. Therefore... The value can be determined based on the specific battery properties.

[0125]

[0126] in, The actual internal resistance at the current moment. This is the coefficient representing the effect of temperature on resistance.

[0127] This formula uses an exponential model, which conforms to the law describing the change of battery internal resistance with temperature as described by the Arrhenius equation. The higher, The resistance decreases because high temperatures increase the conductivity of the electrolyte and reduce the charge transfer impedance, while lower temperatures significantly increase the internal resistance, showing an inverse relationship, which is consistent with experimental observations that the battery's internal resistance increases sharply at low temperatures. It reflects the sensitivity of internal resistance to temperature changes, with a value range of [0.03, 0.08], and is related to the electrolyte characteristics.

[0128] The current battery health status is calculated using a multi-index fusion model, as shown in the following formula:

[0129]

[0130] in, This indicates the current battery health status. This is the actual response time at the current moment. For the initial capacity, The initial internal resistance, For standard response time, All data can be obtained from battery specifications and battery test reports. , .

[0131] Battery health status indicates the battery's current health relative to its initial state, typically ranging from 0% to 100% (100% represents a brand new battery). This indicator integrates three key aging metrics: capacity decay, internal resistance increase, and response time degradation, avoiding the limitations of relying on a single metric. Furthermore, this indicator is updated in real time. , , These parameters reflect the instantaneous state changes of a battery during fatigue applications. Capacity is the most direct indicator of battery aging (e.g., loss of active lithium, reduced usable lithium due to SEI film growth); lower capacity corresponds to worse state of equilibrium (SOH), and the two are positively correlated. Increased internal resistance reflects electrode / electrolyte interface degradation and increased contact impedance, directly affecting power output capability; higher internal resistance corresponds to worse SOH, and the two are inversely correlated. The ion diffusion rate of aged batteries decreases, leading to delayed charge and discharge response, which is crucial, especially in dynamic scenarios such as frequency modulation; longer response times result in worse SOH, and the two are negatively correlated.

[0132] The value range is [0.1, 0.2]. The value range is [0.2, 0.3]. The value range is [0.5, 0.7]. Capacity decay (such as lithium-ion loss and active material failure) is the most direct and stable characteristic of battery aging, and it is strongly correlated with cycle life (e.g., 80% capacity as the end of life). Experimental data shows that the end of life for the vast majority of batteries is determined by capacity decay. The highest weighting ensures that the State of Health (SOH) assessment results are closely linked to the actual usable energy of the battery. Internal resistance growth reflects battery power capability degradation, but its change is typically non-linear (rapid initial growth, then slowing down). Internal resistance has a significant impact on high-rate applications (such as frequency modulation and fast charging), but is less sensitive to energy-type applications (such as energy storage). With a moderate weight, ω3 supplements power performance information on top of capacity-dominated data, avoiding misjudgments caused by sudden changes in internal resistance. Response time variations typically occur in the mid-to-late stages of aging and are significantly affected by instantaneous factors such as temperature and SOC, resulting in high data noise. Its contribution to SOH is more auxiliary, used to capture early polarization phenomena before capacity and internal resistance become explicit. A smaller weight is assigned to ω3 to improve adaptability to dynamic operating conditions without excessively introducing noise.

[0133] In this embodiment, the dynamic allocation of charging and discharging power specifically includes

[0134] The battery's current health status is analyzed to dynamically adjust the maximum charge and discharge power. The calculation formula is as follows:

[0135]

[0136] in, To represent the maximum charging and discharging power at the current moment, Rated power, For safety, a redundancy safety margin is set for the maximum charging and discharging power, with a value range of [1.1, 1.3]. The influence coefficient has a value range of [0.1, 0.3].

[0137] The formula for power distribution during charging and discharging is:

[0138]

[0139]

[0140] in, Allocate power for charging at the current moment. Discharge power allocation for the current moment, This represents the current state of charge.

