A Hydrogen Energy System Configuration Method, Device, Equipment and Storage Medium for Reducing Wind Turbine Power Deviation

Abstract

The present invention discloses a hydrogen energy system configuration method, device, equipment and storage medium for reducing the power deviation of a wind turbine. Existing methods usually do not regard the electrolytic hydrogen production system as a load capable of adapting to the volatility of renewable energy. The hydrogen energy system configuration method of the present invention includes: real-time comparison and dynamic compensation: comparing the predicted data of the wind turbine power with the real-time power of the wind turbine to obtain a real-time error, and then performing dynamic compensation configuration of the hydrogen energy system based on the real-time error; when the real-time power of the wind turbine is greater than the predicted data, the hydrogen energy system starts the electrolytic hydrogen production process to convert the excess electric energy into hydrogen energy; when the real-time power of the wind turbine is less than the predicted data, the hydrogen energy system discharges through a fuel cell to perform power compensation on the wind turbine unit; and complete the configuration of the hydrogen energy system capacity. The present invention enables the electrolytic hydrogen production system to be not only an energy storage link but also a dynamic load, ensuring the continuity and stability of power supply.

Description

Technical Field

[0001] The present invention belongs to the technical field of hydrogen energy systems, and relates to a method, device, equipment and storage medium for configuring a hydrogen energy system to reduce the power deviation of a fan. Background Art

[0002] In the process of addressing climate change and promoting energy transformation, the integration of microgrids and hydrogen energy is regarded as a solution with great potential. Microgrid technology provides a flexible and efficient way of energy management. Integrating electrolytic hydrogen production into an energy storage system can not only promote the utilization of renewable energy but also enhance the stability of the power grid. The electrolytic water hydrogen production technology combined with renewable energy sources such as wind energy and solar energy provides an energy solution with almost zero carbon emissions, which helps to reduce greenhouse gas emissions and improve the efficiency and sustainability of the energy system.

[0003] However, hydrogen storage systems also have some limitations. The storage and transportation of hydrogen are relatively difficult because hydrogen has a low density, an extremely low liquefaction temperature, and is prone to hydrogen embrittlement. In addition, the energy conversion efficiency of hydrogen energy storage systems is usually lower than that of battery energy storage systems. Nevertheless, compared with battery energy storage systems, hydrogen energy storage systems have obvious advantages in cross-seasonal, cross-regional, and large-scale storage. Hydrogen energy storage can achieve long-term energy storage and release, is suitable for large-scale applications, and hydrogen can be flexibly transported in various ways without being restricted by the power transmission and distribution network. These characteristics make hydrogen energy storage systems more advantageous than traditional battery energy storage systems in new energy systems, especially in scenarios that require long-term or large-scale energy storage.

[0004] Zheng Li et al. considered the scheduling optimization of the hydrogen industrial chain under different scenarios in the literature "Coordinated Control Scheme of a Hybrid Renewable Power System Based on Hydrogen Energy Storage". Combining the characteristics of each season, they analyzed the system revenue and operating costs under uncertainty, and achieved 100% wind and photovoltaic power consumption, zero carbon emissions, and economic stability of operation through a variety of energy coupling devices. Tian Tian et al. took a 300MW offshore wind farm as an example in the literature "Comparative Analysis of the Economics of Offshore Wind Power to Hydrogen Production Technology", established economic models such as the equivalent annual value of the total investment cost, equipment investment cost, and annual operation and maintenance cost, and analyzed the economics of three off-grid wind power to hydrogen production systems: the onshore hydrogen production system of offshore wind power, the hydrogen production on offshore platforms and hydrogen transportation by ships system, and the hydrogen production on offshore platforms and hydrogen transportation by pipelines system. Zhao L et al. developed a dynamic system model for a system that uses photovoltaic and wind power generation to supply power to a hydrogen refueling station using proton exchange membrane (PEM) electrolyzers and fuel cells in the literature "Dynamic Operation and Feasibility Study of a Self-sustainable Hydrogen Fueling Station Using Renewable Energy Sources", and also conducted cost and sensitivity analyses to evaluate the average hydrogen cost of different site designs. The wind power-hydrogen energy coupling system proposed by Qin Mengzhu et al. in the literature "Modeling and Simulation of Wind Power-Hydrogen Energy Coupling System" is a microgrid in AC form. The system consists of a doubly-fed wind turbine, a PEM electrolyzer, a PEM fuel cell, and ordinary AC loads. In the system, wind power and fuel cells can be connected to the grid separately, and the excess power can be consumed by the electrolyzer. The stored hydrogen can be converted into electrical energy by the fuel cell when needed; in the off-grid mode, the fuel cell is the main power source, and the system ensures the power supply of local important loads through the AC bus, and the electrolyzer stores the abandoned wind power. Each power generation unit and electrolyzer in the system must be controlled by a double converter for rectification and inversion for grid connection and power supply to AC loads, which will cause relatively large losses, and the electrolyzer only starts when there is power surplus in the system. The PV / SOFC hybrid power system proposed by Guo Wei et al. in the literature "Modeling and Performance Simulation Study of PV / SOFC Hybrid Power System" combines photovoltaic power generation with solid oxide fuel cells and electrolyzers. It is an off-grid system that also supplies power to loads, and the remaining electrical energy is used to produce hydrogen for energy storage, which is used as an intermediate link for the fuel cell to generate electrical energy.

[0005] As described above, it can be seen that the research on renewable energy systems combined with hydrogen energy storage can be divided into off-grid and grid-connected types, and the off-grid hydrogen production system can be further divided into direct hydrogen production systems, fuel cell / electrolyzer hybrid hydrogen production systems, and supercapacitor hybrid hydrogen production systems. In the research on systems involving hydrogen energy storage, most of them combine wind power or photovoltaic power with hydrogen energy storage alone. The electrolytic hydrogen production system is often used as an intermediate energy storage link to replace the traditional battery energy storage link and can achieve power feedback in combination with fuel cells, or as a supplement to the traditional energy storage link. However, these methods usually do not regard the electrolytic hydrogen production system as a load that can adapt to the volatility of renewable energy. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to overcome the defects of the above-mentioned prior art and provide a hydrogen energy system configuration method, device, equipment, and storage medium for reducing the power deviation of a wind turbine, which enables the electrolytic hydrogen production system to be not only an energy storage link but also a dynamic load, and can adjust the hydrogen production rate according to the volatility of wind power and photovoltaic power. When the renewable energy generates excess power, the electrolyzer can increase the hydrogen production speed and convert the excess electrical energy into hydrogen energy for storage. When the power demand is high or the output of renewable energy is low, the stored hydrogen energy is converted back into electrical energy through a fuel cell and supplied to the microgrid to ensure the continuity and stability of power supply.

