A method for energy management of charging pile groups

Abstract

This invention relates to the field of power management technology, specifically to an energy management method for charging pile clusters, comprising the following steps: S1: predicting the short-term power demand and renewable energy availability of each charging pile; S2: adjusting the charging power of each charging pile using a dynamic allocation algorithm; S3: encouraging users to charge during off-peak hours based on a preset pricing algorithm; S4: predicting user charging demand patterns; S5: monitoring power quality in real time during the charging process and automatically adjusting charging equipment parameters using preset power regulation technology; S6: automatically activating a preset backup energy supply scheme in emergency situations of grid failures. This invention optimizes energy allocation through real-time monitoring and intelligent adjustment of charging strategies, improving grid efficiency, reducing costs, and ensuring the continuity and security of charging services through a backup energy system, thereby significantly improving the operational efficiency of charging infrastructure and user satisfaction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power management, in particular to a charging pile group energy management method. BACKGROUND

[0002] With the popularity of electric vehicles and the increasing demand for environmentally friendly travel options, the construction of charging infrastructure has rapidly expanded. However, existing charging infrastructure faces many challenges, particularly in terms of grid load management, energy efficiency optimization, and service reliability. The peaks and valleys of charging demand do not always match the power supply capacity of the grid, resulting in insufficient power supply during high demand periods and energy waste during low demand periods. In addition, grid failures or instability often cause interruptions in the charging process, affecting user experience and the availability of electric vehicles.

[0003] Existing technologies mainly focus on improving the construction efficiency of charging stations and expanding network coverage, with less attention on intelligently managing the charging process to address grid fluctuations and user behavior diversity. The monitoring and management of power quality during the charging process are often insufficient, unable to effectively prevent or mitigate equipment damage and battery performance degradation caused by voltage fluctuations and harmonic pollution. In addition, there is a lack of effective demand response mechanisms to smooth grid load and reduce peak period pressure.

[0004] Therefore, there is an urgent need to develop a comprehensive energy management method to monitor and adjust charging strategies in real time, optimize energy distribution, improve the intelligence level and operational efficiency of the charging network, and ensure the continuity and safety of charging services. SUMMARY

[0005] Based on the above purpose, the present application provides a charging pile group energy management method.

[0006] A charging pile group energy management method, comprising the following steps:

[0007] S1: According to historical charging data and regional grid load information, a hybrid prediction model of SVR and ELM is used to predict the power demand of each charging pile in the short term and the available amount of renewable energy;

[0008] S2: Based on the prediction results in S1, combined with the current grid load and predicted renewable energy output, the charging power of each charging pile is adjusted through a dynamic allocation algorithm to optimize energy distribution;

[0009] S3: During the demand peak period, based on a preset pricing algorithm, users are encouraged to charge during non-peak periods to smooth the power demand curve and reduce grid load;

[0010] S4: Analyze the user's charging behavior and preferences to predict the user's charging demand pattern, and automatically adjust the charging strategy and resource allocation according to the analysis results;

[0011] S5: Real-time monitoring of power quality during charging, including voltage fluctuation and harmonic pollution, automatically adjusting charging device parameters using preset power regulation technology, optimizing power quality, protecting equipment and prolonging battery life;

[0012] S6: In the event of a power grid failure, automatically activate the preset backup power supply scheme to ensure the continuity and safety of the charging service.

[0013] Further, the S1 specifically includes:

[0014] S11: Collect historical charging data of each charging pile, including charging duration, charging start and end time, and charging power information; Collect power grid load data and historical renewable energy output data, and clean and preprocess the collected data to remove outliers and fill in missing values to ensure data quality and integrity;

[0015] S12: Based on the preprocessed data, develop a feature set, including time features, weather conditions, and historical contemporaneous power demand and energy supply data;

[0016] S13: Use support vector regression and extreme learning machine to train the model, where the SVR model is used to analyze the nonlinear trend and complex dependence of power demand, and the ELM model is used to quickly learn and adapt to the fluctuations of renewable energy output, then the prediction results of the two models are weighted and averaged, the weight is determined based on the historical prediction accuracy, to form a hybrid prediction model;

[0017] S14: Apply the hybrid prediction model formed in S13 to real-time data streams to predict the next short-term power demand of each charging pile and the available amount of solar and wind energy.

[0018] Further, the S13 specifically includes:

[0019] S131: First use the SVR model to process and predict power demand data, SVR is an effective nonlinear regression method based on statistical learning theory, which aims to find a function , such that the sum of the squared deviations between the given data points and the function is minimized while ensuring the smoothness of the function;

[0020] S132: Then use ELM to process and predict the renewable energy output, specifically the fluctuations of solar and wind energy, ELM is a single-layer feedforward neural network, the weights and biases of its hidden layer nodes are randomly initialized and fixed, the output weights are quickly calculated by analytical method, the optimization objective of ELM is to minimize the norm of output layer weights and training error, the specific formula is: wherein, is the output matrix of the hidden layer, is the output layer weight, is the target output, is the regularization coefficient, used to control the complexity of the model and avoid overfitting;

[0021] S133: Finally, combine the prediction results of SVR and ELM models to form a hybrid prediction model, achieve by applying weighted average to the prediction results of each model, the weights are calculated based on the prediction performance on the validation set, the specific formula is: wherein, is the final prediction result, and are the prediction results of SVR and ELM models respectively, is the weight determined based on historical prediction accuracy.