[0141] Charging power With remaining capacity (100) The charging power is directly proportional to the State of Charge (SOC), ensuring that the charging power automatically decreases at high SOC levels to avoid the risk of overcharging. Discharge power... The discharge power is proportional to the current SOC; at low SOC, the discharge power is limited to prevent over-discharge. This charging and discharging power distribution achieves automatic SOC balancing and extends battery cycle life. and The relationship is inversely proportional; the higher the SOC, the lower the charging power. and The SOC is directly proportional to the discharge power; the higher the SOC, the greater the discharge power. Many battery systems are equipped with a battery management system, which typically has a built-in SOC estimation function, allowing direct access to battery SOC data.

[0142] And the allocated charging and discharging power is limited:

[0143]

[0144]

[0145] in, The maximum charging power at the current moment. This represents the maximum discharge power at the current moment. Rated charging power, This is the rated discharge power.

[0146] Aging batteries are more prone to lithium plating in the high SOC range, requiring strict limits on charging power. An exponential relationship (1.5 power) amplifies the constraints as SOH decreases. The discharge process is less sensitive to aging, and linear limits are sufficient (e.g., when SOH = 80%, the discharge power is 80% of the rated power). Charging limits are stricter than discharging limits, consistent with the aging characteristics of lithium-ion batteries. and There is a positive correlation (non-linearity); the smaller the SOH, the stricter the charging power limit. and There is a positive correlation (linearity); the smaller the SOH (State of Harm), the proportionally lower the discharge power. Rated charging power and rated discharging power can be directly obtained from the battery's specifications.

[0147] By dynamically adjusting the actual capacity and internal resistance parameters of the battery and introducing a multi-index fusion health status assessment model, the operational reliability and lifespan management capabilities of the energy storage system under complex operating conditions are significantly improved. This is achieved through temperature compensation mechanisms (such as capacity correction coefficients). and internal resistance temperature coefficient Real-time correction of battery parameters solves the problem that traditional static models cannot adapt to environmental temperature fluctuations, especially in extreme high and low temperature scenarios, which can reduce capacity estimation errors; secondly, it innovatively incorporates the actual response time ( The SOH (State of Health) assessment system, together with capacity and internal resistance, forms a multi-dimensional health indicator that can more comprehensively reflect the performance degradation of the battery during frequent charge and discharge. Compared with a single capacity degradation indicator, it can predict the potential failure risk of the battery in advance. Finally, the dynamic power limiting algorithm based on SOH achieves a balance between battery aging and grid demand, which avoids accelerated aging caused by overload and maximizes the utilization of the battery's available capacity.

[0148] Step 5: Obtain the predicted power load for the next moment through a hybrid prediction model, collect the current grid frequency, analyze the impact of the current grid frequency, charging and discharging power, the power load for the next moment, and the current meteorological data on the battery capacity, predict the battery capacity for the next moment, and thus determine whether to execute this charging and discharging power allocation.

[0149] In this embodiment, obtaining the environmental impact index and the power impact index specifically includes:

[0150] Calculate the power impact index at the current moment:

[0151]

[0152] in, The current power impact index. To predict the electricity load at the next time step using a hybrid forecasting model, Based on rated load, Current charging power, The discharge power at the current moment. Based on rated power, The current grid frequency. The rated frequency of the power grid. To allow for frequency deviation (refer to the safe fluctuation range of the power grid frequency). These are the weighting coefficients for the corresponding items; the rated load benchmark and rated power benchmark can be obtained from the technical documents (such as datasheets and specifications) provided by the battery manufacturer, and the grid rated frequency can be obtained from the grid dispatching protocol.