[0007] In a first aspect, the present invention provides a hydrogen energy system configuration method for reducing the power deviation of a wind turbine, which includes:

[0008] Real-time comparison and dynamic compensation: Compare the predicted data of the wind turbine power with the real-time power of the wind turbine to obtain a real-time error, and then perform dynamic compensation configuration of the hydrogen energy system based on the real-time error. When the real-time power of the wind turbine is greater than the predicted data, the hydrogen energy system starts the electrolytic hydrogen production process and converts the excess electrical energy into hydrogen energy. When the real-time power of the wind turbine is less than the predicted data, the hydrogen energy system discharges through a fuel cell to compensate the power of the wind turbine unit.

[0009] Hydrogen energy system capacity configuration: Complete the configuration of the hydrogen energy system capacity according to the above data processing results.

[0010] The present invention can achieve converting the excess part into hydrogen energy for storage when the actual power of the wind turbine is higher than the predicted value, and then converting the hydrogen energy into electrical energy for supplementary output when the actual power of the wind turbine is lower than the predicted value, so as to realize an energy management configuration of a microgrid.

[0011] The present invention enables the electrolytic hydrogen production system to play a key role. It not only stores the electric energy generated by renewable energy (such as wind energy), but also can convert the stored hydrogen back into electric energy through a fuel cell when needed. This ability enables the microgrid to manage its energy more effectively, especially when the output of renewable energy is unstable. The electrolytic hydrogen production system helps to balance supply and demand and enhance the reliability and stability of the power grid.

[0012] Furthermore, the predicted data of the fan power is also subjected to day-ahead prediction before being compared with the real-time power of the fan. The day-ahead prediction uses a long short-term memory network to perform day-ahead prediction on the fan power to obtain a preliminary predicted value of the fan power.

[0013] Even further, on the basis of the day-ahead prediction, the preliminary predicted value of the fan power is further optimized within the day to obtain a predicted value optimized within the day, and then the predicted value optimized within the day is used as the predicted data of the fan power; the within-day optimization uses a rolling time window optimization strategy of machine learning.

[0014] Even further, the goal of within-day optimization is to minimize the prediction error. The within-day optimization corrects the preliminary predicted value of the fan power using a linear regression model with real-time data, and then the corrected predicted value is expressed as:

[0015] ,

[0016] wherein, represents the preliminary predicted value of the fan power, and are correction coefficients obtained by fitting the data within the rolling time window;

[0017] Prediction error minimization: within the rolling window, the corrected predicted value is minimized through a random forest algorithm; assuming the prediction error is defined as , represents the real-time power of the fan, then the optimization goal is to minimize the following loss function:

[0018] ,

[0019] A correction step size is set, and the prediction model is updated every . The updated prediction model makes a prediction again based on the latest corrected predicted value, and finally obtains a predicted value optimized within the day .

[0020] Furthermore, when the hydrogen energy system compensates the power of the wind turbine, the compensated output power P comp is determined by the following formula:

[0021] ,

[0022] Wherein, P comp represents the output power after compensation, k comp represents the error compensation coefficient, represents the real-time error.

[0023] Aiming at the problem that the data accuracy of wind turbines is lower than 85% after day-ahead prediction and intra-day optimization, the present invention proposes a method for power compensation of the wind turbine power grid by using a hydrogen energy system to improve the accuracy of prediction data. The error compensation coefficient is dynamically adjusted according to the current working state of the hydrogen energy system. When the error compensation coefficient is 1, full-power compensation is performed; when the error compensation coefficient is 0, no compensation is performed, so as to flexibly meet the power compensation requirements in different situations and optimize the operation efficiency of the hydrogen energy system and the prediction accuracy of wind turbines. The present invention adjusts the compensation intensity of the hydrogen energy system by setting the error compensation coefficient.

[0024] In some existing power compensation methods, a fixed compensation coefficient may be used, which cannot adapt to different power deviation situations. The dynamic error compensation coefficient of the present invention can avoid energy waste caused by over-compensation while ensuring the power compensation effect, and improve the operation efficiency and economy of the system.

[0025] Furthermore, according to the real-time power of the wind turbine and the predicted value after intra-day optimization the difference between them is used to calculate the real-time error , that is:

[0026] .

[0027] Still further, the process of capacity configuration of the hydrogen energy system is as follows:

[0028] 1) Analyze the maximum power error in historical data ;

[0029] 2) Capacity configuration formula: The calculation formula of the capacity C H2 of the hydrogen energy system is as follows:

[0030]

[0031] Wherein, T is the duration that needs to be compensated; is the efficiency of the electrolyzer; is the efficiency of the fuel cell;

[0032] 3) Optimal configuration: The capacity configuration of the hydrogen energy system is 30%-50% of the installed capacity of the wind farm;

[0033] 4) SOC control: The state of charge (SOC) of the hydrogen energy system is set to 20% - 80% to ensure the effective operation of the electrolyzer and fuel cell.

[0034] In a second aspect, the present invention provides a hydrogen energy system configuration device for reducing the power deviation of a wind turbine, which includes:

[0035] Real-time comparison and dynamic compensation unit: According to the predicted data of the wind turbine power, compare it with the real-time power of the wind turbine to obtain a real-time error, and then perform dynamic compensation configuration of the hydrogen energy system based on the real-time error; when the real-time power of the wind turbine is greater than the predicted data, the hydrogen energy system starts the electrolytic hydrogen production process to convert the excess electric energy into hydrogen energy; when the real-time power of the wind turbine is less than the predicted data, the hydrogen energy system discharges through the fuel cell to perform power compensation on the wind turbine unit;

[0036] Hydrogen energy system capacity configuration unit: Complete the configuration of the hydrogen energy system capacity according to the foregoing data processing results.

[0037] In a third aspect, the present invention provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the method are implemented.

