Charging and discharging management method and device, computer equipment and readable storage medium

Abstract

The invention relates to a charging and discharging management method and device, computer equipment and a readable storage medium. The method comprises the steps of obtaining multi-source data of a target regional power grid in a preset historical time period, preprocessing the multi-source data to obtain a feature vector, and performing feature extraction processing on the feature vector through a pre-trained deep learning model to obtain an electricity price fluctuation prediction result of the power grid in a specified time period, based on the electricity price fluctuation prediction result and the energy storage system parameters of the power grid, an optimization algorithm is adopted to generate a charging and discharging strategy of the energy storage system, and charging and discharging management is carried out on the energy storage system according to the charging and discharging strategy; the method can systematically analyze the long-term trend and periodic change rule of the electricity price fluctuation, makes up for the defect that the traditional method cannot fully reflect the electricity price fluctuation, can accurately capture the characteristics of the electricity price fluctuation, and provides effective support for the construction planning of a new energy power station, the capacity configuration of an energy storage system and the formulation of a charging and discharging strategy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of machine learning, and in particular to a charging and discharging management method and device, computer equipment and a readable storage medium. BACKGROUND

[0002] In recent years, the structure of the electricity market continues to evolve, and large-scale grid-connected access of new energy power generation such as wind power and photovoltaic power makes the operation environment of the power system more complex. Under this background, the nodal price is affected by multiple factors such as energy supply and demand relationship, randomness of new energy output, and policy regulation, and presents significant volatility characteristics. The traditional nodal price prediction method focuses on the single-point price value prediction, and lacks systematic analysis of the long-term trend of price volatility, periodic change rule and coupling relationship of influencing factors. This leads to the difficulty of accurately reflecting the whole picture of price volatility in the prediction result, and cannot provide effective support for new energy power station construction planning, energy storage system capacity configuration and charging and discharging strategy formulation. SUMMARY

[0003] Therefore, it is necessary to provide a charging and discharging management method and device capable of predicting the long-term trend of price volatility based on multi-source data, computer equipment, computer readable storage medium and computer program product.

[0004] In a first aspect, the present application provides a charging and discharging management method, comprising:

[0005] obtaining multi-source data of a target regional power grid in a preset historical time period, and pre-processing the multi-source data to obtain a feature vector;

[0006] performing feature extraction processing on the feature vector through a pre-trained deep learning model to obtain a price volatility prediction result of the power grid in a specified time period; the price volatility prediction result includes price volatility indexes of each node in the power grid at different time points, and the price volatility indexes include at least one of maximum price difference, variance or standard deviation;

[0007] generating a charging and discharging strategy of an energy storage system based on the price volatility prediction result and parameters of the energy storage system of the power grid by using an optimization algorithm; the charging and discharging strategy is a charging and discharging behavior sequence of each time window in the specified time period;

[0008] performing charging and discharging management on the energy storage system according to the charging and discharging strategy.

[0009] In one of the embodiments, the multi-source data includes at least one of historical electricity price, load curve, coal price information, tie-line capacity, network topology, time characteristics, season, holiday, target regional gross domestic product forecast value, newly-built power plant, industrial park information and new energy output data of each node in the power grid; the pre-processing of the multi-source data to obtain the feature vector includes:

[0010] cleaning, denoising and normalizing the multi-source data to obtain the pre-processed multi-source data;

[0011] transforming the pre-processed multi-source data into the feature vector.

[0012] In one of the embodiments, the pre-trained deep learning model includes at least one of a convolutional neural network module, a long short-term memory network and a Transformer; the feature extraction processing of the feature vector by the pre-trained deep learning model to obtain the electricity price volatility prediction result of the power grid in the specified time period includes:

[0013] extracting the spatial feature in the feature vector by the convolutional neural network module;

[0014] modeling the long-term dependence relationship in the time series data in the multi-source data by the long short-term memory network or the Transformer to capture and analyze the inherent nonlinear characteristics and non-stationary characteristics in the electricity price volatility of the power grid in the specified time period, thereby forming an electricity price volatility curve;

[0015] combining the spatial feature and the electricity price volatility curve to obtain the electricity price volatility prediction result.

[0016] In one of the embodiments, based on the electricity price volatility prediction result and the energy storage system parameters of the power grid, an optimization algorithm is used to generate the charge-discharge strategy of the energy storage system, which includes:

[0017] based on the electricity price volatility prediction result and the energy storage system parameters of the power grid, a genetic algorithm in the optimization algorithm is used to generate the charge-discharge strategy of the energy storage system.

[0018] In one of the embodiments, based on the electricity price volatility prediction result and the energy storage system parameters of the power grid, the genetic algorithm is used to generate the charge-discharge strategy of the energy storage system, which includes:

[0019] based on the power and capacity constraints in the energy storage system parameters, a group of initial individuals representing the charge-discharge strategy are randomly generated; each of the initial individuals represents the charge-discharge strategy of each time window in the specified time period;

[0020] The price fluctuation prediction result determines a charging-discharging price difference benefit for each of the coded initial individuals, determines an energy storage system loss cost according to the energy storage system parameters, and obtains a fitness value of the coded initial individual according to the charging-discharging price difference benefit and the loss cost;

[0021] The coded initial individuals are selected according to the fitness values to perform roulette selection, single-point crossover and probability mutation operations, and a new child population satisfying the energy storage system parameter constraint is generated;

[0022] The new child population is taken as a next round population, and the generation of the fitness value and the new child population is repeated until a fitness value change rate is lower than a preset threshold or a maximum iteration number is reached;

[0023] A coded initial individual with the highest fitness value in the final population is selected and decoded into the charging-discharging strategy.

[0024] In one of the embodiments, the training method of the pre-trained deep learning model comprises:

[0025] Historical multi-source sample data of the target regional power grid and a corresponding actual price fluctuation sequence are obtained;

[0026] The historical multi-source sample data are identified by a preset initial model to obtain a predicted price fluctuation sequence;

[0027] A loss function is determined according to the actual price fluctuation sequence, the predicted price fluctuation sequence, parameters of the preset initial model and a regularization coefficient;

[0028] The parameters of the preset initial model are iteratively optimized by using a back propagation algorithm and a gradient descent method until the loss function meets a preset condition.