[0022] Further, the S2 specifically comprises:

[0023] S21: First, integrate the short-term power demand prediction results of each charging pile obtained in the S1 step with real-time power grid load data and renewable energy output prediction results, to ensure that all data are up-to-date and complete before making energy deployment decisions;

[0024] S22 defines a dynamic energy deployment strategy, which determines the charging power of each charging pile according to the current total load of the power grid, the demand prediction of each charging pile, and the availability of renewable energy, this strategy uses the following formula to calculate the charging power allocation of each charging pile: wherein, is the allocated power of the th charging pile, is the predicted demand of the th charging pile, is the total number of charging piles, is the total power available in the current power grid, is the predicted renewable energy power;

[0025] S23: Dynamically adjust the charging power settings of each charging pile according to the strategy defined in S22;

[0026] S24: After the implementation of energy allocation, monitor the deviation between actual power consumption and predicted value, and optimize the allocation strategy according to the deviation result. Specifically, when there is a difference between actual consumption and prediction, automatically fine-tune the dynamic allocation algorithm to reduce future prediction error and optimize energy use.

[0027] Further, the S24 specifically includes:

[0028] S241: First, continuously monitor the deviation between the actual power consumption of each charging pile and the predicted power demand in S1, specifically by calculating the deviation through the following formula: , wherein, represents the power consumption deviation of the th charging pile, is the actual consumption power, and is the predicted power;

[0029] S242: Analyze the pattern and cause of the deviation, including whether there is a persistent over-consumption or under-consumption, and identify whether the prediction model or allocation strategy needs to be adjusted;

[0030] S243: When the analysis result shows that there is a deviation, automatically adjust the parameters of the dynamic allocation algorithm or the prediction model;

[0031] S244: Based on the adjusted prediction model, redefine the energy allocation strategy, specifically by modifying the coefficients in the energy allocation formula to respond to changes in the prediction model, the formula is:

[0032] , wherein, is a newly introduced adjustment factor used to adjust real-time energy distribution based on historical deviation;

[0033] S245: Implement the updated allocation strategy and continuously monitor its effect. When the new strategy effectively reduces prediction error and optimizes energy use, it will continue to be used; when the effect is not good, it will continue to be adjusted iteratively.

[0034] Further, the S3 specifically includes:

[0035] S31: First, determine the peak and off-peak power demand periods according to historical grid data analysis;

[0036] S32: According to the identified peak and off-peak periods, develop a dynamic pricing strategy, including increasing the price during peak periods and reducing the price during off-peak periods, the specific price adjustment formula is:

[0037] , wherein, is the price at time , and is the base price, is the price adjustment coefficient, is the start time of the peak period, and the sign function is used to determine whether the current is a peak period;

[0038] S33: Implement the price strategy in S32 through the user's charging management system or smart metering device, specifically automatically adjust the electricity price, notify the user of the current and expected price changes, to encourage users to choose the lowest cost charging time;

[0039] S34: Monitor the changes in user charging behavior after implementing the dynamic pricing strategy, analyze the impact of charging period adjustment on grid load, including tracking the proportion of charging volume moving from peak to off-peak periods, and the contribution to overall grid load smoothing;

[0040] S35: Based on the monitoring results, adjust the pricing coefficient or modify the definition of peak and off-peak periods to encourage users to charge when the grid load is low, and optimize the power demand curve and reduce the grid load.

[0041] Further, the S4 specifically includes:

[0042] S41: First, regularly collect user charging records for each charging pile, including charging time, charging duration, charging frequency, and charging volume data, in addition to collecting related data that affect user charging behavior, including vehicle type, battery capacity, and historical charging behavior patterns;

[0043] S42: Use data mining and machine learning techniques to analyze user charging behavior patterns, identify user groups with similar charging behavior by grouping users, specifically divide users into night chargers, weekday chargers, and weekend chargers;

[0044] S43: Based on the results of behavior pattern analysis, establish a prediction model to predict the charging needs of different user groups, to predict future charging demand patterns;

[0045] S44: According to the output of the demand prediction model, automatically adjust the charging strategy and resource allocation of the charging pile, specifically when the prediction shows that the demand of night chargers will increase, automatically increase the charging resource allocation at night to ensure demand is met, in addition to dynamically adjusting the electricity price according to user preferences and charging demand to encourage users to charge during off-peak periods.

[0046] Further, the S43 specifically includes:

[0047] S431: Extract features from user charging history records, including charging time, duration, frequency, charging completion, and user type, and normalize the extracted feature data to remove the influence of dimension;

[0048] S432: Construct a Gaussian Process-based regression model to predict the charging demand, which is a random process defined on a continuous input space, and any finite set of observation points follows a multivariate normal distribution. The model is expressed as:

[0049] ;

[0050] ; wherein, is the observed charging demand value; is the input feature vector; is the noise term, representing the random error in the measurement process; represents the Gaussian process; is the mean function, used to describe the average trend of the data; is the covariance function, which defines the covariance between any two points and in the input space;

[0051] S433: Train the model and optimize the hyperparameters by maximizing the log marginal likelihood function, which is expressed as: wherein, represents the log marginal likelihood function; represents the covariance matrix without noise; is the noise variance; represents the covariance matrix with noise, wherein is the identity matrix; is the determinant of the matrix ; is the sample size, representing the total number of data points participating in training;

[0052] S434: Use the trained Gaussian Process regression model to predict future charging demand based on the latest user behavior data.