[0038] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method are implemented.

[0039] The beneficial effects of the present invention are as follows: By predicting the output of wind power generation, the microgrid can adjust the configuration of the electrolytic hydrogen production system in advance to adapt to the expected changes in energy supply and demand; the present invention not only improves the energy utilization efficiency but also helps to reduce energy waste caused by the volatility of renewable energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0041] Figure 1 It is a specific simulation circuit topology diagram of the microgrid energy management configuration of the present invention;

[0042] Figure 2 It is a flowchart of a hydrogen system configuration for reducing the power deviation of a wind turbine according to the present invention;

[0043] Figure 3It is the waveform diagram of wind farm data in the application example of the present invention;

[0044] Figure 4 It is the wind curtailment / cutting load power diagram in the application example of the present invention;

[0045] Figure 5 It is the power prediction error diagram of DC side energy storage with given voltage deviation (available capacity is 0.15 pu) in the simulation example of the present invention;

[0046] Figure 6 It is the power prediction error diagram of DC side energy storage with given voltage deviation (available capacity is 0.38 pu) in the simulation example of the present invention;

[0047] Figure 7 It is the capacity and prediction accuracy curve diagram of DC side energy storage with given voltage deviation in the simulation example of the present invention;

[0048] Figure 8 It is the power prediction error diagram of DC side energy storage with given prediction error deviation (available capacity is 0.15 pu) in the simulation example of the present invention;

[0049] Figure 9 It is the power prediction error diagram of DC side energy storage with given prediction error deviation (available capacity is 0.38 pu) in the simulation example of the present invention;

[0050] Figure 10 It is the comparison diagram of energy storage capacity and prediction accuracy under two control strategies in the simulation example of the present invention;

[0051] Figure 11 It is the three-dimensional diagram of energy storage capacity - prediction error coefficient - accuracy in the simulation example of the present invention;

[0052] Figure 12 It is a schematic diagram of a logical structure of a computer device provided in an embodiment of the present invention. Detailed implementation manners

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0054] Embodiment 1

[0055] The application of electrolytic hydrogen production technology in microgrids, especially as part of an energy storage system, provides new solutions for grid stability and the integration of renewable energy. By predicting the output of wind power generation, the microgrid can adjust the configuration of the electrolytic hydrogen production system in advance to adapt to the expected changes in energy supply and demand. This method not only improves the energy utilization efficiency but also helps reduce energy waste caused by the volatility of renewable energy. The present invention proposes a method for configuring a hydrogen energy storage system to reduce the power deviation of wind turbines. When the actual output of the wind turbine is higher than the predicted value, the excess part can be converted into hydrogen energy for storage. When the actual output of the wind turbine is lower than the predicted value, the hydrogen energy is then converted back into electrical energy for supplementary output, thus realizing an energy management configuration for a microgrid. The specific simulation circuit topology is as Figure 1 shown.

[0056] Before describing a typical wind farm application scenario, the Energy Industry Standard Manual of the People's Republic of China: NB / T 10205-2019 is given. The formula for the comprehensive evaluation index of wind power prediction is shown in Formula (1):

[0057] (1)

[0058] In the formula, S represents the comprehensive evaluation result of wind power prediction; r 1 represents the monthly average accuracy rate, represents the weight of the monthly average accuracy rate; r 2 represents the monthly average qualification rate, is the weight of the monthly average qualification rate; R represents the monthly average correlation coefficient, represents the weight of the monthly average correlation coefficient; E Peak represents the monthly average positive deviation rate during the peak load period, represents the weight of the monthly average positive deviation rate during the peak load period; E Valley represents the monthly average negative deviation rate during the valley load period, represents the weight of the monthly average negative deviation rate during the valley load period; r HW represents the monthly average prediction accuracy rate during the high wind speed period, represents the weight of the monthly average prediction accuracy rate during the high wind speed period; r LW represents the monthly average prediction accuracy rate during the low wind speed period, represents the weight of the monthly average prediction accuracy rate during the low wind speed period; r 3 represents the comprehensive monthly reporting rate of various reported data.

[0059] And a weight coefficient table in the comprehensive evaluation formula of wind power prediction is given. The power dispatching agency can appropriately adjust the index weights according to actual needs, as shown in Table 1.

[0060] Table 1 Weight Table of Comprehensive Evaluation Index of Wind Power Prediction

[0061]

[0062] Accuracy: The monthly (annual) average accuracy is the arithmetic mean of the daily accuracies. The formula is as follows:

[0063] (2)

[0064] In the formula, represents the predicted power; represents the actual power.

[0065] The requirements for prediction accuracy by the criterion are as follows: The monthly average accuracy of short-term wind power prediction in a wind farm should not be lower than 80%, and the coincidence rate should be greater than 80%. The accuracy of a hydrogen energy system configuration method for reducing the power deviation of a wind turbine in the present invention reaches more than 85%, and the rated power of the hydrogen storage system configuration is not less than 30% of the total installed power of wind power generation. The continuous discharge time at the rated power should not be less than 1 h, so as to determine the specific capacity and power selection of the configured hydrogen energy system.

[0066] This embodiment is a hydrogen energy system configuration method for reducing the power deviation of a wind turbine. As Figure 2 shown, the steps are as follows:

[0067] 1. Day-ahead prediction: Perform day-ahead prediction on the wind turbine to obtain a preliminary predicted value of the wind turbine power. The present invention uses a long short-term memory network (LSTM) for wind turbine power prediction.

[0068] The long short-term memory network (LSTM) is a special recurrent neural network (RNN), which is good at processing sequence data and can capture long-term dependencies in time series. Therefore, it is suitable for wind turbine power prediction. The following are the specific steps for applying LSTM to wind turbine power prediction:

[0069] (11) Data preparation: Collect historical meteorological data such as wind speed, temperature, humidity, and air pressure, as well as corresponding wind power output data. The data set needs to be normalized to avoid biases in model training caused by data with different dimensions. Among them, the input data is to use variables such as historical wind speed, temperature, and humidity as the input data sequence, defined as X t ={x 1 ,x 2 ,…,x t Among them, x t represents the multi-dimensional meteorological data collected at time t.