[0029] In a second aspect, the application further provides a charging-discharging management device, comprising:

[0030] A data acquisition module is configured to acquire multi-source data of a target regional power grid in a preset historical time period, and to obtain a feature vector by preprocessing the multi-source data;

[0031] A result prediction module is configured to perform feature extraction processing on the feature vector by using a pre-trained deep learning model to obtain a price fluctuation prediction result of the power grid in a specified time period; the price fluctuation prediction result comprises price fluctuation indexes of each node in the power grid at different time points, and the price fluctuation indexes comprise at least one of a maximum price difference, a variance or a standard deviation;

[0032] a strategy generation module configured to generate a charging and discharging strategy of the energy storage system based on the electricity price fluctuation prediction result and parameters of the energy storage system of the power grid, the charging and discharging strategy being a charging and discharging behavior sequence of each time window in the specified time period;

[0033] a charging and discharging management module configured to perform charging and discharging management on the energy storage system according to the charging and discharging strategy.

[0034] In a third aspect, the present application further provides a computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the following steps when executing the computer program:

[0035] obtaining multi-source data of a target regional power grid in a preset historical time period, and obtaining a feature vector by preprocessing the multi-source data;

[0036] performing feature extraction processing on the feature vector by using a pre-trained deep learning model to obtain an electricity price fluctuation prediction result of the power grid in a specified time period, the electricity price fluctuation prediction result comprising electricity price fluctuation indexes of each node in the power grid at different time points, and the electricity price fluctuation indexes comprising at least one of a maximum price difference, a variance or a standard deviation;

[0037] generating a charging and discharging strategy of the energy storage system based on the electricity price fluctuation prediction result and parameters of the energy storage system of the power grid by using an optimization algorithm, the charging and discharging strategy being a charging and discharging behavior sequence of each time window in the specified time period;

[0038] performing charging and discharging management on the energy storage system according to the charging and discharging strategy.

[0039] In a fourth aspect, the present application further provides a computer readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the following steps:

[0040] obtaining multi-source data of a target regional power grid in a preset historical time period, and obtaining a feature vector by preprocessing the multi-source data;

[0041] performing feature extraction processing on the feature vector by using a pre-trained deep learning model to obtain an electricity price fluctuation prediction result of the power grid in a specified time period, the electricity price fluctuation prediction result comprising electricity price fluctuation indexes of each node in the power grid at different time points, and the electricity price fluctuation indexes comprising at least one of a maximum price difference, a variance or a standard deviation;

[0042] generating a charging and discharging strategy of the energy storage system based on the electricity price fluctuation prediction result and parameters of the energy storage system of the power grid by using an optimization algorithm, the charging and discharging strategy being a charging and discharging behavior sequence of each time window in the specified time period;

[0043] The energy storage system is charged and discharged according to the charging and discharging strategy.

[0044] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the following steps:

[0045] Acquire multi-source data of the target area power grid within a preset historical time period, and pre-process the multi-source data to obtain a feature vector;

[0046] Performing feature extraction processing on the feature vector using a pre-trained deep learning model to obtain a prediction result of electricity price volatility of the power grid within a specified time period; the electricity price volatility prediction result includes an electricity price volatility index of each node in the power grid at different time points, and the electricity price volatility index includes at least one of a maximum price difference, a variance, or a standard deviation;

[0047] Based on the electricity price volatility prediction result and the energy storage system parameters of the power grid, an optimization algorithm is used to generate a charge and discharge strategy for the energy storage system; the charge and discharge strategy is a sequence of charge and discharge behaviors in each time window within the specified time period;

[0048] The energy storage system is charged and discharged according to the charging and discharging strategy.

[0049] The above-mentioned charging and discharging management method, device, computer equipment, computer-readable storage medium and computer program product obtain multi-source data of a preset historical time period of the target area power grid and pre-process it to obtain a feature vector, and use a pre-trained deep learning model to extract features to obtain electricity price volatility prediction results including volatility indicators such as the maximum price difference, variance or standard deviation of each node at different time points, and then use an optimization algorithm in combination with the energy storage system parameters to generate a charging and discharging strategy for the charging and discharging behavior sequence of each time window within the specified time period, thereby realizing the charging and discharging management of the energy storage system; it can systematically analyze the long-term trend, periodic change law and coupling relationship of influencing factors of electricity price fluctuations, make up for the defect of traditional methods in not reflecting the overall picture of electricity price fluctuations, and can accurately capture the characteristics of electricity price fluctuations, provide effective support for the construction planning of new energy power stations, the capacity configuration of energy storage systems and the formulation of charging and discharging strategies, improve the operating economy and reliability of energy storage systems in complex power market environments, and better adapt to the changes in the power system operating environment brought about by the large-scale grid connection of new energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the following will briefly introduce the drawings needed to be used in the description of the embodiments of the present application or the related art. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 An application environment diagram of the charge and discharge management method in an embodiment;

[0052] Figure 2 A flowchart of the charge and discharge management method in an embodiment;

[0053] Figure 3 A flowchart of the training method of the deep learning model in an embodiment;

[0054] Figure 4 A flowchart of the method for generating the charge and discharge strategy in an embodiment;

[0055] Figure 5 A structural block diagram of the charge and discharge management device in an embodiment;

[0056] Figure 6 An internal structure diagram of the computer device in an embodiment. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solutions and advantages of the present application more clear, the following will further describe the present application in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0058] It should be noted that the terms "first", "second", and the like used in the present application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "include" and "have" and any variations thereof used in the present application are intended to cover non-exclusive inclusion. The term "multiple" used in the present application refers to two and more than two. The term "and / or" used in the present application refers to one of the solutions, or any combination of multiple solutions.

[0059] The charge and discharge management method provided by the embodiments of the present application can be applied to, for example, Figure 1The application environment is shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data required by the server 104 to process. The data storage system can be integrated on the server 104, or placed on the cloud or other network servers. The terminal 102 can be used to obtain the multi-source data of the target regional power grid in the preset historical time period, and perform feature extraction processing on the multi-source data to obtain the price volatility prediction result of the power grid in the specified time period. Among them, the terminal 102 can be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, unmanned aerial vehicles and low-altitude aircraft. The server 104 can be a stand-alone physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0060] In an exemplary embodiment, as shown in Figure 2 A charge and discharge management method is provided, and the method is applied to the terminal in Figure 1 for example, including the following steps 201 to 204. Among them:

[0061] Step 201, obtaining the multi-source data of the target regional power grid in the preset historical time period, and pre-processing the multi-source data to obtain the feature vector.

[0062] Among them, the target area refers to a specific power grid area, such as the power grid of a certain province; for example, the target area can be the power grid of A province. The preset historical time period is the historical time period for obtaining multi-source data; for example, it can be multi-source data in the past year. The feature vector is a vector that can reflect the characteristics of the data after preprocessing the multi-source data. For example, the processed load data, coal price data and the like are converted into a feature vector.