[0153] The Power Impact Index quantifies the comprehensive impact of the current grid operating status on battery capacity, reflecting the pressure from three aspects: load demand, charging and discharging behavior, and frequency stability of the power system. A higher Power Impact Index value indicates stronger stress on the battery from the power environment. The technical effects of this index include predicting the potential capacity degradation risk of the battery at the next moment by calculating P(t) in real time, triggering early warnings or adjusting charging and discharging strategies; balancing grid demand and battery life, and avoiding excessive battery use during high-stress periods (such as sudden load increases or large frequency deviations). The load L(t+1) at the next moment directly determines the battery's expected charging and discharging demand. Higher loads may require greater power output from the battery, leading to accelerated capacity degradation. The predicted power load at the next moment is positively correlated with P(t); the higher the power load, the higher the Power Impact Index. The charging and discharging power difference reflects the battery's instantaneous energy throughput intensity. Frequent charging and discharging switching or high power fluctuations exacerbate battery fatigue. The charging and discharging power difference is positively correlated with P(t); the greater the power fluctuation, the higher the Power Impact Index. Grid frequency deviation. This reflects the degree of system imbalance. The greater the frequency deviates from the rated value, the greater the power demand of the battery for frequency regulation. The greater the frequency deviation, the higher the power impact index. The term is positively correlated with P(t).

[0154] In this embodiment, the rated load benchmark is set to 1000MW, the rated power benchmark is set to 500MW, the grid rated frequency is set to 50Hz, the allowable frequency error is set to 0.5Hz, and the weighting factor is... The values ​​were 0.4, 0.3, and 0.3 respectively. Twenty-five sets of data were collected to predict the next moment's power load, charging power, discharging power, and grid frequency for analysis. The specific collected data is shown in the table below:

[0155] Table 1: Statistical Table of Relevant Operational Data for Energy Storage Power Stations

[0156]

[0157] Reference Figures 3-5 As can be seen from the data in the table above, the greater the deviation between the predicted power load, charging power, and discharging power and the grid frequency at the next moment, the greater the corresponding power impact index, reflecting that the system may face problems such as overload, voltage fluctuation, or frequency instability at the current moment. The smaller the deviation between the predicted power load, charging power, and discharging power and the grid frequency at the next moment, the smaller the corresponding power impact index, reflecting that the system is in a low-risk state at the current moment.

[0158] Calculate the environmental impact index at the current moment:

[0159]

[0160] Where E(t) is the environmental impact index at the current moment, and e is the meteorological data index. This represents the total number of meteorological data types. The weight of the e-th type of meteorological data at the current moment, Let the value of the e-th type of meteorological data be the value at the current moment. The optimal value for the suitable range of the e-th type of meteorological data. The maximum value of the suitable range for the e-th type of meteorological data. This represents the minimum value of the suitable range for the e-th type of meteorological data;

[0161] The Environmental Impact Index (E(t)) quantifies the impact of current environmental factors on the operation of an energy storage power station. It considers the influence of multiple meteorological data points (such as temperature, humidity, and rainfall) on battery performance and overall system efficiency. A higher E(t) value indicates a more significant impact of current environmental conditions on the energy storage system, potentially negatively affecting battery charge / discharge efficiency, lifespan, and safety; conversely, a lower value indicates more suitable environmental conditions, which are beneficial to the operation of the energy storage system. Meteorological data are included as an independent variable. These are the fundamental data affecting the dependent variable E(t), determining whether environmental conditions are within a reasonable range. Changes in the independent variable directly affect the value of E(t). For example, if the temperature is higher than the optimal value, E(t) will increase, reflecting the adverse effects of the environment. The independent variable influences the numerator of E(t). This reflects the deviation between environmental data and ideal conditions, thus affecting the calculation results of the environmental impact index. Weighting This is used to adjust the relative importance of different environmental variables to E(t), thereby refining the impact assessment. When a certain environmental data... With the best value The greater the deviation, the higher the absolute value. The increase in environmental impact index E(t) leads to a rise in the value of the environmental impact index, indicating that the negative impact of the environment on batteries or energy storage systems is increasing.