[0070] (12) Model architecture: The architecture of the LSTM network usually includes multiple LSTM layers and fully connected layers. The LSTM layers are used to process time series input data and extract time-related features. The fully connected layer is used to perform regression processing on the extracted features and output the predicted value of the future wind power. The core of the cell state update of the LSTM is through the memory cell c t and the hidden state h t to transmit information. The key formulas of the LSTM are as follows:

[0071] (121) Forget gate f t : Determine the retention degree of the memory cell c t-1 at the current moment:

[0072] (3)

[0073] In the formula, W f is the weight matrix, b f is the bias, is the activation function (sigmoid).

[0074] (122) Input gate and candidate state are used to update the cell state at the current moment:

[0075] (4)

[0076] (5)

[0077] In the formula, : The weight matrix of the input gate; : The bias vector of the input gate; : The weight matrix of the candidate memory cell state; : The bias vector of the candidate memory cell state.

[0078] (123) Update cell state c t : Determine the update of the cell state at the current moment jointly through the forget gate and the input gate:

[0079] (6)

[0080] (124) Output gate o t and hidden state h t are used to generate the output state at the current moment:

[0081] (7)

[0082] (8)

[0083] (125) Output layer: The final output of the LSTM is the preliminary predicted value of the fan power, which is achieved through the fully connected layer. , achieved through the fully connected layer

[0084] (9)

[0085] Where, W out is the weight of the output layer, and b out is the bias of the output layer.

[0086] (13) Training process: The data is divided into a training set and a validation set. The continuous historical data is input into the model using the sliding window method to predict the wind power output for the next one hour or several hours. The loss function can be selected as the mean squared error (MSE), and the model parameters are optimized through the backpropagation algorithm. To improve the prediction accuracy, the most commonly used loss function when training the LSTM model is the mean squared error (MSE), and the formula is as follows:

[0087] (10)

[0088] Where, is the real-time power of the fan, is the preliminary predicted value of the fan power, and N is the number of samples. The MSE is minimized through the backpropagation algorithm to update the model parameters W f 、W i 、W c 、W o 、W out to optimize the prediction results.

[0089] (14) Model evaluation: The model is evaluated using the validation set, and standard evaluation metrics such as the mean squared error (MSE) and the root mean squared error (RMSE) are used to measure the prediction accuracy. After training, the LSTM can be used for real-time prediction of wind power.

[0090] Through the training of the LSTM model, a preliminary predicted value of the fan power with high accuracy can be obtained. These prediction results can be used as the initial reference for subsequent intraday optimization.

[0091] 2. Intraday optimization: Based on the day-ahead prediction, the preliminary predicted value of the fan power is optimized intraday to obtain more accurate data, specifically using the rolling time window optimization strategy of machine learning.

[0092] In the intraday optimization of wind power, since the wind speed and weather conditions may change rapidly in a short period of time, relying solely on the day-ahead prediction cannot guarantee accuracy. The rolling time window optimization strategy improves the prediction accuracy by continuously updating the latest data. This method can be divided into the following steps:

[0093] (21) Real-time data input: Obtain the latest real-time wind speed and meteorological data at regular intervals (e.g., every 15 minutes), compare it with the results of the day-ahead prediction, and calculate the real-time error.

[0094] (22) Rolling window update: Combine the real-time data with historical data to form a rolling time window containing the latest information (e.g., data for the past 1 hour or 3 hours). Use this data for retraining or updating the prediction model. The definition of the rolling time window: Assume the current time is t, then the rolling time window W(t) contains the data for the most recent τ time instants: W(t) = {(P t-τ , X t-τ ), (P t-τ+1 , X t-τ+1 ),..., (P t-1 , X t-1 )}; where P t-i is the actual power value at the past time instant t - i, and X t-i is the corresponding meteorological data.

[0095] (23) Machine learning algorithm selection: Select the random forest algorithm (RF) to train or update on the data in the rolling time window for real-time data update optimization.

[0096] (24) Optimization process: In intraday optimization, based on the latest data in the rolling window, the wind power for the next few hours is corrected through a machine learning model. The optimization goal is to minimize the prediction error and ensure that the predicted value is closer to the actual value. Among them, the error correction for real-time update: Intraday optimization corrects the preliminary predicted value of the fan power through real-time data. Assume that a linear regression model is used for correction, then the corrected predicted value can be expressed as:

[0097] (11)

[0098] In the formula, is the preliminary predicted value of the fan power, and and are the correction coefficients obtained by fitting the data within the rolling time window.

[0099] Error minimization: Within the rolling window, the error of the corrected predicted value is minimized through the random forest algorithm. Assume the error is defined as , represents the real-time power of the fan, and the optimization goal is to minimize the following loss function:

[0100] (12)

[0101] In the rolling time window, the addition of real-time data enables the system to dynamically adjust the prediction results. Set a correction step size , and update the model every . After the update, the model makes predictions again based on the latest error correction data, and finally obtains the optimized predicted value within the day . This can significantly reduce the prediction error caused by drastic changes in wind speed or other meteorological conditions within a short period of time

[0102] During the intra-day optimization process, not only is the linear regression model used to correct the preliminary predicted value of the wind turbine power, but also the random forest algorithm is used to minimize the prediction error. By setting a reasonable loss function and continuously updating the prediction model in combination with the data within the rolling time window, the prediction error can be reduced more effectively. Compared with the traditional technology that only relies on a single correction method, the present invention can more comprehensively consider the influence of various factors on the prediction error, thereby improving the accuracy and reliability of the prediction

[0103] LSTM is good at processing sequence data and can capture long-term dependence relationships, while the rolling time window optimization strategy can continuously update the prediction model according to real-time data to adapt to the rapid changes in wind speed and weather conditions within a short period of time. The combination of LSTM and the rolling time window optimization can significantly improve the prediction accuracy compared with a single prediction method. For example, in some existing technologies, simple statistical methods or a single neural network structure may be used for prediction, which cannot take into account both long-term trends and short-term fluctuations at the same time, while the method of the present invention can better handle the complex change characteristics of wind turbine power

[0104] 3. Real-time comparison and dynamic compensation: Compare the predicted data optimized within the day with the real-time wind turbine power. Then, based on the prediction error, perform dynamic compensation configuration for the hydrogen energy system. When the real-time power of the wind turbine is greater than the predicted data, the hydrogen energy system starts the electrolytic hydrogen production process to convert the excess electric energy into hydrogen energy for storage; when the real-time power of the wind turbine is less than the predicted data, the hydrogen energy system discharges through the fuel cell to perform power compensation for the wind turbine unit, thereby reducing the wind abandonment and load shedding rates and lowering the operation risk cost

[0105] Among them, the error compensation coefficient k comp is the key parameter that determines the compensation strength of the hydrogen energy system. The following are the detailed steps for determining the error compensation coefficient

[0106] Real-time error calculation: According to the difference between the real-time power of the wind turbine and the predicted value optimized within the day , calculate the real-time error , that is:

[0107] (13)

[0108] This error value directly affects the compensation amount of the hydrogen energy system.