[0053] Further, the S5 specifically includes:

[0054] S51: Install power quality monitoring equipment at the charging station to monitor and record voltage fluctuations and harmonic pollution levels in real time during the charging process. The monitoring equipment can continuously collect voltage and current data, and calculate voltage fluctuations and total harmonic distortion through the following formulas:

[0055] ;

[0056] wherein, is the voltage fluctuation rate at a predetermined time ,​ and are the maximum and minimum voltage values in the time, is the nominal voltage, is the total harmonic distortion, is the voltage amplitude of the harmonic, is the fundamental voltage;

[0057] S52: Determine whether there is a power quality problem by setting a threshold value, specifically when the voltage fluctuation exceeds a predetermined threshold or the THD exceeds the standard limit, it is identified as a power quality problem;

[0058] S53: When the power quality problem is identified, the preset power regulation technology will be automatically started to adjust the charging device parameters, the specific technology includes dynamically adjusting the charging current or voltage to match the real-time conditions of the power grid, reducing the impact of voltage fluctuation and harmonics, and applying harmonic filter or phase modifier technology to reduce harmonic content;

[0059] S54: After adjustment, continue to monitor the power quality, and adjust the strategy according to the response of the device and the feedback of the power grid state, to ensure that the power quality is always in the best state, thereby protecting the charging device and prolonging the battery life.

[0060] Further, the S6 specifically includes:

[0061] S61: First, real-time monitoring of power grid state, including voltage, current and frequency parameters, and using power grid monitoring equipment to continuously evaluate the stability and performance of the power grid, once the power grid parameters deviate from the normal range, it is judged as a power grid failure;

[0062] S62: Pre-device energy supply scheme, specifically including energy storage battery system, diesel generator and renewable energy system, wherein the energy storage battery system is used to provide immediate energy support when the power grid is powered off; the diesel generator is used to start in the case of long-term power grid interruption to maintain the operation of the charging station; the renewable energy system is used to cooperate with the energy storage system to provide continuous energy supply for the charging station;

[0063] S63: When the system detects a power grid failure, it will select the predetermined backup energy supply scheme according to the nature and expected duration of the failure;

[0064] S64: Regularly test and maintain the backup energy supply system to ensure reliability and effectiveness in emergency situations, specifically including charge and discharge cycle test of energy storage battery, start and run test of diesel generator, and performance check of renewable energy equipment.

[0065] The beneficial effects of the present application are:

[0066] The present application significantly improves the energy use efficiency of charging infrastructure by real-time monitoring and intelligent prediction of charging demand and available renewable energy, precisely adjusting the charging power of each charging pile and optimizing energy distribution, not only ensuring grid load balancing, but also maximizing the use of renewable energy. This method helps to reduce energy costs, reduce dependence on traditional energy, and support environmental protection and sustainable development goals.

[0067] The present application effectively encourages users to charge during off-peak hours by dynamically adjusting electricity prices and automatically adjusting charging strategies, thereby smoothing the power demand curve and reducing grid load, which not only improves the efficiency and stability of the grid, but also helps to reduce the pressure on the grid during peak hours and avoid overload situations. In addition, by optimizing the distribution of charging periods, charging stations can more effectively manage their operations, improve service quality and user satisfaction.

[0068] The present application can quickly switch to backup energy in the event of grid instability or failure, ensuring the continuity and safety of charging services, and this rapid response capability not only protects charging equipment from voltage fluctuations and harmonic pollution, but also extends the service life of batteries and reduces maintenance costs. BRIEF DESCRIPTION OF DRAWINGS

[0069] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only belong to the present application, and other drawings can also be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0070] Fig. 1 The present application is a charging pile group energy management method schematic diagram.

[0071] Fig. 2 The present application is an automatic adjustment of charging equipment parameter flowchart. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical scheme and advantages of the present application more clear, the following will further illustrate the present application in combination with specific embodiments.

[0073] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present application shall be the usual meanings understood by those skilled in the art to which the present application belongs. The "first", "second" and similar words used in the present application do not represent any order, quantity or importance, but are only used to distinguish different components. "Include" or "contain" and similar words mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects.

[0074] As shown in Figs. 1-2 A charging pile group energy management method, comprising the following steps:

[0075] S1: According to the historical charging data and regional power grid load information, a hybrid prediction model of SVR (Support Vector Regression) and ELM (Extreme Learning Machine) is used to predict the power demand of each charging pile in the short term and the available amount of renewable energy (such as solar and wind energy);

[0076] S2: According to the prediction results in S1, combining the current power grid load and the predicted renewable energy output, the charging power of each charging pile is adjusted through a dynamic allocation algorithm to optimize energy distribution and ensure the optimal load balancing and energy supply of the power grid;

[0077] S3: During the demand peak period, based on the preset pricing algorithm, encourage users to charge during off-peak hours to smooth the power demand curve and reduce the load of the power grid;

[0078] S4: Analyze the charging behavior and preferences of users to predict the user charging demand pattern, and automatically adjust the charging strategy and resource allocation according to the analysis results to improve user experience and system efficiency;

[0079] S5: Real-time monitoring of power quality during charging, including voltage fluctuation and harmonic pollution, using preset power regulation technology to automatically adjust charging device parameters to optimize power quality, protect equipment and prolong battery life;

[0080] S6: In the event of a power grid failure, automatically activate the preset backup energy supply scheme, such as energy storage batteries or diesel generators, to ensure the continuity and safety of the charging service.

