Fault prediction and optimization method for intelligent energy storage system based on artificial intelligence

Abstract

The invention discloses an intelligent energy storage system fault prediction and optimization method based on artificial intelligence, and the method comprises the steps: determining a plurality of wind power plants which are correspondingly connected to an energy storage system, and the historical operation data, high-precision weather forecast data and terrain geographic information of each wind power plant, and constructing a time-space database; constructing a WRF-CFD coupling model embedded based on a neural network prediction model to predict the power generation efficiency of each wind power plant to obtain a second type of prediction result; calculating the total load of the energy storage system based on the second type of prediction results, and predicting the fault risk according to the total load fluctuation rate; and when the fault risk meets a preset condition, generating an advanced coping scheme to optimize and flatten the load fluctuation of the energy storage system. According to the method, accurate prediction and optimal control of the intelligent energy storage system can be realized, the problems that fault early warning is not timely and load fluctuation is difficult to effectively adjust when a traditional energy storage system faces fluctuating power generation of a wind power plant are solved, and the reliability and the operation efficiency of the energy storage system are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of wind power intelligent energy storage systems, and specifically relates to a fault prediction and optimization method for intelligent energy storage systems based on artificial intelligence. Background Art

[0002] As the global energy mix accelerates its transition toward renewable energy, the fluctuating nature of wind farms, as a core clean energy source, poses a significant challenge to grid stability. Intelligent energy storage systems, which achieve spatial and temporal shifting of electrical energy through charge and discharge regulation, have become critical infrastructure for smoothing wind power fluctuations and ensuring grid stability. However, sudden changes in wind farm output and frequent load fluctuations can easily lead to energy storage system overloads and accelerated battery degradation. Traditional fault warning and control methods, lacking dynamic predictive capabilities, struggle to meet high reliability requirements. Intelligent technologies are urgently needed to overcome bottlenecks in energy storage system operational efficiency and lifespan management.

[0003] Current energy storage system fault prediction relies heavily on threshold alarms or statistical analysis based on historical data. Threshold alarms trigger warnings by setting fixed thresholds for parameters like voltage and temperature. However, these methods fail to capture the gradual degradation of the battery's internal state (such as SEI film growth and lithium dendrite precipitation), resulting in a high rate of missed fault notifications. Statistical analysis relies on offline modeling, making it difficult to adapt to dynamic wind farm scenarios in real time and resulting in significant warning lag. Regarding optimal control, traditional methods employ rule-based or PID control strategies. While these methods can handle steady-state loads, they lack a joint modeling of the energy storage unit's health and grid demand when wind farm output fluctuates, leading to regulation oscillations and overcharge / over-discharge risks. For example, when wind power output suddenly drops, fixed-rule charge and discharge strategies may overlook the remaining battery life, forcing peak load regulation, leading to accelerated battery aging and even thermal runaway. The limited adaptability of these methods in dynamic scenarios has become a major obstacle to the reliability and economic viability of energy storage systems.

[0004] To address the above-mentioned shortcomings, there is an urgent need for a dynamic prediction and optimization technology that integrates artificial intelligence. By real-time sensing of the multi-dimensional status data of the energy storage system (such as electrochemical impedance, temperature distribution, and state of charge), combined with wind power output forecasts and grid demand information, a joint optimization model of fault risk and regulation efficiency can be constructed, thereby achieving accurate fault warning and adaptive control. Summary of the Invention

[0005] In view of this, the present invention proposes an artificial intelligence-based intelligent energy storage system fault prediction and optimization method. Through the present invention, accurate prediction and optimized control of the intelligent energy storage system can be achieved, solving the problems of traditional energy storage systems such as untimely fault warnings and difficulty in effectively adjusting load fluctuations when facing fluctuating power generation in wind farms, thereby greatly improving the reliability and operating efficiency of the energy storage system.

[0006] In order to achieve the above object, the present invention provides the following technical solutions: The present invention provides an artificial intelligence-based intelligent energy storage system fault prediction and optimization method, comprising: Step S1: Determine the multiple wind farms connected to the energy storage system, as well as the historical operating data, high-precision weather forecast data, and topographic information of each wind farm, and build a spatiotemporal database; Step S2: constructing a WRF-CFD coupling model embedded in a neural network prediction model based on the spatiotemporal database to predict the power generation efficiency of each wind farm and obtain a second type of prediction result; Step S3: Calculate the total load of the energy storage system based on the second type of prediction results, and predict the failure risk based on the total load fluctuation rate; Step S4: When the failure risk meets the preset conditions, an advance response plan is generated to optimize and smooth the load fluctuation of the energy storage system.

[0007] Preferably, step S2 specifically includes: Step S21: Based on the high-precision weather forecast data and topographic information in the spatiotemporal database, a WRF-CFD coupling model is constructed to perform a gradient-based hierarchical prediction of the wind force and wind direction distribution in any wind farm to obtain a first-class prediction result; Step S22: pre-constructing and training a neural network model for predicting the power generation efficiency of a single wind turbine generator set in the wind farm based on the spatiotemporal database, and using the first type of prediction results to predict the power generation efficiency of the wind turbine generator set; Step S23: Calculate the downstream wind speed and direction of the wind turbine generator using the momentum theory method to obtain a calculation result, and return the calculation result to step S21 to perform real-time correction on the first type of prediction result; Step S24: Based on the predicted power generation efficiency of each wind turbine generator set, statistics are collected on the power generation efficiency of the plurality of wind farms connected to the energy storage system to obtain a second type of prediction result.