[0063] The multi-source data is data containing multiple factors affecting the electricity price; for example, it can be historical electricity prices, load curves, coal price information, tie-line capacity, network topology, time characteristics, seasons, holidays, target regional gross domestic product (GDP) forecast values, newly built power plants, industrial park information, and new energy output data in each node of the target provincial power grid. Among them, the historical electricity prices of each node refer to the electricity trading price data of different geographical location nodes in the target regional power grid in the past time period, reflecting the electricity price level and fluctuation at different time points of each node, and are a direct reference for electricity price prediction. The load curve represents the power load change trend of each node in the target regional power grid within a certain time period (such as day, week, month), presented in the form of a curve with time as the horizontal axis and load as the vertical axis, embodying the regularity and volatility of electricity demand, such as peak load during morning and evening peak. The coal price information refers to the price data of the coal market in the target region, including the purchase price of different coal types, transportation costs, etc. Since coal-fired power accounts for a high proportion in the power structure, fluctuations in coal prices directly affect power generation costs, which in turn are transmitted to node electricity prices. Tie-line capacity refers to the maximum transmission power limit of the transmission line connecting the target regional power grid with other regional power grids, reflecting the inter-regional power exchange capacity. Tie-line capacity limitations can affect power supply and demand balance, and in turn affect node electricity prices, such as when there is a power shortage in the region, insufficient tie-line transmission capacity may lead to an increase in electricity prices. The network topology refers to the physical connection relationship of the target regional power grid, including the connection mode and layout between substations, transmission lines, and nodes. The network topology determines the power transmission path and loss, affecting the price difference of each node, such as remote nodes that may have higher electricity prices due to high transmission losses. The time characteristics include specific time (such as 0:00-23:00 of a day), week, month, and other time dimension information. Electricity demand and power generation costs differ in different time periods, such as the electricity usage patterns differ between weekdays and weekends, which is a key time dimension factor for electricity price prediction. Seasons refer to the division of natural seasons into spring, summer, autumn, and winter. Seasonal changes affect electricity demand and power resources, such as increased air conditioning load in summer and high water power output during the wet season, which in turn affect the periodic characteristics of electricity price fluctuations. Holidays include statutory holidays and special dates. Industrial electricity consumption decreases and residential electricity consumption changes during holidays, leading to abnormal load curves and short-term impacts on node electricity prices, such as lower holiday electricity prices than weekday prices. The target regional GDP forecast value refers to the forecast data of the target regional future economic aggregate. GDP growth is usually accompanied by an increase in electricity demand, reflecting the driving effect of macroeconomics on the power market, and is an important reference for long-term electricity price trend prediction. Newly built power plants refer to the construction information of newly added power facilities in the target region, such as thermal power plants, wind farms, and photovoltaic power stations, including commissioning time, installed capacity, energy type, etc. The input of newly built power plants will change the power supply structure, affect the regional power supply and demand balance, and in turn affect the node electricity price.The industrial park information refers to the distribution, scale, main industry type and power demand characteristics of the newly built or existing industrial parks in the target area. The power consumption of industrial parks is concentrated and large in quantity, and the planning and development of industrial parks will directly affect the regional load distribution and electricity price level. The new energy output data refers to the real-time or historical output power data of renewable energy power generation such as wind power and photovoltaic power in the target area. The output of new energy is random due to weather influence, and the expansion of the grid-connected scale increases the volatility of the power system, which is a key uncertain factor that needs to be considered in node price prediction.

[0064] In one embodiment, taking the power grid of A province as an example, multi-source data of the power grid in the past year is obtained, including historical electricity prices, load curves, coal price information, tie-line capacity, network topology structure, time characteristics, season, holidays, GDP prediction values, and relevant data of newly built power plants and industrial parks. Then, the data is preprocessed, such as cleaning, denoising, and normalization, to finally obtain feature vectors that can represent the characteristics of the data.

[0065] In the above embodiment, by obtaining multi-source data of the target area power grid in a preset historical time period and preprocessing to obtain feature vectors, various factors that affect the electricity price can be comprehensively collected, providing rich and accurate data basis for subsequent electricity price volatility prediction, and ensuring the accuracy and reliability of the prediction results.

[0066] Step 202, performing feature extraction processing on the feature vectors by the pre-trained deep learning model to obtain the electricity price volatility prediction results of the power grid in a specified time period.

[0067] The pre-trained deep learning model is a deep learning model trained in advance using a large amount of sample data. For example, a hybrid model combining Convolutional Neural Network (CNN) and Long Short-Term Memory Network (LSTM), or a hybrid model combining CNN and Transformer. For example, the hybrid model composed of CNN and LSTM is trained using sample data, so that the hybrid model has feature extraction and time series prediction capabilities.

[0068] The electricity price volatility prediction result is an index reflecting the electricity price volatility of each node in the power grid at different time points obtained by model processing, including maximum price difference, variance or standard deviation, etc. For example, the maximum price difference, variance, etc. of a node in the target area power grid in the next month is predicted.

[0069] In an exemplary embodiment, taking the target regional power grid as an example, a pre-trained CNN and LSTM hybrid deep learning model is used to process the pre-processed feature vectors. The CNN part of the hybrid deep learning model extracts the spatial features of the load curve, new energy output and other data, and the LSTM part captures the long-term dependence relationship in the time series data, thereby predicting the price volatility of each node in the provincial power grid at different time points in the next month, and obtaining price volatility prediction results such as the maximum price difference and variance of each node per day.

[0070] In an embodiment, the feature vectors are processed by a pre-trained deep learning model to obtain price volatility prediction results including volatility indicators such as the maximum price difference and variance of each node at different time points. This can make full use of the advantages of deep learning models in feature extraction and time series analysis, effectively capture the nonlinear, non-stationary characteristics and long-term dependence relationship in price volatility, accurately predict the long-term trend, periodic variation law and coupling relationship of each influencing factor of price volatility, and make up for the shortcomings of traditional methods in reflecting the overall picture of price volatility. It provides accurate price volatility information support for subsequent development of energy storage system charging and discharging strategies based on price volatility prediction results, evaluation of energy storage system economy and feasibility, and improves the prediction ability and accuracy of price volatility in complex power market environment.

[0071] Step 203, based on the price volatility prediction results and the energy storage system parameters of the power grid, an optimization algorithm is used to generate the charging and discharging strategy of the energy storage system.