[0081] S1 specifically includes:

[0082] S11: Collect historical charging data of each charging pile, including charging duration, charging start and end time, charging power information; At the same time, collect power grid load data and historical output data of renewable energy (solar and wind energy), and clean and preprocess the collected data to remove outliers and fill in missing values to ensure data quality and integrity;

[0083] S12: Based on the pre-processed data, develop a feature set including temporal features (e.g. hour, week, season), weather conditions (e.g. temperature, wind speed, sunshine duration), and historical same-period power demand and energy supply data; these features help capture key factors affecting charging demand and renewable energy output;

[0084] S13: Train models using Support Vector Regression (SVR) and Extreme Learning Machine (ELM), where the SVR model is used to analyze the nonlinear trends and complex dependencies of power demand, while the ELM model is used to quickly learn and adapt to fluctuations in renewable energy output, then the prediction results of the two models are weighted and averaged, the weight is determined based on historical prediction accuracy, to form a hybrid prediction model;

[0085] S14: Apply the hybrid prediction model formed in S13 to real-time data streams to predict the power demand of each charging pile and the availability of solar and wind energy in the next short term (e.g. within 24 to 48 hours), the hybrid prediction model can also dynamically adjust the prediction in response to immediate data changes to provide the most accurate prediction results; the hybrid prediction model can accurately predict the power demand of each charging pile and the availability of solar and wind energy in the short term, through optimized prediction to improve the efficiency and accuracy of energy distribution, reduce energy waste, enhance the stability and reliability of the power grid, this prediction capability is particularly critical for managing large-scale charging pile networks, helping to achieve more efficient energy utilization and better environmental protection effect.

[0086] S13 specifically includes:

[0087] S131: First use the SVR model to process and predict power demand data, SVR is an effective nonlinear regression method based on statistical learning theory, its goal is to find a function that minimizes the sum of squared deviations between given data points and the function, while ensuring the smoothness of the function, mathematically, SVR is achieved by solving the following optimization problem:

[0088] ;

[0089] ;

[0090] ;

[0091] ; where, is the weight vector, is the bias term, is the regularization parameter, used to control the complexity of the model to prevent overfitting; and is a slack variable used to handle data points that cannot be perfectly inside the error margin; is the error tolerance, defining the width of the penalty-free region, the difference between the predicted and actual values of a data point must be less than this value, otherwise an error cost will be incurred;

[0092] S132: Then use ELM to process and predict the fluctuations of renewable energy output, specifically solar and wind energy, ELM is a single-layer feedforward neural network, the weights and biases of its hidden layer nodes are randomly initialized and fixed, the output weights are quickly calculated by analytical method, the optimization goal of ELM is to minimize the norm of output layer weights and training error, the specific formula is: where, is the output matrix of the hidden layer, is the output layer weight, is the target output, is the regularization coefficient, used to control the complexity of the model and avoid overfitting;

[0093] S133: Finally, combine the prediction results of SVR and ELM models to form a hybrid prediction model, achieve by applying weighted average to the prediction results of each model, the weights are calculated based on their prediction performance on the validation set, the specific formula is: where, is the final prediction result, and are the prediction results of SVR and ELM models respectively, is the weight determined based on historical prediction accuracy; this hybrid prediction model combines the high accuracy of SVR and the fast learning ability of ELM, improves the accuracy and efficiency of the overall prediction, by precisely adjusting the weights of each model, can ensure the optimal prediction result in different application environments, effectively guide the energy management and scheduling strategy of charging piles, optimize the utilization of energy resources, reduce waste.

[0094] S2 specifically includes:

[0095] S21: First, integrate the short-term power demand prediction results of each charging pile obtained in step S1 with real-time grid load data and renewable energy (such as solar and wind energy) output prediction results, ensuring that all data is up-to-date and complete before making energy deployment decisions;

[0096] S22 defines a dynamic energy deployment strategy, which determines the charging power of each charging pile according to the current total grid load, demand prediction of each charging pile, and availability of renewable energy, this strategy uses the following formula to calculate the charging power allocation of each charging pile: where, is the charging power of the i-th charging pile, The allocated power of each charging station It is the first Forecasted demand for each charging station This is the total number of charging stations. It is the total power currently available on the power grid. It is the predicted renewable energy power;

[0097] S23: According to the strategy defined in S22, dynamically adjust the charging power settings of each charging pile. The specific adjustment is executed through a preset central control system, which can receive and process data in real time and adjust the charging speed of each charging pile as needed.

[0098] S24: After implementing energy allocation, monitor the deviation between actual power consumption and predicted values, and optimize the allocation strategy based on the deviation results. Specifically, when there is a difference between actual consumption and prediction, the dynamic allocation algorithm will be automatically fine-tuned to reduce future prediction errors and optimize energy use. Through the above dynamic allocation algorithm, the system can flexibly respond to changes in power demand and supply, optimize the energy allocation of charging piles, improve energy utilization efficiency, and reduce energy waste. In addition, through continuous feedback and optimization, the system can adapt to various environmental changes and improve the operating efficiency and reliability of the entire charging network.