[0008] Preferably, step S21 specifically includes: Based on the high-precision meteorological forecast data in the spatiotemporal database, the temperature field, pressure gradient, and humidity field data are determined and regional dynamic downscaling is performed using the constructed WRF model to generate initial wind field data of 1 km grid. The terrain information within the preset range around the wind farm is collected in advance through synthetic aperture radar images, and the surface roughness parameters are calculated using the following formula: in, is the surface roughness parameter, Indicates the vertical height of the surface obstacle. Represents the surface coverage. The surface coverage parameter is determined by the type of surface cover. Each type of surface cover corresponds to a unique surface coverage parameter. The obtained surface roughness parameters are input into the CFD model to complete the 10m-level micro-meteorological modeling within the wind farm; The wind farm is graded according to the wind direction, and the wind force and direction distribution in the outermost wind farm are predicted based on 10m-level micro-meteorological modeling to obtain the third type of prediction results. Waiting for the returned calculation results to be cyclically predicted for the wind force and wind direction distribution within the next gradient to obtain multiple third-category prediction results until the wind force and wind direction distribution within multiple wind farm gradients are predicted; The obtained multiple third-category prediction results are counted and integrated to obtain the first-category prediction results.

[0009] Preferably, step S22 includes: Determine the historical operating data of each wind turbine based on the spatiotemporal database, including external wind conditions and internal states. External wind conditions include wind speed, wind direction, and turbulence intensity, and internal states include pitch angle, speed, temperature, and power output. Data cleaning removes invalid periods of wind speed below the cut-in speed or above the cut-out speed, and fills missing values ​​with interpolation based on physical constraints. Key features, including wind energy density, pitch efficiency factor, and sliding window statistics, are then constructed to complete time series alignment and normalization. A hybrid neural network model with a bidirectional LSTM and a temporal convolutional network is used. The LSTM layer captures long- and short-term wind speed time series characteristics. The TCN uses dilated convolution to expand the receptive field to identify periodic patterns. The output layer embeds the static parameters of the turbine unit into a vector, and the fully connected layer uses regression to predict power generation efficiency. The static parameters of the turbine unit include rotor diameter and rated power. During training, the Smooth L1 loss function is used to reduce outlier interference. The AdamW optimizer and cosine annealing learning rate scheduling are used together with the introduction of time series dropout and label smoothing to improve the generalization ability of the model. During the verification phase, the data set was divided by season to avoid climate model leakage. The test covered unit operating conditions under three typical wind conditions: steady state, turbulent, and cut-out. The goal was to control the mean absolute error within 3% of the rated power. Finally, a neural network model for predicting the power generation efficiency of a single wind turbine in a wind farm is trained. The external wind conditions and internal states faced by the wind turbine in the first type of prediction results are used as model inputs to predict the power generation efficiency of the wind turbine and obtain the prediction results.

[0010] Preferably, step S23 includes: The improved Jensen wake model is used to calculate the velocity attenuation and turbulence enhancement after the wind passes through the wind turbine. The conservation of wind energy and momentum change are determined based on the momentum theory method, and the interference effect of the upstream wind turbine on the downstream wind field is quantified. The velocity attenuation formula is expressed as: in, represents the wind speed at the hub height of the downstream wind turbine, represents the wind speed at the hub height of the upstream wind turbine, is the angle between the wind direction and the line connecting the two units, is the wake overlap area, is the impeller swept area, Indicates the impeller diameter; The turbulence enhancement formula is expressed as: in, represents the downstream turbulence intensity, represents the upstream turbulence intensity; The wind speed at the hub height of the downstream wind turbine and downstream turbulence intensity The uniform wind speed and turbulence intensity in the direction of the natural wind after it passes through the wind turbine are calculated and the results are generated.

[0011] Preferably, step S3 includes: The total load calculation process determines the predicted power generation value of the i-th wind farm at the future time t based on the second type of prediction results , based on historical data statistics, predict the power demand of the power grid at time t , combined with the power generation forecast value to calculate the total load of the energy storage system at time t : in, Indicates the total number of wind farms that input electricity into the energy storage system; when When the energy storage system enters the charging mode, the charging power , Indicates the maximum charging power of the energy storage system; when When the energy storage system enters the discharge mode, the discharge power , Indicates the maximum discharge power of the energy storage system; The total load fluctuation calculation process calculates the range of load power within the statistical period : The main frequency component is extracted through Fourier transform, and the high frequency (>0.1Hz) energy ratio is calculated: in, represents the fast Fourier transform, represents the sum of high-frequency energy, Represents the energy density of different frequency f components in the signal, Represents the total energy of the entire frequency band, It represents the proportion of high-frequency energy. A value close to 1 indicates that the power fluctuation is mainly dominated by high-frequency components, while a value close to 0 indicates that the fluctuation is gentle. Calculating the Composite Volatility Index : in, Indicates the average load power during the period, Indicates the total number of time points in the time series; Construct a failure risk prediction model and use the comprehensive volatility index , battery health parameters and timing correlation characteristics to predict the failure risk of the energy storage system and obtain the prediction results.