[0072] The energy storage system refers to a device installed in the power grid for storing and releasing electrical energy, which can adjust power supply and demand through charging and discharging. For example, a lithium battery energy storage system configured in the power grid. The energy storage system parameters of the power grid represent various data characterizing the characteristics of the energy storage system, including capacity, maximum charging and discharging power, efficiency, initial power, etc. For example, the capacity of a certain energy storage system is 100 megawatt-hours (MWh), the maximum charging and discharging power is 50 megawatts (MW), and the charging and discharging efficiency is 90%. The optimization algorithm is an intelligent algorithm used to solve the optimal charging and discharging strategy, such as genetic algorithm, particle swarm optimization algorithm, etc. For example, a genetic algorithm can be used to optimize the charging and discharging behavior of the energy storage system. The charging and discharging strategy is a sequence of charging and discharging behavior in each time window within a specified time period, which clearly specifies the charging / discharging state and power of the energy storage system in each time window. For example, in the next 24 hours, each hour is a time window, and the charging and discharging strategy can be: 0:00-1:00 charge 20MW, 1:00-2:00 no charge and no discharge, …, 23:00-24:00 discharge 30MW.

[0073] In one embodiment, taking a lithium battery energy storage system of a certain power grid as an example, its parameters are known, such as capacity of 100 MWh, maximum charge and discharge power of 50 MW, efficiency of 90%, etc. Based on the prediction results of the price volatility of the power grid in the next month, such as the maximum price difference and variance of each node per day, etc., a genetic algorithm is used for optimization to generate the charge and discharge behavior sequence of the energy storage system in the next month, i.e. the charge and discharge strategy. For example, charge in the low price period (such as early morning) and discharge in the high price period (such as evening) to realize peak-valley arbitrage.

[0074] In the above embodiment, based on the price volatility prediction results and the parameters of the energy storage system of the power grid, the charge and discharge strategy of the charge and discharge behavior sequence of each time window in the specified period is generated by using an optimization algorithm, which can deeply combine the price volatility characteristics with the characteristics of the energy storage system, and solve the optimal charge and discharge scheme considering economy and feasibility through an intelligent optimization algorithm. On the one hand, the price volatility index is used to accurately capture the peak-valley price difference to guide the energy storage system to charge in the low price period and discharge in the high price period, maximizing the spot price difference income; on the other hand, the parameters of the energy storage system are combined to ensure the operability of the strategy, avoiding excessive charging and discharging to damage the energy storage system equipment or violate the physical constraints. The charge and discharge strategy generation method can effectively improve the operation economy of the energy storage system in the complex power market, provide a scientific basis for capacity configuration and charge and discharge scheduling of the energy storage system, and realize the optimized operation of the power system under large-scale grid connection of new energy.

[0075] Step 204, charging and discharging management of the energy storage system according to the charge and discharge strategy.

[0076] Exemplarily, taking the energy storage system of a certain provincial power grid as an example, it is assumed that the charge and discharge strategy for the next 7 days is generated through the previous steps, which divides each day into 24 time windows (each hour as a window) and clearly indicates the charge and discharge behavior of each time window. For example, in the early morning of 2:00-5:00 when the price is low, the charge and discharge strategy instructs the energy storage system to charge at a power of 30 MW; in the evening of 18:00-20:00 when the price is high, to discharge at a power of 40 MW; and in other periods, to perform corresponding charging, discharging or standby operations according to the price volatility and the state of the energy storage system. The energy storage management system automatically controls the charge and discharge equipment of the energy storage system in each time window according to the charge and discharge strategy, executes charging, discharging or stopping actions, and monitors the parameters such as power and capacity of the energy storage system in real time to ensure that the charging and discharging process meets the requirements of the charge and discharge strategy and the safety operation constraints of the energy storage system.

[0077] In the above embodiments, by managing the charging and discharging of the energy storage system according to the generated charging and discharging strategy, the charging and discharging strategy optimized based on the prediction of price volatility and the characteristics of the energy storage system can be actually implemented, realizing the automatic and intelligent operation of the energy storage system in the electricity market. By accurately implementing the charging and discharging strategy, the price peak-valley difference can be effectively utilized for arbitrage, maximizing the economic benefits of the energy storage system; at the same time, by monitoring the real-time power and power of the energy storage system during the strategy implementation process, the energy storage system can be ensured to operate under safe constraint conditions, avoiding abnormal conditions such as overcharging and overdischarging, and improving the operation reliability and service life of the energy storage system. In addition, the management process can be matched with the actual operation state of the power system in real time, adapt to the complex environment of new energy output randomness and price volatility, provide strong support for the stable operation of the power system, and help realize the optimal scheduling and efficient operation of the power system.

[0078] The charging and discharging management method described above, by obtaining the multi-source data of the target regional power grid in a preset historical time period and preprocessing to obtain a feature vector, using a pre-trained deep learning model to extract features to obtain a price volatility prediction result containing volatility indicators such as maximum price difference, variance or standard deviation of each node at different time points, and then using an optimization algorithm to generate a charging and discharging strategy for the charging and discharging behavior sequence of each time window in a specified time period, the charging and discharging management of the energy storage system is realized; it can systematically analyze the long-term trend, periodic change rule and coupling relationship of influencing factors of price volatility, make up for the defects of traditional methods that cannot fully reflect the overall situation of price volatility, accurately capture the characteristics of price volatility, provide effective support for new energy power station construction planning, energy storage system capacity configuration and charging and discharging strategy formulation, and improve the operation economy and reliability of the energy storage system in the complex electricity market environment, better adapt to the changes in the operation environment of the power system brought by large-scale grid connection of new energy.

[0079] In an exemplary embodiment, the pre-trained deep learning model is a hybrid model combining CNN and LSTM, or a hybrid model combining CNN module and Transformer. By using the pre-trained deep learning model to extract features from the feature vector, the price volatility prediction result of the power grid in a specified time period can include: extracting spatial features in the feature vector through the CNN module, modeling the long-term dependence relationship in the time series data in the multi-source data through LSTM or Transformer, to capture and analyze the inherent nonlinear characteristics and non-stationary characteristics in the price volatility of the power grid in the specified time period, thereby forming a price volatility curve, and combining the spatial features and the price volatility curve to obtain the price volatility prediction result.

[0080] The nonlinear characteristic refers to the fact that the electricity price fluctuation and the influencing factors such as the load curve, coal price information, and new energy output data do not have a simple linear mapping relationship, but have a complex coupling effect. For example, when the grid load reaches the peak, the electricity price may rise exponentially rather than being proportional to the load growth. The randomness of new energy output such as wind power and photovoltaic power can cause irregular jumps in electricity price in a short period of time, which cannot be accurately described by a linear model. This nonlinear characteristic makes it difficult to fit the electricity price fluctuation with a simple linear equation, and a deep learning model needs to be used to capture the nonlinear mapping rule.