[0099] S24 specifically includes:

[0100] S241: First, continuously monitor the deviation between the actual power consumption of each charging station and the predicted power demand in S1. Specifically, the deviation is calculated using the following formula: ,in, Indicates the first The power consumption deviation of each charging station It is the actual power consumed, and It is a predicted power;

[0101] S242: Analyze the patterns and causes of deviations, including whether there is persistently high or low consumption, and identify whether the forecasting model or allocation strategy needs to be adjusted.

[0102] S243: When the analysis results show a bias, the parameters of the dynamic allocation algorithm or prediction model will be automatically adjusted. The adjustment strategies include:

[0103] Adjust weights In predictive models, in order to better balance the influence of SVR and ELM models;

[0104] Adjusting regularization parameters and In SVR and ELM models, to improve model fit and generalization ability;

[0105] S244: Based on the adjusted prediction model, the energy allocation strategy is redefined, specifically by modifying the coefficients in the energy allocation formula to respond to changes in the prediction model. The formula is as follows:

[0106] ,in, It is a newly introduced adjustment factor used to adjust real-time energy allocation based on historical deviations;

[0107] S245: Implement the updated allocation strategy and continuously monitor its effectiveness. If the new strategy effectively reduces prediction errors and optimizes energy use, it will continue to be adopted; if the effect is not good, it will continue to be iteratively adjusted. Through the dynamic monitoring and adjustment mechanism of the above steps, it can be ensured that the energy allocation strategy is always based on the most accurate prediction data and responds to various changes in a timely manner. This not only improves the efficiency of energy use, but also enhances the system's adaptability to uncertainty, ultimately achieving higher economic benefits and user satisfaction.

[0108] S3 specifically includes:

[0109] S31: First, based on historical power grid data analysis, determine the peak (such as morning and evening) and off-peak (such as night) periods of electricity demand. This step is based on the overall load level of the power grid and aims to identify the periods when electricity use is most concentrated and least concentrated.

[0110] S32: Based on the identified peak and off-peak periods, formulate a dynamic electricity pricing strategy, including increasing electricity prices during peak periods and decreasing electricity prices during off-peak periods. The specific price adjustment formula is as follows:

[0111] ,in, It is in time Electricity price, It is the base electricity price. It is a price adjustment coefficient. It is the start time of the peak period. The sign function is used to determine whether the current period is a peak period.

[0112] S33: Implement the S32 mid-price strategy through the user's charging management system or smart metering device. Specifically, the electricity price will be automatically adjusted, and the user will be notified of the current and expected electricity price changes to incentivize the user to choose the charging time with the lowest cost.

[0113] S34: Monitor changes in user charging behavior after the implementation of dynamic pricing strategies, and analyze the impact of adjustments to charging periods on grid load, including tracking the proportion of charging volume shifting from peak to off-peak periods, and its contribution to the overall grid load smoothness.

[0114] S35: Adjust pricing coefficients based on monitoring results Alternatively, the definitions of peak and off-peak periods can be modified to encourage users to charge when grid load is low, thereby optimizing the electricity demand curve and reducing grid load. This time-based dynamic pricing strategy effectively guides users to charge during off-peak hours, helping to smooth the load curve of the entire grid and reduce grid pressure during peak periods. This strategy not only reduces grid operating costs but also improves grid operating efficiency and reliability, while providing users with the possibility of saving energy costs.

[0115] S4 specifically includes:

[0116] S41: First, regularly collect charging records from users of each charging station, including charging time, charging duration, charging frequency, and charging amount data. In addition, collect relevant data that affect user charging behavior, including vehicle model, battery capacity, and historical charging behavior patterns.

[0117] S42: Use data mining and machine learning techniques, such as cluster analysis (K-means), decision trees or neural networks, to analyze users’ charging behavior patterns. By grouping users, identify user groups with similar charging behaviors, specifically categorizing users into nighttime chargers, weekday chargers, and weekend chargers.

[0118] S43: Based on the behavioral pattern analysis results, establish a predictive model to predict the charging needs of different user groups in order to predict future charging demand patterns.

[0119] S44: Based on the output of the demand forecasting model, the charging strategy and resource allocation of charging piles are automatically adjusted. Specifically, when the forecast indicates that the demand of nighttime charging users will increase, the nighttime charging resource allocation will be automatically increased to ensure that the demand is met. In addition, the electricity price is dynamically adjusted according to user preferences and charging demand to incentivize users to charge during off-peak hours. Through the above-mentioned refined user charging behavior analysis and demand forecasting, this invention can effectively predict and meet the charging needs of different user groups and optimize the allocation of charging resources. This not only improves the operational efficiency of charging stations but also enhances user satisfaction, reduces the overall load on the power grid, and ultimately achieves the optimization of energy use.

[0120] S43 specifically includes:

[0121] S431: Extract features from the user's charging history, including charging time, duration, frequency, charging completion rate, and user type (commercial user, residential user, etc.), and normalize the extracted feature data to remove the influence of units and ensure the model's generalization ability.