[0162] The suitable temperature range is set at [0℃, 40℃], with an optimal value of 25℃. Most lithium-ion batteries exhibit best charge and discharge performance within this temperature range. Excessively low temperatures can decrease the electrochemical reaction rate, affecting battery capacity and charge / discharge efficiency; excessively high temperatures can cause overheating, increasing internal resistance, shortening battery life, and even triggering safety hazards (such as thermal runaway). 25°C is considered room temperature and is the standard condition for battery testing and verification. At this temperature, most batteries achieve optimal charging efficiency and discharge performance, ensuring battery health and longevity. The suitable humidity range is set at [20%, 80%], with an optimal value of 50%. A relative humidity range of 20% to 80% effectively avoids static electricity caused by excessive dryness or corrosion and short circuits caused by excessive moisture. High humidity environments can cause corrosion of the battery casing and electrical connections, affecting battery performance and safety. 50% humidity is considered a balance point, reducing the risk of corrosion without interfering with the battery's electrochemical reaction, helping to maintain device stability and reliability. The suitable rainfall range is set to [0, 5 mm / day], with the optimal value set to 0. Rainfall directly affects batteries and electrical equipment, potentially causing moisture damage, short circuits, or corrosion. Significant rainfall (especially intrusive rainfall) can cause direct physical damage to battery components. The suitable wind speed range is set to [0, 10 m / s], with the optimal value set to 3 m / s. Moderate wind speed helps dissipate heat from the battery, preventing performance degradation due to overheating. Excessive wind speed may cause equipment vibration and physical damage, affecting the stability and safety of the equipment. A wind speed of 3 m / s provides good heat dissipation without negatively impacting the equipment, making it a relatively safe and efficient operating wind speed.

[0163] This introduces dynamic weight adjustment, based on the following formula:

[0164]

[0165] in, This indicates the current battery health status. The mixing coefficient, .

[0166] The State of Health (SOH) value indicates the current health status of the battery and is typically used to assess its remaining lifespan and performance. Changes in SOH directly affect battery efficiency and safety. Indicating SOH and meteorological factors The partial derivative of this factor reflects the degree of influence of a specific meteorological factor on the battery's health status. If this value is positive, it indicates that the meteorological factor is beneficial to the state of health (SOH); if it is negative, it indicates that the environmental factor is detrimental to the SOH. This indicates the deviation between current environmental data and optimal values, reflecting the suitability of environmental conditions. The larger the deviation, the less ideal the environmental conditions. The deviation is standardized to ensure the value is between 0 and 1, facilitating comparison with calculations from other parts. A mixing coefficient controls the weighting ratio of the two parts, ranging from 0 to 1. This allows for flexible weighting of the SOH influence and the environmental factor influence in the formula. The second part of the formula... This ensures that the impact of environmental factors is fully considered, adapting to different environmental changes. This combination allows the formula to effectively reflect changes in battery health under various environmental conditions. The formula is intuitive and can theoretically explain the relationship between environmental factors and battery health. The combination of partial derivatives and standardized deviations clearly demonstrates how environmental changes affect the battery's state of health (SOH), consistent with physical principles.

[0167] In this embodiment, predicting the battery capacity at the next moment specifically includes:

[0168] A capacity decay model is constructed based on the current environmental impact index and power impact index to perform real-time battery capacity correction and predict the battery capacity at the next moment. The calculation formula is as follows:

[0169]

[0170] in, To predict the battery capacity at the next moment, This represents the actual capacity at the current moment. for The environmental compensation function at time t. for The power compensation function at time t. It is a time variable;

[0171] Represents the actual capacity at the current moment. This value, combining environmental and electrical stress cumulative effects, predicts the future usable capacity of the battery. It reflects the dynamic degradation trend of the battery in fatigue-prone applications, rather than its static capacity. This value can be used to identify risks of rapid capacity degradation in advance (such as high-temperature + high-load scenarios), triggering warnings or power limiting. High-temperature environments accelerate electrolyte decomposition and SEI film growth, while low-temperature environments hinder lithium-ion diffusion, exacerbating capacity degradation. The quantification of temperature correction for the degradation rate is as follows: 0.8 is used at high temperatures to mitigate the integral term, while 1.2 is used at low temperatures to amplify the integral term due to faster actual degradation. The power impact index reflects the combined stress of grid load, power fluctuations, and frequency deviations; a higher value indicates more severe battery operating conditions. P(t) is mapped to a nonlinear decay coefficient by the hyperbolic tangent function (tanh), when P(t) is low (e.g., close to 0). It is also close to 0, so the decay is negligible; however, when P(t) is high, such as greater than 1, It's also close to 1, indicating accelerated capacity decay. The integral term calculates the cumulative effect of environmental and power conditions within a past time window N, reflecting the historical dependence of capacity decay. The higher the current capacity, the higher the predicted future capacity, therefore... and They are positively correlated. and They are inversely correlated, especially under extreme temperatures. A deviation from 1 accelerates capacity decay; the greater the electrical stress P(t), the greater the capacity decay. The closer it is to 1, the faster the capacity decays.