[0081] The non-stationary characteristic refers to the fact that the statistical characteristics of the electricity price fluctuation such as the mean, variance, and periodicity change over time, and do not meet the conditions of a stationary random process. For example, the electricity price fluctuation amplitude during the summer high-temperature period is usually greater than that during the spring and autumn seasons, that is, the variance changes with the season. After large-scale grid connection of new energy, the traditional peak-valley period periodicity of electricity price may be broken, and the time-varying characteristic is presented. Due to the existence of the non-stationary characteristic, the traditional time series model based on the stationary assumption cannot accurately predict the electricity price, while the LSTM or Transformer can capture the long-term dependence relationship in the multi-source data through the memory unit or attention mechanism, and adapt to the time-varying characteristic of the electricity price.

[0082] The deep learning model can effectively analyze the nonlinear and non-stationary characteristics in the electricity price fluctuation by extracting spatial features such as load distribution and network topology and modeling long-term dependence relationships of time series, thereby generating more accurate electricity price fluctuation prediction results and providing a scientific basis for formulating the charging and discharging strategy of the energy storage system.

[0083] In an exemplary embodiment, Figure 3 The flowchart of the training method of the deep learning model in one embodiment is shown in FIG. 1. Figure 3 The training method of the deep learning model shown in FIG. 1 involves a specific implementation of training the deep learning model through historical multi-source sample data. This method is applied to the application environment in Figure 1 For example, the following steps 301 to 304 are included. Among them:

[0084] Step 301: Obtain historical multi-source sample data of the target regional power grid and the corresponding actual electricity price fluctuation sequence.

[0085] The historical multi-source sample data refers to sample data containing various factors affecting the electricity price in the target regional power grid over a period of time in the past. For example, the historical electricity price, load curve, coal price information, tie-line capacity, network topology structure, time characteristics, season, holiday, GDP forecast value, new power plant, industrial park information, and new energy output data of each node of a provincial power grid in the past three years.

[0086] The actual price fluctuation sequence is a sequence formed by arranging the actual price fluctuation of the corresponding historical multi-source sample data in chronological order; for example, the sequences of the actual maximum price difference, variance or standard deviation of each node of the provincial power grid at each time point in the past three years are arranged in chronological order.

[0087] For example, the historical multi-source sample data of a provincial power grid in the past three years is obtained, including the historical price of each node, the load curve of different time periods, the coal price information of each period, the capacity change of the tie line, the network topology data, the specific time characteristics (such as time, week, month, etc.), seasonal division, holiday information, GDP prediction value, the commissioning time and installed capacity of newly built power plants, the power demand characteristics of industrial parks, and the output data of new energy such as wind power and photovoltaic power. At the same time, the actual price fluctuation sequence of each node of the provincial power grid at each time point in the past three years corresponding to these historical multi-source sample data is obtained, such as the sequence of the actual maximum price difference, variance, etc. of each node per day arranged in chronological order.

[0088] Step 302, identifying the historical multi-source sample data by a preset initial model to obtain a predicted price fluctuation sequence.

[0089] The preset initial model refers to a deep learning model that is set in advance and has not been fully trained, which is used to preliminarily process the historical multi-source sample data; for example, it can be a hybrid model combining CNN and LSTM, or a hybrid model combining CNN module and Transformer. For example, a CNN+LSTM hybrid model with random initial parameters, in which CNN is used to extract spatial features and LSTM is used to process time series.

[0090] The predicted price fluctuation sequence refers to a sequence formed by arranging the prediction result of the price fluctuation of the corresponding historical time period after processing the historical multi-source sample data by the preset initial model in chronological order, which contains the price fluctuation index of each node at different time points; for example, the predicted maximum price difference sequence of each node per day of the provincial power grid corresponding to the historical time period is obtained by processing the historical multi-source sample data of the provincial power grid by the initial model.

[0091] Exemplarily, taking a provincial power grid as an example, a preset initial model combining CNN and LSTM is selected, and historical multi-source sample data of the power grid in the past three years is input into the preset initial model. The CNN module in the preset initial model extracts spatial features in the historical multi-source sample data, such as load distribution and network topology structure features, and the LSTM module models long-term dependence relationships in time series data, identifies and processes the historical data, thereby obtaining a predicted price fluctuation sequence of the provincial power grid in the corresponding time period, such as a sequence of predicted maximum price difference, variance and the like of each node at each time point in the past three years arranged in time sequence. In an embodiment, the mathematical formula of the prediction process can be expressed as:

[0092] Volatilityt,i=DLP(Featurest,i)

[0093] wherein Volatilityt,i represents the price volatility of node i at time t, DLP represents a deep learning prediction function of the preset initial model, and Featurest,i represents a feature vector of all historical multi-source sample data.

[0094] In step 303, a loss function is determined according to the actual price fluctuation sequence, the predicted price fluctuation sequence, the parameters of the preset initial model and the regularization coefficient.

[0095] The parameters of the preset initial model refer to parameters such as weights, biases and the like that can be adjusted in the preset initial model, such as convolution kernel parameters of CNN, gate unit parameters of LSTM and the like; for example, the size and number of convolution kernels in CNN, the number of neurons in the LSTM hidden layer and the like.

[0096] The regularization coefficient is a hyperparameter used to prevent overfitting of the preset initial model, and the complexity of the model parameters is penalized in the loss function; for example, when an L2 regularization term is added to the loss function, the regularization coefficient λ is used to control the penalty strength.

[0097] The loss function is a function used to measure the difference between the predicted price fluctuation sequence and the actual price fluctuation sequence, and is constructed in combination with the parameters of the preset initial model and the regularization coefficient, which is the optimization target of the model; for example, a mean square error (MSE) loss function is used to calculate the average of the sum of squares of the difference between the predicted value and the actual value.

[0098] In an embodiment, the calculation formula of the loss function includes:

[0099] L=min t,i ∑(Volatility t,i pred-Volatility t,i actual)2+λ||θ||2

[0100] wherein L is a loss function, Volatility t,i pred is a predicted electricity price volatility sequence, Volatility t, i actual is an actual electricity price volatility sequence, λ is a regularization coefficient, θ is a parameter of a preset initial model, min t,i is a function of taking the minimum value.

[0101] Step 304, the parameters of the preset initial model are optimized iteratively by using a back propagation algorithm and a gradient descent method until the loss function meets a preset condition.