[0122] S432: Construct a regression model based on Gaussian processes to predict charging demand. A Gaussian process is a stochastic process defined on a continuous input space, and any finite set of observations follows a multivariate normal distribution. The model formula is expressed as follows:

[0123] ;

[0124] ;in, The observed charging demand value is represented by the actual charging amount of each charging station within a specific time period. The input feature vector includes features related to user charging behavior, such as charging start time, charging duration, calendar features (weekdays or weekends), and weather conditions. The noise term represents random errors in measurement and process, and is usually assumed to follow a Gaussian distribution. It expresses the deviation between the model output and the actual observation. A Gaussian process is a stochastic process in which any finite set of observations is assumed to follow a multivariate normal distribution. It is the mean function, used to describe the average trend of data. In many practical applications, it can be simply set to zero, indicating that there is no prior trend. Let covariance function be defined for any two points in the input space. and The covariance between input features is the core of the model and is used to describe how input features affect the correlation between outputs. Commonly used covariance functions include radial basis function (RBF) and Matérn function.

[0125] S433: The model is trained and hyperparameters are optimized by maximizing the logarithmic marginal likelihood function. The specific formula is as follows: ,in, The logarithmic marginal likelihood function is used for model training and hyperparameter optimization, measuring the likelihood of a given input. Lower output The probability of; Represents the covariance matrix without noise, whose elements are... The calculation yields a correlation used to describe the relationship between different data points; Noise variance represents the level of random noise in the data and is a key hyperparameter of the GPR model. This represents the covariance matrix with added noise. ,in It is the identity matrix, used to account for the effects of noise in prediction; For matrix The determinant is used to calculate a component of the log-likelihood function, which affects parameter optimization; The number of samples represents the total number of data points used in the training process.

[0126] S434: Using a trained Gaussian process regression model, future charging demand is predicted based on the latest user behavior data. Gaussian process regression can provide the mean and variance for each prediction point. The variance can be used as a measure of prediction uncertainty to assist in making more prudent strategic decisions. Due to its excellent prediction performance and ability to provide prediction uncertainty, the Gaussian process regression model is particularly suitable for application scenarios such as demand forecasting. Through this model, charging station operators can more accurately predict charging demand, optimize the allocation of charging resources, improve service quality and operational efficiency, and reduce energy waste.

[0127] S5 specifically includes:

[0128] S51: Install power quality monitoring equipment at charging stations to monitor and record voltage fluctuations and harmonic pollution levels in real time during the charging process. This monitoring equipment can continuously collect voltage and current data, and specifically calculate voltage fluctuations and total harmonic distortion using the following formulas. :

[0129] ;

[0130] ,in, It is the scheduled time voltage fluctuation rate, and These are the maximum and minimum voltage values ​​within that time period. It is the nominal voltage. It is total harmonic distortion. It is the first The voltage amplitude of the second harmonic It is the fundamental voltage;

[0131] S52: Determine whether there is a power quality problem by setting a threshold. Specifically, when the voltage fluctuation exceeds a predetermined threshold (e.g., 5%) or the THD exceeds the standard limit (e.g., 5% recommended by IEEE), it is identified as a power quality problem.

[0132] S53: When a power quality problem is detected, preset power regulation technology will be automatically activated to adjust the charging equipment parameters. Specific technologies include dynamically adjusting the charging current or voltage to match the real-time conditions of the power grid, reducing the impact of voltage fluctuations and harmonics, and applying harmonic filters or phase modulators to reduce harmonic content. The specific calculation formula is as follows:

[0133] ;

[0134] ,in, and These are the adjusted charging current and phase angle, respectively. and These are the settings under normal operating conditions. and It is an adjustment factor;

[0135] S54: After adjustment, continue to monitor power quality and adjust strategies based on equipment response and grid status feedback to ensure that power quality is always kept in optimal condition, thereby protecting charging equipment and extending battery life. By implementing the above steps, charging facilities can effectively deal with power quality problems, improve the stability and safety of the charging process, extend the service life of batteries and charging equipment, and improve overall charging efficiency. This intelligent response and automatic adjustment mechanism of the system significantly improves user satisfaction and operational efficiency.

[0136] S6 specifically includes:

[0137] S61: First, monitor the power grid status in real time, including parameters such as voltage, current and frequency, and use power grid monitoring equipment to continuously evaluate the stability and performance of the power grid. Once a power grid parameter is detected to deviate from the normal range (such as a voltage drop exceeding a predetermined threshold or an abnormal frequency), it is judged as a power grid fault.

[0138] S62: A preset backup energy supply plan, specifically including an energy storage battery system, a diesel generator, and a renewable energy system (such as solar panels and wind power). The energy storage battery system is used to provide immediate energy support when the grid is down; the diesel generator is used to start up in the event of a prolonged grid outage to maintain the operation of the charging station; and the renewable energy system (such as solar panels and wind power) is used in conjunction with the energy storage system to provide a continuous energy supply for the charging station.

[0139] S63: When the system detects a grid fault, it will select a predetermined backup energy supply scheme based on the nature of the fault and its expected duration. For example, a short-term grid fluctuation may only require the activation of the energy storage battery system, while a long-term power outage will require the start of a diesel generator.

[0140] S64: Regularly test and maintain the backup energy supply system to ensure its reliability and effectiveness in emergencies. This includes charge-discharge cycle testing of energy storage batteries, start-up and operation testing of diesel generators, and performance checks of renewable energy equipment. By implementing the above steps, charging stations can respond quickly to grid failures and maintain the continuity and safety of charging services through pre-set backup energy supply plans. This not only ensures that the charging needs of electric vehicle users are met, but also improves the reliability of charging facilities and users' trust in the charging service.