[0109] Initial value setting of the error compensation coefficient: According to the actual operation of the wind farm, the initial value can be set for the error compensation coefficient k comp Usually, the initial error compensation coefficient is set to 1, which means that all real-time errors are compensated by the hydrogen energy system.

[0110] Dynamic adjustment: In practical applications, the compensation coefficient can be dynamically adjusted according to the current working state of the hydrogen energy system. The specific adjustment method is as follows:

[0111] (31) If the actual error is small, it indicates high prediction accuracy. At this time, the compensation coefficient can be appropriately reduced to prevent over-compensation of the hydrogen energy system. For example, k comp can be set to 0.8.

[0112] (32) If the error is large and the capacity and SOC state of the hydrogen energy system permit, the compensation coefficient can be increased to 1 or a higher value to ensure the timeliness of power compensation.

[0113] Error compensation formula: The power P that the hydrogen energy system needs to compensate comp is determined by the following formula:

[0114] (14)

[0115] In the formula, P comp represents the power that the hydrogen energy system needs to compensate, and k comp is the dynamic compensation coefficient.

[0116] Through the analysis of historical data, the optimal value of the compensation coefficient can be found to ensure that the hydrogen energy system can effectively reduce the wind power prediction error during actual operation without causing energy waste due to over-compensation.

[0117] 4. Capacity configuration of the hydrogen energy system: According to the above steps and data processing results, complete the capacity configuration of the hydrogen energy system to ensure its effective operation and the realization of the power compensation function. The capacity configuration of the hydrogen energy system involves finding a balance between the wind power prediction error and the cost of the hydrogen energy system. The following are the specific configuration methods:

[0118] (41) Power error analysis: First, analyze the maximum power error in the historical data . This is the basis for configuring the hydrogen energy system to ensure that the hydrogen energy system has sufficient capacity to handle the most extreme power fluctuations.

[0119] (42) Capacity configuration formula: The capacity C of the hydrogen energy system H2 is calculated as follows:

[0120] (15)

[0121] Wherein: is the maximum power error.

[0122] T is the duration that needs to be compensated, usually 1 hour or longer.

[0123] is the efficiency of the electrolyzer, usually 70% - 80%.

[0124] is the efficiency of the fuel cell, usually 50% - 60%.

[0125] (43) Optimization configuration: Generally, the capacity of the hydrogen energy system should be configured as 30% - 50% of the installed capacity of the wind farm. This configuration can ensure that in most cases, the hydrogen energy system has sufficient capacity to absorb excess power or release power when needed.

[0126] (44) SOC control: To avoid overcharging and over-discharging of the hydrogen energy system, the state (SOC, State of Charge) of the hydrogen energy system should be controlled within a reasonable range, usually set to 20% - 80%. When the SOC is too high or too low, the hydrogen energy system should adjust its working state in a timely manner to ensure the effective operation of the electrolyzer and the fuel cell.

[0127] In the capacity configuration of the hydrogen energy system, not only the size of the capacity is considered, but also the state of charge (SOC) is reasonably controlled. Setting the SOC to 20% - 80% ensures the effective operation of the electrolyzer and the fuel cell. This coordination of SOC control and capacity configuration can improve the service life and operation efficiency of the hydrogen energy system.

[0128] During the capacity configuration process of the hydrogen energy system, first analyze the maximum power error in the historical data , and then perform capacity calculation by combining multiple factors such as the duration to be compensated, the efficiency of the electrolyzer, and the efficiency of the fuel cell. This capacity configuration method based on historical data and actual system parameters can more accurately determine the capacity required for the hydrogen energy system.

[0129] Compared with some methods that determine the capacity only based on experience or simple proportional relationships, the capacity configuration of the present invention is more scientific and reasonable, which can ensure that the hydrogen energy system has sufficient capacity to handle power fluctuations under different working conditions, and at the same time avoid cost waste caused by excessive capacity.

[0130] The present invention adopts a multi-level prediction and feedback mechanism, which can more accurately estimate the change of wind turbine power, adjust the working state of the hydrogen energy system in advance, thereby improving the energy management efficiency of the entire system and reducing energy waste and system operation instability caused by inaccurate prediction.

[0131] Application Example: The wind power data used in the present invention are the operation data of a wind farm from September 18, 2023 to September 20, 2023. The time interval is 15 minutes. The day-ahead prediction is updated at 5:40 pm, and the latest prediction is updated hourly. The capacity of the wind farm is 5315.3 MW. As shown in the appendix Figure 3 As shown, it shows the corresponding waveform diagram. After normalizing the wind power data, it can be seen that the actual wind power fluctuations are relatively large. The day-ahead prediction value can follow the actual value changes relatively accurately, and its prediction accuracy reaches 75.63%. The real-time prediction is relatively better than the day-ahead prediction, and its accuracy reaches 79.41%, but neither meets the criterion requirements. Therefore, only date prediction and intraday optimization cannot meet the national standards.

[0132] By taking the difference between the actual wind power data and the day-ahead prediction value, the curtailment / load-shedding power diagram is obtained, as shown in the appendix Figure 4 As shown. At this time, when time = 85, the actual prediction error value of the wind power is 0.085 pu, and there is a situation of wind power underestimation. The curtailment power reaches 0.085 * 5315.3 = 451 MW, and the curtailment rate reaches 0.085 / 0.6638 = 12.8%. Another example is when t = 147, the actual prediction error value of the wind power is -0.227 pu, and there is a situation of wind power overestimation. At this time, it is necessary to shed a load of 0.227 * 5315.3 = 1205 MW, and the load-shedding rate is as high as 0.227 / 0.493 = 46%.