[0102] The back propagation algorithm is an algorithm for optimizing the parameters of a neural network, which calculates the gradient of the loss function with respect to the parameters of each layer, propagates from the output layer to the input layer, and guides the updating direction of the parameters; for example, in a CNN+LSTM hybrid model, the gradients of the parameters of each layer such as the convolution layer and the LSTM layer are calculated by back propagation.

[0103] The gradient descent method is an optimization method for iteratively updating the parameters of the model based on the gradient direction, which adjusts the parameters in the direction of the gradient descent of the loss function to minimize the loss function; for example, in each iteration, the weight parameters of the model are updated in the opposite direction of the gradient, and the step size is determined by the learning rate.

[0104] The preset condition is a condition set in advance to determine whether the optimization of the model stops, such as the value of the loss function being less than a certain threshold, the number of iterations reaching a maximum value, or the decrease in the loss function being less than a set value. For example, the optimization stops when the value of the loss function is less than 0.03 or the number of iterations reaches 800.

[0105] In one embodiment, taking the CNN+LSTM hybrid model as an example, after the loss function is determined, the back propagation algorithm is used to calculate the gradient of the loss function with respect to the parameters of the preset initial model, and then the gradient descent method is used to update the parameters of the model in the opposite direction of the gradient, such as subtracting the product of the learning rate and the gradient from the convolution kernel weight. The value of the loss function is recalculated after each iteration and compared with the preset condition, such as the loss function being less than 0.03 or the number of iterations not exceeding 800. If the loss function does not meet the condition, the optimization continues to iterate; if the loss function is less than 0.03 or the number of iterations reaches 800, the optimization stops, and the optimized model parameters are obtained, which enable the model to more accurately predict the electricity price volatility sequence.

[0106] In one exemplary embodiment, Figure 4 is a flowchart of the method for generating the charging and discharging strategy in one embodiment, as Figure 4 The method for generating the charging and discharging strategy involves using an optimization algorithm to generate the charging and discharging strategy of the energy storage system based on the electricity price volatility prediction result and the parameters of the energy storage system of the power grid, and the specific implementation is described in the aboveFigure 2 For aspects of the embodiments that have not been described in detail, refer to the description in Figure 2 and apply the method to Figure 1 application environments. The following describes the method in the application environment, including the following steps 401 to 405. Among them:

[0107] Step 401, based on the power and capacity constraints in the energy storage system parameters, a set of initial individuals representing the charging and discharging strategy are randomly generated.

[0108] Among them, the initial individual code is a set of codes randomly generated based on the power and capacity constraints in the energy storage system parameters, used to represent the charging and discharging strategy of each time window in a specified time period. Each initial individual code corresponds to a charging and discharging plan, and meets the power and capacity constraints of the energy storage system. For example, the charging and discharging behavior is represented in the form of binary code or real number code, where each code bit corresponds to the charging and discharging state and power size of each time window. Each initial individual code represents the charging and discharging plan of each time window in a specified time period.

[0109] In one embodiment, a set of initial individual codes representing the charging and discharging strategy are randomly generated according to the power and capacity constraints in the energy storage system parameters, and each individual defines the charging and discharging action of each time window in the specified time period.

[0110] Step 402, for each initial individual code, the price fluctuation prediction result determines the charging and discharging price difference income, the energy storage system loss cost is determined according to the energy storage system parameters, and the fitness value of the initial individual code is obtained according to the charging and discharging price difference income and the loss cost.

[0111] Among them, the fitness value is a numerical value used to evaluate the pros and cons of the initial individual code, which is obtained by determining the charging and discharging price difference income through the price fluctuation prediction result, determining the energy storage system loss cost according to the energy storage system parameters, and subtracting the loss cost from the charging and discharging price difference income. The higher the fitness value, the better the economic benefit of the charging and discharging strategy corresponding to the initial individual code. For example, the charging and discharging price difference income of a certain initial individual code in a specified time period is 100,000 yuan, and the energy storage system loss cost is 20,000 yuan, then its fitness value is 80,000 yuan.

[0112] Step 403, according to the fitness value, the initial individual code is selected to perform roulette selection, single-point crossover and probability mutation operations to generate a new sub-population that meets the energy storage system parameter constraints.

[0113] Wherein, the population is a set of coded initial individuals. The new offspring population is a set of new coded individuals generated after performing roulette selection, single-point crossover and probability mutation operations on coded initial individuals with fitness values greater than a specified threshold, which meet the energy storage system parameter constraints. The new offspring population inherits part of the excellent characteristics of the parent population and introduces new possibilities through mutation to better search for the optimal charging and discharging strategy. For example, individuals with high fitness values are selected from the parent population, part of the code bits are exchanged through crossover, and some code bits are mutated with a certain probability to generate multiple new coded individuals, which form the new offspring population.

[0114] Wherein, the roulette selection operation refers to screening individuals with higher fitness values according to the fitness values using the roulette strategy; the single-point crossover operation is to generate new offspring by performing single-point crossover on selected individuals; and the probability mutation operation is to mutate the offspring individuals according to the probability and verify the energy storage parameter constraints.

[0115] Step 404, taking the new offspring population as the next round population, repeating the generation of fitness values and new offspring populations until the fitness value change rate is lower than the preset threshold or the maximum iteration number is reached.

[0116] Wherein, the preset threshold is one of the conditions set in advance for judging whether to stop iteration to generate new offspring populations, i.e., the threshold of the fitness value change rate. When the fitness value change rate is lower than the preset threshold, it indicates that the fitness value of the population tends to be stable, and further iteration does not obviously improve the optimization result, so the iteration can be stopped. For example, the preset threshold is 0.5%, and when the maximum fitness value change rate of the populations of adjacent two iterations is less than 0.5%, the iteration is stopped.

[0117] In one embodiment, steps 402 and 403 are repeatedly performed to repeatedly generate fitness values and new offspring populations until the fitness value change rate is lower than the preset threshold or the maximum iteration number is reached.

[0118] Step 405, selecting the coded initial individual with the highest fitness value in the final population and decoding it into the charging and discharging strategy.

[0119] Wherein, the final population is the last generation population obtained after several rounds of iteration optimization, which is a set of coded individuals; for example, the population composed of multiple coded individuals obtained after 100 iterations.

[0120] In an embodiment, the final population is obtained when the iteration process meets the stop condition that the rate of change of the fitness value is lower than the preset threshold or the maximum number of iterations is reached. The initial individual with the highest fitness value is found in the final population, such as the individual with a fitness value of 35000 yuan. The initial individual is decoded, assuming that it uses real number coding, and each coding bit corresponds to the charging and discharging power of a time window. The charging and discharging power of each time window in the next 24 hours is obtained after decoding, such as charging 40 MW in the first hour, charging 35 MW in the second hour, discharging 20 MW in the third hour, etc., to form a specific charging and discharging strategy. The strategy is the optimal charging and discharging strategy obtained by the genetic algorithm and can be used for charging and discharging management of the energy storage system.