[0012] Preferably, step S4 includes: When the fault risk meets the preset conditions, an advance response plan is generated to optimize the energy storage system load fluctuations, including limiting the charging power when the overall SOC of the energy storage system is predicted to approach a preset upper limit: in, Indicates the rated capacity of the energy storage system under standard conditions. Indicates the preset upper limit value. represents the overall SOC of the energy storage system at time t, Indicates the duration of the charging and discharging process; Limit the power change rate by sliding window: in, Indicates the maximum allowable power change rate, is the power value at time t.

[0013] Preferably, the fault risk prediction model predicts the fault risk of the energy storage system, and the prediction results include: Pre-build a hybrid model composed of gradient boosting tree GBDT or LSTM-attention mechanism to calculate the probability of failure of the energy storage system in the next 24 hours. As the target output, the load characteristics, battery health parameters, and time series correlation characteristics in the historical data are used as input for training to obtain a fault risk prediction model, where: Load characteristics include comprehensive volatility indicators and the number of charge and discharge cycles of the batteries in the energy storage system : Battery health parameters include internal resistance growth rate and battery temperature gradient : in, represents the internal resistance of the battery at time t, express The internal resistance of the battery at all times, represents the temperature of the i-th battery cell; The time series correlation feature includes the statistically obtained high fluctuation duration of the continuous total load fluctuation.

[0014] The present invention has achieved at least the following beneficial effects: 1. It enables accurate prediction and optimized control of intelligent energy storage systems, resolving the issues of traditional energy storage systems such as delayed fault warnings and difficulty in effectively adjusting load fluctuations when facing fluctuating wind farm power generation, significantly improving the reliability and operational efficiency of energy storage systems.

[0015] 2. It can achieve high-precision prediction and real-time correction of wind farm power generation efficiency, solving the problem of wind power prediction accuracy under complex terrain and meteorological conditions. At the same time, it improves the fault prediction and optimization capabilities of the energy storage system, and enhances the operating efficiency and reliability of the entire intelligent energy storage system.

[0016] Other advantages, objectives and features of the present invention will be described in the following description and will be apparent to those skilled in the art to some extent, or those skilled in the art can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to make the purpose, technical solutions and beneficial effects of the present invention more clear, the present invention provides the following drawings for illustration: Figure 1 This is a flowchart of a method for predicting and optimizing faults of an intelligent energy storage system based on artificial intelligence in an embodiment of the present invention; Figure 2 This is a flow chart of the steps for predicting the power generation efficiency of each wind farm in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0019] The present invention provides an artificial intelligence-based intelligent energy storage system fault prediction and optimization method, referring to Figure 1 ,include: Step S1: Determine the multiple wind farms connected to the energy storage system, as well as the historical operating data, high-precision weather forecast data, and topographic information of each wind farm, and build a spatiotemporal database; Step S2: constructing a WRF-CFD coupling model embedded in a neural network prediction model based on the spatiotemporal database to predict the power generation efficiency of each wind farm and obtain a second type of prediction result; Step S3: Calculate the total load of the energy storage system based on the second type of prediction results, and predict the failure risk based on the total load fluctuation rate; Step S4: When the failure risk meets the preset conditions, an advance response plan is generated to optimize and smooth the load fluctuation of the energy storage system.

[0020] The working principle and beneficial effects of the above technical solution are as follows: Step S1 collects historical operating data, high-precision weather forecasts, and topographic information from each wind farm and constructs a spatiotemporal database, providing a comprehensive data foundation for subsequent predictions and enabling efficient data integration and management. Step S2 utilizes a WRF-CFD coupling model embedded in a neural network prediction model to accurately predict the power generation efficiency of each wind farm, resulting in a second-category prediction result. This not only improves prediction accuracy but also provides critical data support for the optimized operation of the energy storage system. Step S3 calculates the total load of the energy storage system based on the prediction results and analyzes the load fluctuation rate to predict failure risks, enabling real-time monitoring of the energy storage system's operating status and risk warnings. When the failure risk reaches a preset condition, step S4 generates an advance response plan to optimize and smooth the energy storage system's load fluctuations, effectively reducing the probability of failure and ensuring the stable operation of the energy storage system. This technical solution enables accurate prediction and optimized control of the intelligent energy storage system, resolving the problems of traditional energy storage systems facing fluctuating wind farm power generation, such as delayed fault warnings and difficulty in effectively adjusting load fluctuations, significantly improving the reliability and operational efficiency of the energy storage system.

[0021] In a specific embodiment, referring to Figure 2 , step S2 specifically includes: Step S21: Based on the high-precision weather forecast data and topographic information in the spatiotemporal database, a WRF-CFD coupling model is constructed to perform a gradient-based hierarchical prediction of the wind force and wind direction distribution in any wind farm to obtain a first-class prediction result; Step S22: pre-constructing and training a neural network model for predicting the power generation efficiency of a single wind turbine generator set in the wind farm based on the spatiotemporal database, and using the first type of prediction results to predict the power generation efficiency of the wind turbine generator set; Step S23: Calculate the downstream wind speed and direction of the wind turbine generator using the momentum theory method to obtain a calculation result, and return the calculation result to step S21 to perform real-time correction on the first type of prediction result; Step S24: Based on the predicted power generation efficiency of each wind turbine generator set, statistics are collected on the power generation efficiency of the plurality of wind farms connected to the energy storage system to obtain a second type of prediction result.