[0141] This invention is intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A charging pile group energy management method, characterized in that, The method comprises the following steps: S1: According to historical charging data and regional power grid load information, a hybrid prediction model of SVR and ELM is used to predict the power demand of each charging pile in the short term and the available amount of renewable energy; Specifically, it includes: S11: Collect the historical charging data of each charging pile, including charging duration, charging start and end time, and charging power information; At the same time, collect the power grid load data and the historical output data of renewable energy, and clean and preprocess the collected data to remove outliers and fill in missing values, so as to ensure data quality and integrity; S12: Based on the preprocessed data, develop a feature set, which includes time features, weather conditions, and historical contemporaneous power demand and energy supply data; S13: Use support vector regression and extreme learning machine to train the model, wherein the SVR model is used to analyze the nonlinear trend and complex dependence of power demand, and the ELM model is used to quickly learn and adapt to the fluctuations of renewable energy output, and then the prediction results of the two models are weighted and averaged, and the weight is determined based on the historical prediction accuracy to form a hybrid prediction model; S14: Apply the hybrid prediction model formed in S13 to real-time data streams to predict the power demand of each charging pile in the next short term and the available amount of solar and wind energy; S2: According to the prediction results in S1, combine the current power grid load and the predicted renewable energy output, and adjust the charging power of each charging pile through a dynamic allocation algorithm to optimize energy distribution; S3: During the demand peak period, based on the preset pricing algorithm, encourage users to charge during off-peak hours to smooth the power demand curve and reduce the load on the power grid; S4: Analyze the charging behavior and preferences of users to predict user charging demand patterns, and automatically adjust charging strategies and resource allocation based on the analysis results; S5: Real-time monitoring of power quality during charging, including voltage fluctuations and harmonic pollution, using preset power regulation techniques to automatically adjust charging device parameters to optimize power quality, protect equipment and extend battery life; S6: In the event of a power grid failure, automatically activate the preset backup energy supply scheme to ensure the continuity and safety of the charging service.

2. The charging pile group energy management method according to claim 1, characterized in that, The S13 specifically includes: S131: First, the power demand data is processed and predicted using an SVR model, which is an effective nonlinear regression method based on statistical learning theory, whose goal is to find a function that minimizes the sum of the squared deviations between the given data points and the function while ensuring the smoothness of the function; S132: Then use ELM to process and predict the renewable energy output, specifically the fluctuations of solar and wind energy, ELM is a single-layer feedforward neural network, the weights and biases of its hidden layer nodes are randomly initialized and fixed, the output weights are quickly calculated by analytical method, the optimization objective of ELM is to minimize the norm of output layer weights and training error, the specific formula is: where, is the output matrix of the hidden layer, is the output layer weight, is the target output, is the regularization coefficient, used to control the complexity of the model and avoid overfitting; S133: Finally, combine the prediction results of the SVR and ELM models to form a hybrid prediction model, which is achieved by applying weighted averaging to the prediction results of each model, and the weight is calculated based on the prediction performance of each model on the validation set, and the specific formula is: where, is the final prediction result, and are the prediction results of SVR and ELM model respectively, is the weight determined based on historical prediction accuracy.

3. The charging pile group energy management method of claim 1, wherein, The S2 specifically includes: S21: First, integrate the short-term power demand prediction results of each charging pile obtained in step S1 with real-time power grid load data and renewable energy output prediction results to ensure that all data are up-to-date and complete before making energy allocation decisions; S22 defines a dynamic energy allocation strategy that determines the charging power of each charging pile based on the current total load of the power grid, the demand prediction of each charging pile, and the availability of renewable energy. This strategy uses the following formula to calculate the charging power allocation of each charging pile: wherein, is the allocated power for the th charging station, is the predicted demand for the th charging station, is the total number of charging stations, is the total power available from the current grid, is the predicted renewable energy power; S23: Dynamically adjust the charging power settings of each charging pile according to the strategy defined in S22; S24: After the implementation of the energy allocation, monitor the deviation between the actual power consumption and the predicted value, and optimize the allocation strategy according to the deviation result. Specifically, when there is a difference between the actual consumption and the prediction, the dynamic allocation algorithm will be automatically fine-tuned to reduce future prediction errors and optimize energy use.

4. The charging pile group energy management method according to claim 3, characterized in that, The S24 specifically includes: S241: First, continuously monitor the deviation between the actual power consumption of each charging pile and the predicted power demand in S1, specifically calculate the deviation by the following formula: wherein, represents the power consumption deviation of the th charging pile, is the actual consumed power, and is the predicted power; S242: Analyze the pattern and cause of the deviation, including whether there is a persistent over-consumption or under-consumption, and identify whether the prediction model or allocation strategy needs to be adjusted; S243: When the analysis result shows that there is a deviation, automatically adjust the parameters of the dynamic allocation algorithm or the prediction model; S244: Based on the adjusted prediction model, redefine the energy allocation strategy, specifically respond to the changes in the prediction model by modifying the coefficients in the energy allocation formula, the formula is: wherein, is a newly introduced adjustment factor for adjusting real-time energy distribution according to historical bias; S245: Implement the updated allocation strategy and continuously monitor its effect, when the new strategy effectively reduces the prediction error and optimizes the energy use, it will continue to be adopted; when the effect is not good, it will continue to be iteratively adjusted.