[0121] In the above embodiments, the fluctuation characteristics such as the price peak-valley difference are accurately captured based on the price fluctuation prediction result, the constraint conditions of the grid energy storage system parameters are combined, and the charging and discharging strategy is iteratively optimized by means of selection, crossover, and mutation operations of the genetic algorithm. The charging and discharging price difference income can be maximized and the loss cost can be minimized under the premise of meeting the physical limitations of the energy storage system.

[0122] It should be understood that although each step in the flowchart involved in each of the above embodiments is displayed in sequence according to the arrow, these steps are not necessarily executed in sequence according to the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other sequences. Moreover, at least part of the steps in the flowchart involved in each of the above embodiments can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be alternately executed with at least part of other steps or steps or stages in other steps. It can be understood that the steps in different embodiments can be freely combined as needed, and various non-contradictory schemes formed by the combination are within the scope of protection of the present application.

[0123] Based on the same inventive concept, the embodiments of the present application also provide a charging and discharging management device for implementing the above-mentioned charging and discharging management method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more charging and discharging management device embodiments provided below can refer to the limitations of the charging and discharging management method described above, which will not be repeated here.

[0124] In an exemplary embodiment, as shown in Figure 5 a charging and discharging management device 500 is provided, which includes a data acquisition module 510, a result prediction module 520, a strategy generation module 530, and a charging and discharging management module 540, wherein:

[0125] The data acquisition module 510 is configured to acquire multi-source data of the target regional power grid in a preset historical time period, and obtain a feature vector by preprocessing the multi-source data.

[0126] The result prediction module 520 is configured to perform feature extraction processing on the feature vector by using a pre-trained deep learning model, and obtain a price fluctuation prediction result of the power grid in a specified time period; the price fluctuation prediction result includes price fluctuation indexes of each node in the power grid at different time points, and the price fluctuation indexes include at least one of a maximum price difference, a variance or a standard deviation.

[0127] The strategy generation module 530 is configured to generate a charging and discharging strategy of the energy storage system based on the price fluctuation prediction result and parameters of the energy storage system of the power grid, and by using an optimization algorithm; the charging and discharging strategy is a charging and discharging behavior sequence of each time window in the specified time period.

[0128] The charging and discharging management module 540 is configured to perform charging and discharging management on the energy storage system according to the charging and discharging strategy.

[0129] The charging and discharging management device described above, the data acquisition module 510 acquires multi-source data of a target regional power grid in a preset historical time period and obtains a feature vector by preprocessing, the result prediction module 520 extracts features by using a pre-trained deep learning model to obtain a price fluctuation prediction result including fluctuation indexes such as maximum price difference, variance or standard deviation of each node at different time points, the strategy generation module 530 generates a charging and discharging strategy of each time window in a specified time period by using an optimization algorithm in combination with parameters of the energy storage system, and then the charging and discharging management module 540 performs charging and discharging management on the energy storage system; the long-term trend, periodic variation law and coupling relationship of influencing factors of price fluctuation can be systematically analyzed, the defects of the traditional method that the overall situation of price fluctuation is not reflected are made up, the characteristics of price fluctuation can be accurately captured, effective support is provided for new energy power station construction planning, energy storage system capacity configuration and charging and discharging strategy formulation, and the operation economy and reliability of the energy storage system in a complex power market environment are improved, and the energy storage system better adapts to the change of the power system operation environment brought by large-scale grid connection of new energy.

[0130] In one of the embodiments, the data acquisition module 510 is further configured to perform cleaning, denoising and normalization preprocessing on the multi-source data to obtain preprocessed multi-source data.

[0131] The preprocessed multi-source data is converted into a feature vector.

[0132] In one of the embodiments, the result prediction module 520 is further configured to extract spatial features in the feature vector by using the convolutional neural network module.

[0133] modeling long-term dependencies in time series data in the multi-source data by the long short-term memory network or the Transformer to capture and analyze nonlinear and non-stationary characteristics inherent in price fluctuation of the power grid within the specified time period, thereby forming a price fluctuation curve;

[0134] combining the spatial features and the price fluctuation curve to obtain the price fluctuation prediction result.

[0135] In one of the embodiments, the strategy generation module 530 is further configured to generate the charging and discharging strategy of the energy storage system based on the price fluctuation prediction result and the energy storage system parameters of the power grid by using a genetic algorithm in the optimization algorithm.

[0136] In one of the embodiments, the strategy generation module 530 is further configured to randomly generate a group of initial individuals representing the charging and discharging strategy according to power and capacity constraints in the energy storage system parameters; each of the initial individuals represents the charging and discharging strategy of each time window within the specified time period.

[0137] For each of the initial individuals, the price fluctuation prediction result determines a charging and discharging price difference benefit, the energy storage system parameters determine an energy storage system loss cost, and the initial individual's fitness value is obtained according to the charging and discharging price difference benefit and the loss cost.

[0138] According to the fitness value, the initial individuals are selected to perform roulette selection, single-point crossover and probabilistic mutation operations to generate a new child population that satisfies the energy storage system parameter constraints; the population is a set of the initial individuals.

[0139] The new child population is taken as the next round of population, and the fitness value and the new child population are repeatedly generated until the fitness value change rate is lower than a preset threshold or the maximum number of iterations is reached.

[0140] The initial individual with the highest fitness value in the final population is selected and decoded as the charging and discharging strategy.

[0141] In one of the embodiments, the data acquisition module 510 is further configured to acquire historical multi-source sample data and a corresponding actual price fluctuation sequence of the target regional power grid.

[0142] The historical multi-source sample data is identified by a preset initial model to obtain a predicted price fluctuation sequence.

[0143] According to the actual price fluctuation sequence, the predicted price fluctuation sequence, parameters of the preset initial model and a regularization coefficient, a loss function is determined.

[0144] The parameters of the preset initial model are iteratively optimized by using a back propagation algorithm and a gradient descent method until the loss function meets a preset condition.

[0145] The modules in the charge and discharge management apparatus can be implemented by software, hardware, or a combination thereof. The modules can be embedded in or independent of a processor in a computer device in hardware form, or stored in a memory in the computer device in software form, so as to be called and executed by the processor.