[0022] The working principle and beneficial effects of the above technical solution are as follows: through step S21, based on the high-precision meteorological forecast data and topographic information in the spatiotemporal database, a WRF-CFD coupling model is constructed to perform gradient-based hierarchical predictions on the wind force and wind direction distribution in any wind farm, and obtain the first type of prediction results, thereby achieving fine modeling of the wind farm characteristics and providing high-precision meteorological data support for subsequent power generation efficiency predictions. Through step S22, a neural network model for predicting the power generation efficiency of a single wind turbine in a wind farm is pre-constructed and trained, and the first type of prediction results are used to predict the power generation efficiency of the wind turbine, thereby achieving an accurate prediction of the power generation efficiency and providing a key data basis for the optimized operation of the energy storage system. In step S23, the momentum theory method is used to calculate the downstream wind speed and direction of the wind turbine, and the calculation results are returned to step S21 to make real-time corrections to the first type of prediction results, thereby forming a closed-loop feedback mechanism that can dynamically adjust and optimize the accuracy of wind farm predictions, thereby improving the adaptability and reliability of the entire prediction system. Finally, in step S24, based on the predicted power generation efficiency of each wind turbine, the power generation efficiency of the multiple wind farms connected to the energy storage system is statistically analyzed to obtain a second type of prediction result. This provides comprehensive power generation efficiency data for the scheduling and optimization of the energy storage system, facilitating the advance planning of the energy storage system's charging and discharging strategies and ensuring the stable operation of the power grid. This technical solution enables high-precision prediction and real-time correction of wind farm power generation efficiency, solving the problem of wind power prediction accuracy under complex terrain and meteorological conditions. It also improves the energy storage system's fault prediction and optimization capabilities, enhancing the operating efficiency and reliability of the entire intelligent energy storage system.

[0023] In a specific embodiment, step S21 specifically includes: Based on the high-precision meteorological forecast data in the spatiotemporal database, the temperature field, pressure gradient, and humidity field data are determined and regional dynamic downscaling is performed using the constructed WRF model to generate initial wind field data of 1 km grid. The terrain information within the preset range around the wind farm is collected in advance through synthetic aperture radar images, and the surface roughness parameters are calculated using the following formula: in, is the surface roughness parameter, Indicates the vertical height of the surface obstacle. Represents the surface coverage. The surface coverage parameter is determined by the type of surface cover. Each type of surface cover corresponds to a unique surface coverage parameter. The obtained surface roughness parameters are input into the CFD model to complete the 10m-level micro-meteorological modeling within the wind farm; The wind farm is graded according to the wind direction, and the wind force and direction distribution in the outermost wind farm are predicted based on 10m-level micro-meteorological modeling to obtain the third type of prediction results. Waiting for the returned calculation results to be cyclically predicted for the wind force and wind direction distribution within the next gradient to obtain multiple third-category prediction results until the wind force and wind direction distribution within multiple wind farm gradients are predicted; The obtained multiple third-category prediction results are counted and integrated to obtain the first-category prediction results.

[0024] The working principle and beneficial effects of the above technical solution are as follows: First, in step S21, high-precision meteorological forecast data from a spatiotemporal database is used to extract temperature, pressure gradient, and humidity field information. This data is then processed using the WRF model for regional dynamic downscaling to generate initial wind field data for a 1km grid, providing basic meteorological data support for subsequent refined wind field predictions. Next, topographic information surrounding the wind farm is acquired using synthetic aperture radar imagery, and a surface roughness parameter is calculated using a given formula. The surface roughness parameter comprehensively considers the vertical height and coverage of surface obstacles. Different surface cover types correspond to unique coverage parameters, accurately reflecting the impact of wind farm topographic characteristics on the wind farm. The resulting surface roughness parameter is input into the CFD model, completing 10m-level micrometeorological modeling within the wind farm and enabling a detailed simulation of the microscale wind field within the wind farm. Subsequently, the wind turbines within the wind farm are graded according to wind direction, and the wind force and direction distribution of the outermost wind farm is predicted based on the 10m-level micrometeorological model, resulting in preliminary third-category prediction results. During the prediction process, the system waits in a loop to receive the calculated results of downstream wind speed and direction, and based on this, it corrects the wind force and direction predictions within the subsequent gradients in real time until the predictions for all gradients are completed, generating multiple third-category prediction results. Finally, these results are statistically integrated to obtain comprehensive first-category prediction results, providing a high-precision wind farm data foundation for the power generation efficiency prediction of the entire wind farm. This technical solution enables refined modeling and prediction of wind farm wind farms, effectively improving the accuracy of wind turbine power generation efficiency predictions and enhancing the reliability of energy storage system fault prediction and optimization.