5. The charging pile group energy management method according to claim 1, characterized in that, The S3 specifically includes: S31: First, determine the peak and off-peak power demand periods according to historical power grid data analysis; S32: According to the identified peak and off-peak periods, develop a dynamic pricing strategy, including increasing the price during peak periods and reducing the price during off-peak periods, the specific price adjustment formula is: ,in, It is in time Electricity price, It is the base electricity price. It is a price adjustment coefficient. It is the start time of the peak period. The sign function is used to determine whether the current period is a peak period. S33: Implement the price strategy in S32 through the user's charging management system or intelligent metering equipment, specifically automatically adjust the price, notify the user of the current and expected price changes, to encourage users to choose the lowest cost charging time; S34: Monitor the changes in user charging behavior after implementing the dynamic pricing strategy, analyze the impact of charging period adjustment on power grid load, including tracking the proportion of charging volume moving from peak period to off-peak period, and the contribution to overall power grid load smoothing; S35: Adjust the pricing coefficient based on the monitoring result or modify the definition of peak and off-peak hours to encourage users to charge when the grid load is lower and optimize the power demand curve and reduce the grid load.

6. The charging pile group energy management method according to claim 1, characterized in that, The S4 specifically includes: S41: First, regularly collect charging records of users of each charging pile, including charging time, charging duration, charging frequency, and charging volume data, in addition to collecting related data that affect user charging behavior, including vehicle type, battery capacity, and historical charging behavior patterns; S42: Use data mining and machine learning techniques to analyze user charging behavior patterns, identify user groups with similar charging behavior by grouping users, specifically divide users into night chargers, weekday chargers, and weekend chargers; S43: Based on the results of behavior pattern analysis, establish a prediction model to predict the charging demand of different user groups, to predict future charging demand patterns; S44: According to the output of the demand prediction model, automatically adjust the charging strategy and resource allocation of the charging pile, specifically when the prediction shows that the demand of night chargers will increase, automatically increase the charging resource allocation at night to ensure demand is met, in addition to dynamically adjusting the price according to user preferences and charging demand to encourage users to charge during off-peak periods.

7. The charging pile group energy management method according to claim 6, characterized in that, The S43 specifically includes: S431: Extract features from the user's charging history, including charging time, duration, frequency, charging completion degree and user type, and normalize the extracted feature data to remove the influence of dimension; S432: Construct a regression model based on Gaussian process to predict charging demand, Gaussian process is a random process defined on continuous input space, any finite set of observation points of which obeys multivariate normal distribution, the model formula is expressed as: ； ; wherein is the observed charging demand value; is the input feature vector; is the noise term representing random errors in the measurement process; denotes a Gaussian process; is the mean function used to describe the average trend of the data; is the covariance function that defines the covariance between any two points and in the input space; S433: Train the model and optimize the hyperparameters by maximizing the log marginal likelihood function, specifically: where, represents the log marginal likelihood function; represents the covariance matrix without noise; is the noise variance; represents the covariance matrix with the addition of the noise term, where is the identity matrix; is the determinant of the matrix ; and is the number of samples, representing the total number of data points participating in the training. S434: Use the trained Gaussian process regression model to predict future charging demand according to the latest user behavior data.

8. The charging pile group energy management method according to claim 1, characterized in that, The S5 specifically includes: S51: Install power quality monitoring equipment at the charging station for real-time monitoring and recording of voltage fluctuations and harmonic pollution levels during charging. This monitoring equipment is capable of continuously collecting voltage and current data, specifically calculating voltage fluctuations and total harmonic distortion through the following formulas: ; wherein, is the voltage fluctuation rate, of a predetermined time and are the maximum and minimum voltage values, respectively, over that time, is the nominal voltage, is the total harmonic distortion, is the voltage amplitude of the th harmonic, is the fundamental voltage; S52: Determine whether there is a power quality problem by setting a threshold, specifically when the voltage fluctuation exceeds the predetermined threshold or the THD exceeds the standard limit, it is identified as a power quality problem; S53: When the power quality problem is identified, the preset power regulation technology will be automatically started to adjust the charging equipment parameters, the specific techniques include dynamically adjusting the charging current or voltage to match the real-time conditions of the power grid, reducing the influence of voltage fluctuation and harmonics, and applying harmonic filter or phase modifier technology to reduce harmonic content; S54: Continue to monitor the power quality after adjustment, and adjust the strategy according to the response of the equipment and the feedback of the power grid state, to ensure that the power quality always remains in the optimal state, thereby protecting the charging equipment and prolonging the battery life.

9. The charging pile group energy management method of claim 1, wherein, The S6 specifically includes: S61: First, real-time monitoring of power grid state, including voltage, current and frequency parameters, and using grid monitoring equipment to continuously evaluate the stability and performance of the power grid, once the detection of power grid parameters deviates from the normal range, it is judged as power grid failure; S62: Pre-device energy supply scheme, including energy storage battery system, diesel generator and renewable energy system, among which the energy storage battery system is used to provide immediate energy support when the power grid is powered off; the diesel generator is used to start in the case of long-term power grid interruption to maintain the operation of the charging station; the renewable energy system is used to cooperate with the energy storage system to provide continuous energy supply for the charging station; S63: When the system detects a power grid failure, it will select the predetermined backup energy supply scheme according to the nature and expected duration of the failure; S64: Regularly test and maintain the backup energy supply system to ensure reliability and effectiveness in emergency situations, including charge and discharge cycle test of energy storage battery, start and run test of diesel generator, and performance check of renewable energy equipment.

Images