[0146] In an exemplary embodiment, a computer device, which can be a terminal, has an internal structure as shown in Figure 6 The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, the memory, and the input / output interface are connected through a system bus, and the communication interface, the display unit, and the input device are connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to perform wired or wireless communication with external terminals. The wireless communication can be achieved through WIFI, mobile cellular network, near field communication (NFC), or other technologies. The computer program is executed by the processor to implement a charge and discharge management method. The display unit of the computer device is configured to form a visually visible picture, which can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball, or touchpad arranged on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0147] Those skilled in the art can understand that Figure 6 The structure shown in the above description is only a block diagram of part of the structure related to the scheme of the present application, and does not limit the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0148] In an exemplary embodiment, a computer device is provided, comprising a memory and a processor, the memory storing a computer program, and the processor implementing the steps in the above method embodiments when executing the computer program.

[0149] In an exemplary embodiment, a computer readable storage medium is provided, storing a computer program, and the computer program implementing the steps in the above method embodiments when executed by a processor.

[0150] In an exemplary embodiment, a computer program product is provided, comprising a computer program, and the computer program implementing the steps in the above method embodiments when executed by a processor.

[0151] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0152] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. The non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. The volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, the RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., without being limited thereto.

[0153] The technical features of the above embodiments can be combined in any manner. To make the description concise, all possible combinations of the technical features in the above embodiments are not described, but as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the present application.

[0154] The above-described embodiments are merely illustrative of several embodiments of the present application, and the description is relatively specific and detailed, but should not be understood as a limitation on the scope of the patent. It should be noted that for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of the present application. Therefore, the scope of protection of the present application should be subject to the appended claims.

Claims

1. A charge and discharge management method, characterized in that: The method comprises: Acquire multi-source data of the target area power grid within a preset historical time period, and pre-process the multi-source data to obtain a feature vector; Performing feature extraction processing on the feature vector using a pre-trained deep learning model to obtain a prediction result of electricity price volatility of the power grid within a specified time period; the electricity price volatility prediction result includes an electricity price volatility index of each node in the power grid at different time points, and the electricity price volatility index includes at least one of a maximum price difference, a variance, or a standard deviation; Based on the electricity price volatility prediction result and the energy storage system parameters of the power grid, an optimization algorithm is used to generate a charge and discharge strategy for the energy storage system; the charge and discharge strategy is a sequence of charge and discharge behaviors in each time window within the specified time period; The energy storage system is charged and discharged according to the charging and discharging strategy.

2. The method according to claim 1, characterized in that The multi-source data includes at least one of historical electricity prices of each node in the power grid, load curves, coal price information, tie line capacity, network topology, time characteristics, seasons, holidays, target regional GDP forecasts, newly built power plants, industrial park information, and new energy output data; The preprocessing of the multi-source data to obtain a feature vector includes: Performing cleaning, denoising, and normalization preprocessing on the multi-source data to obtain preprocessed multi-source data; The preprocessed multi-source data is converted into a feature vector.

3. The method according to claim 1, characterized in that The pre-trained deep learning model includes: at least one of a convolutional neural network module, a long short-term memory network, and a transformer; the feature extraction processing performed on the feature vector by the pre-trained deep learning model to obtain a prediction result of the electricity price volatility of the power grid within a specified time period includes: Extracting spatial features from the feature vector using the convolutional neural network module; Modeling the long-term dependencies in the time series data in the multi-source data by the long short-term memory network or the transformer to capture and analyze the inherent nonlinear and non-stationary characteristics of the electricity price fluctuations of the power grid within the specified time period, thereby forming an electricity price fluctuation curve; The electricity price volatility prediction result is obtained by combining the spatial characteristics and the electricity price fluctuation curve.

4. The method according to claim 1, wherein The method of generating a charge and discharge strategy for the energy storage system using an optimization algorithm based on the electricity price volatility prediction result and the energy storage system parameters of the power grid includes: Based on the electricity price volatility prediction result and the energy storage system parameters of the power grid, a genetic algorithm in the optimization algorithm is used to generate a charging and discharging strategy for the energy storage system.

5. The method according to claim 4, characterized in that The step of using the genetic algorithm to generate a charge and discharge strategy for the energy storage system based on the electricity price volatility prediction result and the energy storage system parameters of the power grid includes: Based on the power and capacity constraints in the energy storage system parameters, a set of coded initial individuals representing the charge and discharge strategy is randomly generated; each of the coded initial individuals represents the charge and discharge strategy of each time window within the specified time period; For each of the encoded initial individuals, the electricity price volatility prediction result determines a charge-discharge price difference benefit, the energy storage system loss cost is determined according to the energy storage system parameters, and the fitness value of the encoded initial individual is obtained according to the charge-discharge price difference benefit and the loss cost; Selecting the encoded initial individuals according to the fitness value to perform roulette wheel selection, single-point crossover and probabilistic mutation operations to generate a new offspring population that meets the energy storage system parameter constraints; the population is a collection of the encoded initial individuals; Using the new offspring population as the next round population, repeatedly generating the fitness value and the new offspring population until the fitness value change rate is lower than a preset threshold or the maximum number of iterations is reached; The encoded initial individual with the highest fitness value in the final population is selected and decoded into the charging and discharging strategy.

6. The method according to claim 1, characterized in that The training method of the pre-trained deep learning model includes: Obtaining historical multi-source sample data of the target area power grid and the corresponding actual electricity price fluctuation series; Identifying the historical multi-source sample data through a preset initial model to obtain a predicted electricity price fluctuation sequence; Determining a loss function based on the actual electricity price fluctuation sequence, the predicted electricity price fluctuation sequence, the parameters of the preset initial model, and the regularization coefficient; The parameters of the preset initial model are optimized and iterated using the back propagation algorithm and the gradient descent method until the loss function meets the preset conditions.

7. A charge and discharge management device, characterized in that: The device comprises: A data acquisition module is used to acquire multi-source data of the target area power grid within a preset historical time period and pre-process the multi-source data to obtain a feature vector; a result prediction module, configured to perform feature extraction processing on the feature vector using a pre-trained deep learning model to obtain a prediction result of electricity price volatility of the power grid within a specified time period; the electricity price volatility prediction result includes an electricity price volatility index of each node in the power grid at different time points, the electricity price volatility index including at least one of a maximum price difference, a variance, or a standard deviation; a strategy generation module for generating a charge and discharge strategy for the energy storage system using an optimization algorithm based on the electricity price volatility prediction result and the energy storage system parameters of the power grid; the charge and discharge strategy is a sequence of charge and discharge behaviors for each time window within the specified time period; A charge and discharge management module is used to manage the charge and discharge of the energy storage system according to the charge and discharge strategy.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images