[0133] Simulation Example 1: In this example, the maximum power control strategy is adopted for the machine-side control of the wind turbine, and the grid-side adopts the grid-voltage-oriented vector control strategy. The active power reference is the wind power prediction value. The hydrogen energy system adopts power feedforward control considering SOC, and the power reference value of the hydrogen energy system is given according to the DC-side voltage deviation.

[0134] It can be seen from the appendix Figure 5 that configuring hydrogen energy on the DC side can eliminate part of the power prediction error, but does not reduce the overall power prediction error. Since the control purpose of configuring the hydrogen energy system on the DC side is to maintain the DC-side voltage constant, when a prediction error occurs, unbalanced power will be generated on the machine side and the grid side, which will cause the hydrogen energy system to work, either for electrolytic hydrogen production or fuel cell discharge, until the SOC threshold is reached and it stops working. The initial SOC is set to 50%, the rated value of the hydrogen energy capacity is set to 0.25, the available capacity range is 0.04 - 0.2 in per-unit value, and the available capacity is 0.15 pu. By the hydrogen energy compensation prediction error waveform, it is calculated that after adding the hydrogen energy system on the DC side, the accuracy of the day-ahead power prediction value is 77.13%, which is 1.5 percentage points higher than before adding. When the available capacity is increased to 0.38 pu, the power prediction error diagram of configuring hydrogen energy on the DC side is as shown in the appendix Figure 6As shown, the error curve after compensation is significantly more eliminated than that of the previously configured 0.15 pu system, and the prediction accuracy has also increased to 80.04%, which is 2.91 percentage points higher than the former.

[0135] If hydrogen energy is configured on the DC side and the control objective is to maintain the stability of the DC side voltage, then in the case of power imbalance, the hydrogen energy system will store all the unbalanced power in the system or release it to the power grid. In this case, due to this, the effect that the hydrogen energy system cannot compensate the power error as a whole will occur. Although it can completely compensate for some prediction errors, it will not reduce the error peak. Attached Figure 5 Among them, some prediction errors are completely compensated, but the maximum and minimum values of the prediction errors are not compensated. The wind curtailment rate at t = 85 reaches 12.8%; for another example, at t = 147, the load shedding rate is as high as 46%. Figure 6 Among them, when the capacity of the hydrogen energy system increases from 0.15 pu to 0.38 pu, the wind curtailment rate decreases to 9.76% at 23.28 h, a decrease of 3.2 percentage points. The load shedding rate does not change at 36.5 h, showing a certain phenomenon of reducing wind curtailment and load shedding. It can be seen that doubling the capacity of the hydrogen energy system does not significantly change the wind curtailment volume and load shedding volume.

[0136] By setting the capacity of the hydrogen energy system from 0.1 pu to 0.5 pu, sampling the capacity and accuracy at intervals of 0.05 pu, and obtaining the attached Figure 7 Among them, the curve of the capacity and accuracy of the hydrogen energy system under the DC side voltage stabilization control strategy. Through this curve, the capacity configuration when the accuracy is greater than 85% is 0.465 pu.

[0137] It can be seen that at low capacity, the operating risk cost of the system is not improved much. The reason is not that the capacity configuration is insufficient, but that the amount of electricity that the hydrogen energy system can absorb or emit in a short time is limited, that is, it is restricted by the SOC threshold of the hydrogen energy battery. Therefore, when the hydrogen energy capacity configuration is too high, the improvement of the system is not obvious, but instead increases the cost of hydrogen energy configuration. Therefore, not only the capacity of hydrogen energy needs to be reasonably configured, but also the control strategy needs to be changed.

[0138] Application Example 2: In this example, the hydrogen energy system is used for charge and discharge to compensate for the prediction error. The prediction errors before and after hydrogen energy compensation are as attached Figure 8 As shown, in this example, the maximum power control strategy is adopted for the machine side control of the wind turbine, and the grid side adopts the vector control strategy based on the grid voltage orientation. The power reference is the product of the predicted error value of the day-ahead and the compensation coefficient. Its compensation effect will change with the change of the compensation coefficient. The hydrogen energy system is configured on the DC side, and the control strategy is the power feedforward control considering the SOC. The control objective is to stabilize the DC side voltage. The set initial SOC is 50%, and the available capacity of the hydrogen energy system is 0.15 pu.

[0139] Appendix Figure 8 As can be seen, after modifying the control strategy, the compensation effect of the hydrogen energy system can act on the overall prediction error, but it is still limited by the hydrogen energy configuration capacity. At 21.28 h, the day-ahead prediction error value decreased by 29.9%, and the wind curtailment rate decreased from 12.8% to 8.97% at this time; at 36.5 h, the prediction error value decreased by 29.89%, and the load shedding rate decreased from 46% to 24.03% at this time. It can be seen that the prediction error value uses the error coefficient as the control strategy for the active power setting of the hydrogen energy system, resulting in a long compensation response time of the hydrogen energy system, reducing wind curtailment and load shedding phenomena, and the prediction accuracy reaching 80.87%.

[0140] When the hydrogen energy capacity is increased to 0.38 pu and the error coefficient is set to 0.6, the prediction accuracy value is 88.69%. Appendix Figure 9 As can be seen, at 21.28 h, the day-ahead prediction error value decreased by 59.98%, and the wind curtailment rate decreased from 12.8% to 5.13% at this time; at 36.5 h, the prediction error value decreased by 59.78%, and the load shedding rate decreased from 46% to 13.78% at this time, meeting the criterion index.

[0141] Appendix Figure 10 The figure shows a comparison diagram of the energy storage configuration capacity and prediction accuracy under two control strategies. It can be seen that using the prediction error as the active power setting of the DC-side hydrogen energy system, the accuracy of this strategy is generally greater than that of the strategy using the DC voltage deviation as the active power setting of the DC-side hydrogen energy system. According to the requirement that the power prediction accuracy of the wind farm should reach more than 85%, the available energy storage capacity is given as 0.265 pu.

[0142] Since the compensation effect of hydrogen energy under this strategy is related not only to the hydrogen energy configuration capacity but also to the error compensation coefficient, a three-dimensional diagram of energy storage capacity - prediction error coefficient - prediction accuracy is established, as shown in Appendix Figure 11 As shown, at the same hydrogen energy capacity, the prediction error accuracies generated by different error coefficients are different, and there will be a maximum value. It can be known that the maximum accuracy value increases with the increase of the energy storage capacity and decreases with the increase of the error coefficient. The error compensation coefficient is given as 0.46, and the capacity configuration is 0.265 pu.