[0172] The environmental compensation function is determined based on the environmental impact index value, as follows:

[0173]

[0174] in, The environmental sensitivity coefficient has a value range of [0.5, 1.5]. The preset environmental impact threshold, The value can be referenced from the 95th percentile of historical E(t);

[0175] The environmental compensation function directly reflects the linear positive correlation between environmental stress and capacity decay rate, consistent with experimental observations of the effects of most environmental factors (such as temperature, humidity, wind speed, and rainfall) on battery aging. When the environment deteriorates, the compensation amount increases proportionally, avoiding calibration difficulties caused by complex nonlinearities. The power compensation function is determined based on the current power impact index, as shown in the following formula:

[0176]

[0177] in, for The power impact index at any given time;

[0178] When P(t) is low (close to 0) The value is close to 0, consistent with experimental observations that small loads have no significant effect on capacity. When P(t) is under moderate stress, Approaching 0.5, it decays linearly. When P(t) is high, A value close to 1 indicates that the capacity decay rate tends to stabilize under high stress. The design avoids over-prediction by the linear model in high-stress regions. A nonlinear stress-degradation mapping is achieved through the tanh function, which better reflects the actual aging curve of the battery.

[0179] If the predicted battery capacity at the next moment is not lower than the preset battery capacity, the current charge / discharge power allocation will be executed normally. If the predicted battery capacity at the next moment is lower than the preset battery capacity, a red warning will be issued and the current charge / discharge power allocation will not be executed.

[0180] Please see Figure 2 The present invention also provides a decision system for optimizing the operation mode of an energy storage power station. This system is used to implement the aforementioned method for making decisions on optimizing the operation mode of an energy storage power station, and includes:

[0181] The data acquisition module is used to collect historical power data of the batteries in the energy storage power station and historical meteorological data of the location of the energy storage power station. The power data includes voltage, current, power load and battery temperature.

[0182] The hybrid prediction model construction module is used to preprocess the power load data in historical power data and perform spectrum analysis on the preprocessed power load data to obtain high-frequency and low-frequency components. The hybrid prediction model is constructed by combining LSTM-Attention and an improved ARIMAX model.

[0183] The parameter estimation module is used to estimate the internal resistance at the current moment based on the voltage and current data in the historical power data, using the recursive least squares method, and to obtain the estimated battery capacity at the current moment.

[0184] The health status analysis module is used to obtain the current battery temperature, analyze its impact on the battery capacity and internal resistance, correct the battery capacity and internal resistance, obtain the actual battery capacity and actual internal resistance at the current moment, and record the actual battery response time at the current moment. By comprehensively analyzing the actual battery capacity, actual internal resistance and actual response time, the battery health status is determined, and charging and discharging power is allocated according to the battery health status.

[0185] The comprehensive judgment module is used to obtain the predicted power load for the next moment through a hybrid prediction model, collect the current grid frequency, analyze the impact of the current grid frequency, charging and discharging power, the power load for the next moment, and the current meteorological data on the battery capacity, predict the battery capacity for the next moment, and thus determine whether to execute the current charging and discharging power allocation.