[0025] In a specific embodiment, step S22 includes: Determine the historical operating data of each wind turbine based on the spatiotemporal database, including external wind conditions and internal states. External wind conditions include wind speed, wind direction, and turbulence intensity, and internal states include pitch angle, speed, temperature, and power output. Data cleaning removes invalid periods of wind speed below the cut-in speed or above the cut-out speed, and fills missing values ​​with interpolation based on physical constraints. Key features, including wind energy density, pitch efficiency factor, and sliding window statistics, are then constructed to complete time series alignment and normalization. A hybrid neural network model with a bidirectional LSTM and a temporal convolutional network is used. The LSTM layer captures long- and short-term wind speed time series characteristics. The TCN uses dilated convolution to expand the receptive field to identify periodic patterns. The output layer embeds the static parameters of the turbine unit into a vector, and the fully connected layer uses regression to predict power generation efficiency. The static parameters of the turbine unit include rotor diameter and rated power. During training, the Smooth L1 loss function is used to reduce outlier interference. The AdamW optimizer and cosine annealing learning rate scheduling are used together with the introduction of time series dropout and label smoothing to improve the generalization ability of the model. During the verification phase, the data set was divided by season to avoid climate model leakage. The test covered unit operating conditions under three typical wind conditions: steady state, turbulent, and cut-out. The goal was to control the mean absolute error within 3% of the rated power. Finally, a neural network model for predicting the power generation efficiency of a single wind turbine in a wind farm is trained. The external wind conditions and internal states faced by the wind turbine in the first type of prediction results are used as model inputs to predict the power generation efficiency of the wind turbine and obtain the prediction results.

[0026] The working principle and beneficial effects of the above technical solution are as follows: First, through step S22, historical operating data for each wind turbine is determined from the spatiotemporal database, covering external wind conditions (wind speed, wind direction, turbulence intensity) and internal states (pitch angle, rotational speed, temperature, and power output), providing a comprehensive data foundation for model training. Next, data cleaning is performed to remove invalid records below the cut-in wind speed or above the cut-out wind speed. Missing values ​​are filled using a physically constrained interpolation method to ensure data integrity and accuracy. Key features, including wind energy density, pitch efficiency factor, and sliding window statistics, are constructed, and time series alignment and normalization are performed. This extracts features that are important for power generation efficiency prediction and improves the quality of model input data.

[0027] The model architecture utilizes a hybrid neural network model combining bidirectional LSTM and a temporal convolutional network (TCN). The LSTM layer effectively captures the long- and short-term temporal characteristics of wind speed, while the TCN, through dilated convolutions, expands the receptive field and accurately identifies periodic patterns. The combination of these two layers takes into account both the long-term dependencies and periodic characteristics of time series data. The output layer incorporates the embedded vector of the turbine's static parameters, and a fully connected layer implements regression prediction of power generation efficiency, fully accounting for the differences in static characteristics of different wind turbines.

[0028] During training, the Smooth L1 loss function was used to effectively reduce outlier interference. Combining the AdamW optimizer with cosine annealing learning rate scheduling improved the model's convergence speed and stability. The introduction of time-series dropout and label smoothing significantly enhanced the model's generalization capabilities, enabling it to maintain good forecasting performance under varying wind conditions.

[0029] During the validation phase, the dataset was divided by season to effectively prevent climate pattern leakage and ensure the model's applicability across different seasonal scenarios. Testing covered turbine operating conditions under three typical wind conditions: steady-state, turbulent, and cut-out. This comprehensively evaluated the model's performance under diverse wind conditions, with the goal of keeping the mean absolute error within 3% of rated power, thereby ensuring high accuracy in model predictions.

[0030] Ultimately, the trained neural network model for predicting the power generation efficiency of a single wind turbine within a wind farm uses the external wind conditions and internal state of the wind turbine in the first category of predictions as model inputs, accurately predicting its power generation efficiency and providing reliable data support for the optimized scheduling of the energy storage system. This technical solution achieves high-precision predictions of wind turbine power generation efficiency, solving the challenge of power generation efficiency prediction under complex wind conditions and improving the operational efficiency and reliability of the entire intelligent energy storage system.

[0031] In a specific embodiment, step S23 includes: The improved Jensen wake model is used to calculate the velocity attenuation and turbulence enhancement after the wind passes through the wind turbine. The conservation of wind energy and momentum change are determined based on the momentum theory method, and the interference effect of the upstream wind turbine on the downstream wind field is quantified. The velocity attenuation formula is expressed as: in, represents the wind speed at the hub height of the downstream wind turbine, represents the wind speed at the hub height of the upstream wind turbine, is the angle between the wind direction and the line connecting the two units, is the wake overlap area, is the impeller swept area, Indicates the impeller diameter; The turbulence enhancement formula is expressed as: in, represents the downstream turbulence intensity, represents the upstream turbulence intensity; The wind speed at the hub height of the downstream wind turbine and downstream turbulence intensity The uniform wind speed and turbulence intensity in the direction of the natural wind after it passes through the wind turbine are calculated and the results are generated.