[0143] Embodiment 2

[0144] This embodiment provides a hydrogen energy system configuration device for reducing the power deviation of a wind turbine, which includes:

[0145] 1. Day-ahead prediction unit: Using a long short-term memory network to perform day-ahead prediction on the wind turbine power to obtain the LSTM prediction value of the wind turbine power.

[0146] 2. Intra - day Optimization Unit: Based on the day - ahead prediction, the LSTM predicted value of the wind turbine power is optimized intra - day to obtain the intra - day optimized predicted value, and then the intra - day optimized predicted value is used as the predicted data of the wind turbine power; the intra - day optimization adopts the rolling time - window optimization strategy of machine learning.

[0147] The goal of intra - day optimization is to minimize the prediction error. The intra - day optimization corrects the LSTM predicted value of the wind turbine power using a linear regression model with real - time data, and the corrected predicted value is expressed as:

[0148] ,

[0149] where, represents the preliminary predicted value of the wind turbine power (i.e., the LSTM predicted value of the wind turbine power), and are the correction coefficients obtained by fitting the data within the rolling time - window;

[0150] Prediction error minimization: Within the rolling window, the prediction error of the current wind turbine power is minimized through the random forest algorithm; assuming the prediction error is defined as , represents the real - time power of the wind turbine, then the optimization goal is to minimize the following loss function:

[0151] ,

[0152] Set a correction step size △t, update the model every △t, and the updated model makes predictions again based on the latest corrected predicted value, and finally obtains the intra - day optimized predicted value .

[0153] 3. Real - time Comparison and Dynamic Compensation Unit: According to the predicted data of the wind turbine power, compare it with the real - time power of the wind turbine to obtain the real - time error, and then perform dynamic compensation configuration of the hydrogen energy system based on the real - time error; when the real - time power of the wind turbine is greater than the predicted data, the hydrogen energy system starts the electrolytic hydrogen production process to convert the excess electric energy into hydrogen energy; when the real - time power of the wind turbine is less than the predicted data, the hydrogen energy system discharges through the fuel cell to compensate the power of the wind turbine unit.

[0154] When the hydrogen energy system compensates the power of the wind turbine unit, the compensated output power P comp is determined by the following formula:

[0155] ,

[0156] where, P comp represents the compensated output power, k comp represents the error compensation coefficient, Indicates the real-time error;

[0157] The error compensation coefficient is dynamically adjusted according to the current working state of the hydrogen energy system. When the error compensation coefficient is 1, full-power compensation is performed; when the error compensation coefficient is 0, no compensation is performed.

[0158] According to the real-time power of the fan And the predicted value optimized within the day Calculate the real-time error based on the difference between them , that is:

[0159] .

[0160] 4. Hydrogen energy system capacity configuration unit: Complete the configuration of the hydrogen energy system capacity according to the aforementioned data processing results.

[0161] The process of configuring the hydrogen energy system capacity is as follows:

[0162] 1) Analyze the maximum power error in historical data ;

[0163] 2) Capacity configuration formula: The calculation formula for the capacity C of the hydrogen energy system H2 is as follows:

[0164]

[0165] Among them, T is the duration that needs to be compensated; is the efficiency of the electrolyzer; η fc is the efficiency of the fuel cell;

[0166] 3) Optimal configuration: The capacity configuration of the hydrogen energy system is 30%-50% of the installed capacity of the wind farm;

[0167] 4) SOC control: The state of charge SOC of the hydrogen energy system is set to 20%-80% to ensure the effective operation of the electrolyzer and the fuel cell.

[0168] It should be noted that each unit in the above hydrogen energy system configuration device for reducing the power deviation of the fan can be implemented in whole or in part by software, hardware, and their combination. The above units can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each of the above units. For the specific limitations of a hydrogen energy system configuration device for reducing the power deviation of the fan, refer to the limitations on a hydrogen energy system configuration method for reducing the power deviation of the fan in the above text. The two have the same functions and effects and will not be elaborated here.

[0169] Example 3

[0170] This embodiment provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, and when the computer program is executed by the at least one processor, it is used to cause the electronic device to execute the method according to Embodiment 1 of the present invention.

[0171] Example 4

[0172] This embodiment provides a non-transitory computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor of a computer, it is used to cause the computer to execute the method according to Embodiment 1 of the present invention.

[0173] Reference Figure 12 , the structural block diagram of an electronic device 400 that can be used as a server or a client of the present invention will now be described. It is an example of a hardware device that can be applied to various aspects of the present invention. The electronic device is intended to represent various forms of digital electronic computer devices, such as, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, a personal digital processor, a cellular phone, a smart phone, a wearable device, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are only examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0174] As Figure 12 shown, the electronic device 400 includes a computing unit 401, which can execute various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 402 or a computer program loaded from a storage unit 408 into a random access memory (RAM) 403. In the RAM 403, various programs and data required for the operation of the electronic device 400 can also be stored. The computing unit 401, the ROM 402, and the RAM 403 are connected to each other through a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0175] Multiple components in the electronic device 400 are connected to the I / O interface 405, including: an input unit 406, an output unit 407, a storage unit 408, and a communication unit 409. The input unit 406 can be any type of device capable of inputting information into the electronic device 400. The input unit 406 can receive input digital or character information, and generate key signal inputs related to the user settings and / or function controls of the electronic device. The output unit 407 can be any type of device capable of presenting information, and can include, but is not limited to, a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 408 can include, but is not limited to, magnetic disks and optical discs. The communication unit 409 allows the electronic device 400 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks, and can include, but is not limited to, a modem, a network card, an infrared communication device, a wireless communication transceiver, and / or a chipset, such as a Bluetooth™ device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0176] The computing unit 401 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 401 executes the various methods and processes described above. For example, in some embodiments, the aforementioned hydrogen energy system configuration method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 400 via the ROM 402 and / or the communication unit 409. In some embodiments, the computing unit 401 can be configured to execute the aforementioned hydrogen energy system configuration method by any other suitable means (e.g., by means of firmware).