[0186] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0187] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0188] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0189] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A decision-making method for optimizing the operation mode of an energy storage power station, characterized in that, The specific steps include: Step 1: Collect historical power data of the batteries in the energy storage power station and historical meteorological data of the location of the energy storage power station. The power data includes voltage, current, power load and battery temperature. Step 2: Preprocess the power load data in the historical power data, and perform spectrum analysis on the preprocessed power load data to obtain high-frequency and low-frequency components. Combine LSTM-Attention and the improved ARIMAX model to construct a hybrid prediction model. Step 3: Based on the voltage and current data in the historical power data, estimate the internal resistance at the current moment using the recursive least squares method, and obtain the estimated battery capacity at the current moment. Step 4: Obtain the current battery temperature, analyze its impact on the battery capacity and internal resistance, correct the battery capacity and internal resistance to obtain the actual battery capacity and actual internal resistance at the current moment, and record the actual battery response time at the current moment. By comprehensively analyzing the actual battery capacity, actual internal resistance and actual response time, the battery health status is determined, and the charging and discharging power is allocated according to the battery health status. Step 5: Obtain the predicted power load for the next moment through a hybrid prediction model, collect the current grid frequency, analyze the impact of the current grid frequency, charging and discharging power, the power load for the next moment, and the current meteorological data on the battery capacity, predict the battery capacity for the next moment, and thus determine whether to execute this charging and discharging power allocation. Constructing the hybrid prediction model specifically includes: The power load data in the historical power data is traversed, and an improved Hample filter is used to correct outliers in the power load data. Dynamic standardization is then applied to standardize the power load data in the collected historical power data, as shown in the expression: in, For dynamic standardization Power load at any given time for Power load at any given time , The mean and standard deviation of electricity load in historical electricity data. It is a time variable; Spectral analysis is performed on the power load data after outlier correction to determine the number of decomposition layers. Then, wavelet decomposition is performed based on the standardized power load data and the number of decomposition layers to obtain high-frequency and low-frequency components. The high-frequency components are predicted using an LSTM-Attention model, and the low-frequency components are predicted using an improved ARIMAX model. Finally, the predicted high-frequency and low-frequency components are analyzed through adaptive reconstruction to obtain the predicted power load for the next time step.

2. The method for optimizing the operation mode of an energy storage power station according to claim 1, characterized in that, The collection of historical power data and historical meteorological data specifically includes: Voltage sensors are installed at the charging and discharging interfaces of the energy storage device to collect the charging and discharging voltages of the energy storage device. Current sensors are installed on the charging and discharging lines to collect the charging and discharging current data of the energy storage device. Temperature sensors are installed on the surface of the battery module of the energy storage device to collect the temperature data of the energy storage device. The meteorological data includes the ambient temperature, humidity, rainfall, and wind speed of the location of the energy storage power station.

3. The method for optimizing the operation mode of an energy storage power station according to claim 1, characterized in that, Obtaining the current internal resistance and battery capacity specifically includes: The internal resistance is estimated using the recursive least squares method, and the calculation formula is as follows: in, To estimate the internal resistance at the current moment, N is the length of the sliding window. for The current battery terminal voltage is applicable to both charging and discharging. for The battery current at any given time is positive when charging and negative when discharging. For a moment At the time The average voltage, For a moment At the time The average current is obtained, where t represents the current time; and the battery capacity at the current time is obtained as the estimated battery capacity.

4. The method for optimizing the operation mode of an energy storage power station according to claim 3, characterized in that, Determining the battery health status specifically includes Collect the battery's temperature data at the current moment, analyze the impact of temperature on the battery, and correct the battery's capacity and internal resistance using the following formula: in, This represents the actual capacity at the current moment. Let T(t) be the estimated battery capacity at the current moment, and T(t) be the battery temperature at the current moment. This is the coefficient representing the effect of temperature on battery capacity. The optimal operating temperature for the battery; in, The actual internal resistance at the current moment. This is the coefficient representing the effect of temperature on resistance. The current battery health status is calculated using a multi-index fusion model, as shown in the following formula: in, This indicates the current battery health status. This is the actual response time at the current moment. For the initial capacity, The initial internal resistance, For standard response time, , .