[0032] The working principle and beneficial effects of the above technical solution are as follows: the use of the improved Jensen wake model can accurately calculate the speed attenuation and turbulence enhancement after the wind passes through the wind turbine, thereby quantifying the interference effect of the upstream wind turbine on the downstream wind field. Based on the momentum theory method, the conservation of wind energy and momentum change are determined. Specifically, the speed attenuation formula specifically calculates the speed change of the wind after passing through the wind turbine through parameters such as the wind speed at the hub height of the downstream wind turbine, the wind speed at the hub height of the upstream wind turbine, the angle between the wind direction and the line connecting the two units, the wake overlap area, and the impeller swept area. Similarly, the turbulence enhancement formula quantifies the impact of the wake on the turbulence intensity by comparing the downstream turbulence intensity with the upstream turbulence intensity. The downstream wind turbine hub height wind speed and the downstream turbulence intensity are used as the uniform wind speed and turbulence intensity in the direction of the natural wind after passing through the wind turbine generator set, and the corresponding calculation results are generated, and the process returns to step S21 to make real-time corrections to the first type of prediction results. This technical solution enables accurate modeling and real-time correction of the wake effect within a wind farm, thereby effectively improving the accuracy of the entire wind farm's power generation efficiency prediction, optimizing the energy storage system's operating strategy, and enhancing its adaptability to wind power fluctuations.

[0033] In a specific embodiment, step S3 includes: The total load calculation process determines the predicted power generation value of the i-th wind farm at the future time t based on the second type of prediction results , based on historical data statistics, predict the power demand of the power grid at time t , combined with the power generation forecast value to calculate the total load of the energy storage system at time t : in, Indicates the total number of wind farms that input electricity into the energy storage system; when When the energy storage system enters the charging mode, the charging power , Indicates the maximum charging power of the energy storage system; when When the energy storage system enters the discharge mode, the discharge power , Indicates the maximum discharge power of the energy storage system; The total load fluctuation calculation process calculates the range of load power within the statistical period : The main frequency component is extracted through Fourier transform, and the high frequency (>0.1Hz) energy ratio is calculated: in, represents the fast Fourier transform, represents the sum of high-frequency energy, Represents the energy density of different frequency f components in the signal, Represents the total energy of the entire frequency band, It represents the proportion of high-frequency energy. A value close to 1 indicates that the power fluctuation is mainly dominated by high-frequency components, while a value close to 0 indicates that the fluctuation is gentle. Calculating the Composite Volatility Index : in, Indicates the average load power during the period, Indicates the total number of time points in the time series; Construct a failure risk prediction model and use the comprehensive volatility index , battery health parameters and timing correlation characteristics to predict the failure risk of the energy storage system and obtain the prediction results.

[0034] The working principle and beneficial effects of the above technical solution are as follows: Based on the second-category prediction results, the predicted power generation value for each wind farm at a future time is determined. Combined with the grid's power demand forecast, the total load of the energy storage system at that moment is accurately calculated. When the generated power exceeds the power demand, the energy storage system enters charging mode, with the charging power limited by the maximum charging power; otherwise, it enters discharging mode, with the discharge power limited by the maximum discharge power. The total load fluctuation rate calculation process comprehensively evaluates the load fluctuation characteristics by statistically analyzing the load power range and extracting the high-frequency energy ratio using a Fourier transform. In particular, the high-frequency energy ratio indicator clearly indicates the dominant frequency component of power fluctuations. A value close to 1 indicates that the fluctuation is primarily driven by high-frequency components, while a value close to 0 indicates that the fluctuation is relatively gentle. Finally, a fault risk prediction model is constructed that integrates comprehensive fluctuation rate indicators, battery health parameters, and time-series correlation characteristics to accurately predict the failure risk of the energy storage system. The working principle of the entire technical solution is to comprehensively monitor and evaluate the load conditions and fluctuation characteristics of the energy storage system, and timely predict potential failure risks, thereby achieving refined management and optimized scheduling of the energy storage system, effectively improving the operational stability and reliability of the energy storage system, and reducing the probability of failure.

[0035] In a specific embodiment, step S4 includes: When the fault risk meets the preset conditions, an advance response plan is generated to optimize the energy storage system load fluctuations, including limiting the charging power when the overall SOC of the energy storage system is predicted to approach a preset upper limit: in, Indicates the rated capacity of the energy storage system under standard conditions. Indicates the preset upper limit value. represents the overall SOC of the energy storage system at time t, Indicates the duration of the charging and discharging process; Limit the power change rate by sliding window: in, Indicates the maximum allowable power change rate, is the power value at time t.

[0036] The working principle and beneficial effects of the above technical solution are as follows: The advance response plan generated by step S4 can timely optimize the load fluctuations of the energy storage system when the failure risk meets the preset conditions. When the overall SOC of the energy storage system is predicted to be close to the preset upper limit, the charging power is limited to avoid overcharging of the energy storage system and ensure its safe operation. The specific measures for limiting the charging power are determined based on the rated capacity of the energy storage system under standard conditions, the preset upper limit, the current SOC, and the charge and discharge time, thereby achieving precise power control.

[0037] At the same time, a sliding window limits the rate of power change, effectively smoothing power output and preventing sudden power changes from impacting the system. The maximum allowable power rate of change, combined with the current power value, ensures that the energy storage system's output power fluctuates within a reasonable range, improving system stability.

[0038] This optimization approach works by monitoring the energy storage system's status parameters in real time and dynamically adjusting the charging and discharging strategies and power output based on pre-set conditions, effectively managing load fluctuations and reducing the risk of failures. This significantly improves the energy storage system's operational safety, extends battery life, and enhances the system's adaptability to grid fluctuations, ensuring a stable and reliable power supply.