[0177] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the program codes are executed by the processor or controller, the functions / operations specified in the flowchart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, executed partially on the machine as an independent software package and partially on a remote machine, or executed entirely on a remote machine or server.

[0178] In the context of the present invention, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0179] As used in the present invention, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus, and / or device (e.g., a disk, optical disk, memory, programmable logic device (PLD)) used to provide machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0180] In order to provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, speech input, or tactile input).

[0181] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), and the Internet.

[0182] A computer system can include clients and servers. The clients and servers are generally remote from each other and typically interact through a communication network. The client - server relationship is created by computer programs that run on the respective computers and have a client - server relationship with each other.

[0183] Persons skilled in the art can obviously make various modifications to the above - mentioned embodiments easily and apply the general principles described herein to other embodiments without creative efforts. Therefore, the present invention is not limited to the above - mentioned embodiments, and all improvements and modifications made by persons skilled in the art to the present invention based on the disclosure of the present invention should be within the protection scope of the present invention.

Claims

1. A method for configuring a hydrogen energy system to reduce wind turbine power deviation, characterized in that: include: Real-time comparison and dynamic compensation: The predicted data of wind turbine power is compared with the real-time power of the wind turbine to obtain the real-time error, and then the dynamic compensation configuration of the hydrogen energy system is performed based on the real-time error; when the real-time power of the wind turbine is greater than the predicted data, the hydrogen energy system starts the electrolysis hydrogen production process to convert the excess electrical energy into hydrogen energy; when the real-time power of the wind turbine is less than the predicted data, the hydrogen energy system discharges the fuel cell to compensate the wind turbine for power; Hydrogen energy system capacity configuration: Based on the above data processing results, complete the configuration of hydrogen energy system capacity; The predicted data of the wind turbine power is also predicted before being compared with the real-time power of the wind turbine. The prediction of the day-ahead adopts a long short-term memory network to predict the wind turbine power and obtain a preliminary prediction value of the wind turbine power. On the basis of the day-ahead forecast, the preliminary forecast value of wind turbine power is also optimized intraday to obtain the forecast value after intraday optimization, and then the forecast value after intraday optimization is used as the forecast data of wind turbine power; The intraday optimization described adopts a rolling time window optimization strategy based on machine learning; The goal of intraday optimization is to minimize the prediction error. Intraday optimization uses a linear regression model to correct the initial prediction value of wind turbine power through real-time data. The corrected prediction value It is expressed as: ， In the formula, represents the preliminary predicted value of wind turbine power, and is the correction coefficient obtained by fitting the data in the rolling time window; Minimize the prediction error: In the rolling window, the corrected prediction value is minimized by the random forest algorithm. Minimize the error; assume that the prediction error is defined as , represents the real-time power of the wind turbine, then the optimization objective is to minimize the following loss function: ， Set a correction step size , every Update the forecast model once, and then make another forecast based on the latest revised forecast value, and finally get the forecast value after intraday optimization. .

2. A hydrogen energy system configuration method for reducing wind turbine power deviation according to claim 1, characterized in that: When the hydrogen energy system performs power compensation on the wind turbine, the output power P after compensation is comp Determined by the following formula: ， Where P comp represents the output power after compensation, k comp represents the error compensation coefficient, Indicates real-time error; The error compensation coefficient is dynamically adjusted according to the current working state of the hydrogen energy system. When the error compensation coefficient is 1, full power compensation is performed; when the error compensation coefficient is 0, no compensation is performed.

3. A hydrogen energy system configuration method for reducing wind turbine power deviation according to claim 2, characterized in that: According to the real-time power of the fan Compared with the predicted value after intraday optimization The difference between the two is used to calculate the real-time error ,Right now: 。 4. A hydrogen energy system configuration method for reducing wind turbine power deviation according to claim 3, characterized in that: The process of hydrogen energy system capacity configuration is as follows: 1) Analyze the maximum power error in historical data ; 2) Capacity configuration formula: Hydrogen energy system capacity C H2 The calculation formula is as follows: ， in, T is the length of time that needs to be compensated; is the efficiency of the electrolyzer; is the efficiency of the fuel cell; 3) Optimized configuration: The capacity of the hydrogen energy system is configured to be 30%-50% of the installed capacity of the wind farm; 4) SOC control: The state of charge (SOC) of the hydrogen energy system is set to 20%-80% to ensure the effective operation of the electrolyzer and fuel cell.

5. A hydrogen energy system configuration device for reducing wind turbine power deviation, characterized in that: include: Day-ahead prediction unit: uses long short-term memory network to predict wind turbine power and obtains preliminary prediction value of wind turbine power; Intraday optimization unit: Based on the day-ahead prediction, the LSTM prediction value of the wind turbine power is optimized intraday to obtain the intraday optimized prediction value, and then the intraday optimized prediction value is used as the prediction data of the wind turbine power; the intraday optimization adopts the rolling time window optimization strategy of machine learning; The goal of intraday optimization is to minimize the prediction error. Intraday optimization uses a linear regression model to correct the initial prediction value of wind turbine power through real-time data. The corrected prediction value It is expressed as: ， In the formula, represents the preliminary predicted value of wind turbine power, and is the correction coefficient obtained by fitting the data in the rolling time window; Minimize the prediction error: In the rolling window, the corrected prediction value is minimized by the random forest algorithm. Minimize the error; assume that the prediction error is defined as , represents the real-time power of the wind turbine, then the optimization objective is to minimize the following loss function: ， Set a correction step size , every Update the forecast model once, and then make another forecast based on the latest revised forecast value, and finally get the forecast value after intraday optimization. ; Real-time comparison and dynamic compensation unit: According to the predicted data of wind turbine power, it is compared with the real-time power of the wind turbine to obtain the real-time error, and then the dynamic compensation configuration of the hydrogen energy system is performed based on the real-time error; when the real-time power of the wind turbine is greater than the predicted data, the hydrogen energy system starts the electrolysis hydrogen production process to convert the excess electrical energy into hydrogen energy; when the real-time power of the wind turbine is less than the predicted data, the hydrogen energy system discharges the fuel cell to compensate the wind turbine for power; Hydrogen energy system capacity configuration unit: completes the configuration of the hydrogen energy system capacity based on the above data processing results.

6. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

7. 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 4 are implemented.

Images