5. The method for optimizing the operation mode of an energy storage power station according to claim 4, characterized in that, The dynamic allocation of charging and discharging power specifically includes The battery's current health status is analyzed to dynamically adjust the maximum charge and discharge power. The calculation formula is as follows: in, To represent the maximum charging and discharging power at the current moment, Rated power, For safety reasons, This is the influence coefficient; The formula for power distribution during charging and discharging is: in, Allocate power for charging at the current moment. Discharge power allocation for the current moment, This represents the current state of charge. And limit the allocated charging and discharging power: in, The maximum charging power at the current moment. This represents the maximum discharge power at the current moment. Rated charging power, This is the rated discharge power.

6. The method for optimizing the operation mode of an energy storage power station according to claim 3, characterized in that, The acquisition of the environmental impact index and the power impact index specifically includes: Calculate the power impact index at the current moment: in, The current power impact index. To predict the electricity load at the next time step using a hybrid forecasting model, Based on rated load, Current charging power, The discharge power at the current moment. Based on rated power, The current grid frequency. The rated frequency of the power grid. To allow for frequency deviation, These are the weight coefficients for the corresponding items; Calculate the environmental impact index at the current moment: Where E(t) is the environmental impact index at the current moment, and e is the meteorological data index. This represents the total number of meteorological data types. The weight of the e-th type of meteorological data at the current moment, Let the value of the e-th type of meteorological data be the value at the current moment. The optimal value for the suitable range of the e-th type of meteorological data. The maximum value of the suitable range for the e-th type of meteorological data. This represents the minimum value of the suitable range for the e-th type of meteorological data; This introduces dynamic weight adjustment, based on the following formula: in, This indicates the current battery health status. This is the mixing coefficient.

7. The method for optimizing the operation mode of an energy storage power station according to claim 6, characterized in that, The prediction of the battery capacity at the next moment specifically includes: A capacity decay model is constructed based on the current environmental impact index and power impact index to perform real-time battery capacity correction and predict the battery capacity at the next moment. The calculation formula is as follows: in, To predict the battery capacity at the next moment, This represents the actual capacity at the current moment. for The environmental compensation function at time t. for The power compensation function at time t. It is a time variable; The environmental compensation function is determined based on the environmental impact index, as follows: in, Environmental sensitivity coefficient, The preset environmental impact threshold; The power compensation function is determined based on the power impact index at the current moment, as shown in the following formula: in, for The power impact index at any given time; If the predicted battery capacity at the next moment is not lower than the preset battery capacity, the current charge / discharge power allocation will be executed normally. If the predicted battery capacity at the next moment is lower than the preset battery capacity, a red warning will be issued and the current charge / discharge power allocation will not be executed.

8. A decision-making system for optimizing the operation mode of an energy storage power station, characterized in that, The energy storage power station optimized operation mode decision system is used to implement the energy storage power station optimized operation mode decision method according to any one of claims 1-7, including: The data acquisition module is used to collect historical power data of the batteries in the energy storage power station and historical meteorological data of the location of the energy storage power station. The power data includes voltage, current, power load and battery temperature. The hybrid prediction model construction module is used to preprocess the power load data in historical power data and perform spectrum analysis on the preprocessed power load data to obtain high-frequency and low-frequency components. The hybrid prediction model is constructed by combining LSTM-Attention and an improved ARIMAX model. The parameter estimation module is used to estimate the internal resistance at the current moment based on the voltage and current data in the historical power data, using the recursive least squares method, and to obtain the estimated battery capacity at the current moment. The health status analysis module is used to obtain the current battery temperature, analyze its impact on the battery capacity and internal resistance, correct the battery capacity and internal resistance, obtain the actual battery capacity and actual internal resistance at the current moment, and record the actual battery response time at the current moment. By comprehensively analyzing the actual battery capacity, actual internal resistance and actual response time, the battery health status is determined, and charging and discharging power is allocated according to the battery health status. The comprehensive judgment module is used to obtain the predicted power load for the next moment through a hybrid prediction model, collect the current grid frequency, analyze the impact of the current grid frequency, charging and discharging power, the power load for the next moment, and the current meteorological data on the battery capacity, predict the battery capacity for the next moment, and thus determine whether to execute the current charging and discharging power allocation.

Images