[0039] In a specific embodiment, the fault risk prediction model predicts the fault risk of the energy storage system, and the prediction results include: Pre-build a hybrid model composed of gradient boosting tree GBDT or LSTM-attention mechanism to calculate the probability of failure of the energy storage system in the next 24 hours. As the target output, the load characteristics, battery health parameters, and time series correlation characteristics in the historical data are used as input for training to obtain a fault risk prediction model, where: Load characteristics include comprehensive volatility indicators and the number of charge and discharge cycles of the batteries in the energy storage system : Battery health parameters include internal resistance growth rate and battery temperature gradient : in, represents the internal resistance of the battery at time t, express The internal resistance of the battery at all times, represents the temperature of the i-th battery cell; The time series correlation feature includes the statistically obtained high fluctuation duration of the continuous total load fluctuation.

[0040] The working principle and beneficial effects of the above technical solution are as follows: A pre-built failure risk prediction model, based on a gradient boosted tree (GBDT) or LSTM-attention hybrid model, uses the failure probability of the energy storage system in the next 24 hours as the target output. This model is trained using load characteristics, battery health parameters, and time-correlation features from historical data as input, resulting in a highly accurate failure risk prediction model. Load characteristics include a comprehensive volatility index and the number of charge and discharge cycles of the batteries in the energy storage system. The comprehensive volatility index comprehensively reflects load fluctuations, while the number of charge and discharge cycles correlates with battery usage. Battery health parameters include internal resistance growth rate and battery temperature gradient. The internal resistance growth rate reflects changes in the battery's internal resistance, while the temperature gradient reflects the uniformity of the battery's temperature distribution. Time-correlation features measure the duration of continuous high fluctuations to capture the persistence of load fluctuations. This model can accurately predict failure risks in advance, providing critical decision support for energy storage system maintenance and management, ensuring reliable system operation.

[0041] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. A method for fault prediction and optimization of intelligent energy storage system based on artificial intelligence, characterized in that: include: Step S1: Determine the multiple wind farms connected to the energy storage system, as well as the historical operating data, high-precision weather forecast data, and topographic information of each wind farm, and build a spatiotemporal database; Step S2: constructing a WRF-CFD coupling model embedded in a neural network prediction model based on the spatiotemporal database to predict the power generation efficiency of each wind farm and obtain a second type of prediction result; Step S3: Calculate the total load of the energy storage system based on the second type of prediction results, and predict the failure risk based on the total load fluctuation rate; Step S4: When the failure risk meets the preset conditions, an advance response plan is generated to optimize and smooth the load fluctuation of the energy storage system.

2. The method for predicting and optimizing faults of an intelligent energy storage system based on artificial intelligence according to claim 1, characterized in that: Step S2 specifically includes: Step S21: Based on the high-precision weather forecast data and topographic information in the spatiotemporal database, a WRF-CFD coupling model is constructed to perform a gradient-based hierarchical prediction of the wind force and wind direction distribution in any wind farm to obtain a first-class prediction result; Step S22: pre-constructing and training a neural network model for predicting the power generation efficiency of a single wind turbine generator set in the wind farm based on the spatiotemporal database, and using the first type of prediction results to predict the power generation efficiency of the wind turbine generator set; Step S23: Calculate the downstream wind speed and direction of the wind turbine generator using the momentum theory method to obtain a calculation result, and return the calculation result to step S21 to perform real-time correction on the first type of prediction result; Step S24: Based on the predicted power generation efficiency of each wind turbine generator set, statistics are collected on the power generation efficiency of the plurality of wind farms connected to the energy storage system to obtain a second type of prediction result.

3. The method for predicting and optimizing faults of an intelligent energy storage system based on artificial intelligence according to claim 2, characterized in that: Step S21 specifically includes: Based on the high-precision meteorological forecast data in the spatiotemporal database, the temperature field, pressure gradient, and humidity field data are determined and regional dynamic downscaling is performed using the constructed WRF model to generate initial wind field data of 1 km grid. The terrain information within the preset range around the wind farm is collected in advance through synthetic aperture radar images, and the surface roughness parameters are calculated using the following formula: in, is the surface roughness parameter, Indicates the vertical height of the surface obstacle. Represents the surface coverage. The surface coverage parameter is determined by the type of surface cover. Each type of surface cover corresponds to a unique surface coverage parameter. The obtained surface roughness parameters are input into the CFD model to complete the 10m-level micro-meteorological modeling within the wind farm; The wind farm is graded according to the wind direction, and the wind force and direction distribution in the outermost wind farm are predicted based on 10m-level micro-meteorological modeling to obtain the third type of prediction results. Waiting for the returned calculation results to be cyclically predicted for the wind force and wind direction distribution within the next gradient to obtain multiple third-category prediction results until the wind force and wind direction distribution within multiple wind farm gradients are predicted; The obtained multiple third-category prediction results are counted and integrated to obtain the first-category prediction results.

4. The method for predicting and optimizing faults of an intelligent energy storage system based on artificial intelligence according to claim 2, characterized in that: Step S22 includes: Determine the historical operating data of each wind turbine based on the spatiotemporal database, including external wind conditions and internal states. External wind conditions include wind speed, wind direction, and turbulence intensity, and internal states include pitch angle, speed, temperature, and power output. Data cleaning removes invalid periods of wind speed below the cut-in speed or above the cut-out speed, and fills missing values ​​with interpolation based on physical constraints. Key features, including wind energy density, pitch efficiency factor, and sliding window statistics, are then constructed to complete time series alignment and normalization. A hybrid neural network model with a bidirectional LSTM and a temporal convolutional network is used. The LSTM layer captures long- and short-term wind speed time series characteristics. The TCN uses dilated convolution to expand the receptive field to identify periodic patterns. The output layer embeds the static parameters of the turbine unit into a vector, and the fully connected layer uses regression to predict power generation efficiency. The static parameters of the turbine unit include rotor diameter and rated power. During training, the Smooth L1 loss function is used to reduce outlier interference. The AdamW optimizer and cosine annealing learning rate scheduling are used together with the introduction of time series dropout and label smoothing to improve the generalization ability of the model. During the verification phase, the data set was divided by season to avoid climate model leakage. The test covered unit operating conditions under three typical wind conditions: steady state, turbulent, and cut-out. The goal was to control the mean absolute error within 3% of the rated power. Finally, a neural network model for predicting the power generation efficiency of a single wind turbine in a wind farm is trained. The external wind conditions and internal states faced by the wind turbine in the first type of prediction results are used as model inputs to predict the power generation efficiency of the wind turbine and obtain the prediction results.

5. The method for fault prediction and optimization of an intelligent energy storage system based on artificial intelligence according to claim 2, characterized in that: Step S23 includes: The improved Jensen wake model is used to calculate the velocity attenuation and turbulence enhancement after the wind passes through the wind turbine. The conservation of wind energy and momentum change are determined based on the momentum theory method, and the interference effect of the upstream wind turbine on the downstream wind field is quantified. The velocity attenuation formula is expressed as: in, represents the wind speed at the hub height of the downstream wind turbine, represents the wind speed at the hub height of the upstream wind turbine, is the angle between the wind direction and the line connecting the two units, is the wake overlap area, is the impeller swept area, Indicates the impeller diameter; The turbulence enhancement formula is expressed as: in, represents the downstream turbulence intensity, represents the upstream turbulence intensity; The wind speed at the hub height of the downstream wind turbine and downstream turbulence intensity The uniform wind speed and turbulence intensity in the direction of the natural wind after it passes through the wind turbine are calculated and the results are generated.

6. The method for predicting and optimizing faults of an intelligent energy storage system based on artificial intelligence according to claim 2, characterized in that: Step S3 includes: The total load calculation process determines the predicted power generation value of the i-th wind farm at the future time t based on the second type of prediction results , based on historical data statistics, predict the power demand of the power grid at time t , combined with the power generation forecast value to calculate the total load of the energy storage system at time t : in, Indicates the total number of wind farms that input electricity into the energy storage system; when When the energy storage system enters the charging mode, the charging power , Indicates the maximum charging power of the energy storage system; when When the energy storage system enters the discharge mode, the discharge power , Indicates the maximum discharge power of the energy storage system; The total load fluctuation calculation process calculates the range of load power within the statistical period : The main frequency component is extracted through Fourier transform, and the high frequency (>0.1Hz) energy ratio is calculated: in, represents the fast Fourier transform, represents the sum of high-frequency energy, Represents the energy density of different frequency f components in the signal, Represents the total energy of the entire frequency band, It represents the proportion of high-frequency energy. A value close to 1 indicates that the power fluctuation is mainly dominated by high-frequency components, while a value close to 0 indicates that the fluctuation is gentle. Calculating the Composite Volatility Index : in, Indicates the average load power during the period, Indicates the total number of time points in the time series; Construct a failure risk prediction model and use the comprehensive volatility index , battery health parameters and timing correlation characteristics to predict the failure risk of the energy storage system and obtain the prediction results.

7. The method for fault prediction and optimization of an intelligent energy storage system based on artificial intelligence according to claim 6, characterized in that: Step S4 includes: When the fault risk meets the preset conditions, an advance response plan is generated to optimize the energy storage system load fluctuations, including limiting the charging power when the overall SOC of the energy storage system is predicted to approach a preset upper limit: in, Indicates the rated capacity of the energy storage system under standard conditions. Indicates the preset upper limit value. represents the overall SOC of the energy storage system at time t, Indicates the duration of the charging and discharging process; Limit the power change rate by sliding window: in, Indicates the maximum allowable power change rate, is the power value at time t.

8. The method for predicting and optimizing faults of an intelligent energy storage system based on artificial intelligence according to claim 7, characterized in that: The fault risk prediction model predicts the failure risk of the energy storage system and obtains the following prediction results: Pre-build a hybrid model composed of gradient boosting tree GBDT or LSTM-attention mechanism to calculate the probability of failure of the energy storage system in the next 24 hours. As the target output, the load characteristics, battery health parameters, and time series correlation characteristics in the historical data are used as input for training to obtain a fault risk prediction model, where: Load characteristics include comprehensive volatility indicators and the number of charge and discharge cycles of the batteries in the energy storage system : Battery health parameters include internal resistance growth rate and battery temperature gradient : in, represents the internal resistance of the battery at time t, express The internal resistance of the battery at all times, represents the temperature of the i-th battery cell; The time series correlation feature includes the statistically obtained high fluctuation duration of the continuous total load fluctuation.

Images