Hydrogen production control method, system, device and storage medium based on predicted power

Abstract

The present application relates to a hydrogen production control method, system, device and storage medium based on predicted power, which simulates the environment around the offshore wind turbine using the obtained surrounding weather data to obtain predicted surrounding weather data; uses the known wind turbine performance curve to determine the predicted power of the offshore wind turbine wind power output; extracts the feature vector of the wind power output power value; obtains the hydrogen production power of the hydrogen production device within a predetermined time period and arranges it into a hydrogen production power input vector according to the time dimension, and extracts the feature vector of the hydrogen production power; calculates the difference feature vector between the feature vector of the wind power output power value and the feature vector of the hydrogen production power; inputs the difference feature vector into the classifier, and controls the hydrogen production device according to the classification result; the present application can realize accurate prediction and efficient control of the offshore wind turbine hydrogen production device, improve the response speed and flexibility of the system, and at the same time ensure the stability and efficiency of the hydrogen production process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wind power hydrogen production, in particular to a hydrogen production control method, system, device and storage medium based on predicted power. BACKGROUND

[0002] Offshore wind power, as a clean and sustainable energy form, has received extensive attention and application in recent years. Offshore wind power hydrogen production system is a new energy system that integrates wind energy conversion, power storage and hydrogen energy production. It uses the electricity generated by offshore wind power to produce hydrogen through water electrolysis, thereby realizing energy storage and conversion and providing a sustainable solution for future energy demand. However, the existing operation scheme for offshore wind power hydrogen production system lacks effective measures to deal with complex situations, thereby reducing the stability of hydrogen production.

[0003] For example, patent publication No. CN114204602B discloses a "wind power hydrogen production control method and system based on real-time meteorological data". According to the real-time meteorological data, the instantaneous power prediction value of the wind field is determined. The instantaneous power prediction value of the wind field is subtracted from the instantaneous demand power prediction value of the power grid to form a difference prediction curve. According to the difference prediction curve within the set time range, the average power difference within the set time range is obtained. According to the average power difference, the operation of the energy storage device and the hydrogen production device under different conditions is adjusted.

[0004] For example, patent publication No. CN113516274A discloses a "hydrogen energy system and its balance control method". The supply and demand balance strategy that can maximize the use of new energy, minimize the use of electricity in the power grid, and meet the hydrogen demand is determined according to the new energy power supply prediction and the hydrogen demand prediction. The corresponding subsystems in the hydrogen energy system are controlled to operate according to the determined strategy to realize hydrogen production.

[0005] Although the existing method can effectively evaluate the hydrogen production feasibility of offshore wind farms, the deviation of the operation scheme of the offshore wind power hydrogen production system is large, which causes the hydrogen production device to not be able to maintain in the best working state. SUMMARY

[0006] In view of the deficiencies of the prior art, the present application provides a hydrogen production control method, system, device and storage medium based on predicted power to solve the above problems.

[0007] The present application provides the following technical solutions:

[0008] The hydrogen production control method based on predicted power comprises the following steps:

[0009] S1, obtaining the surrounding meteorological data D(T) of the offshore wind turbine within a predetermined time period;

[0010] S2, using the obtained peripheral weather data D(T), an atmosphere-ocean coupled model is constructed to simulate the environment around the offshore wind turbine, and the predicted peripheral weather data D(T future ) is obtained;

[0011] S3, according to the measured peripheral weather data D(T future ), the known wind turbine performance curve is used to determine the offshore wind turbine wind power output prediction power P=f[D(T future )]=[x1, x2,..., x t ], wherein x t represents the wind power output power value at time point t;

[0012] S4, the feature vector A=[a1, a2,...a t ] of the wind power output power value is extracted;

[0013] S5, the hydrogen production power of the hydrogen production device in a predetermined time period is obtained, and is arranged into a hydrogen production power input vector according to time dimension, and the feature vector B=[b1, b2,...b t ] of the hydrogen production power is extracted;

[0014] S6, the difference feature vector C=[c1, c2,..., c t ] between the feature vector A of the wind power output power value and the feature vector B of the hydrogen production power is calculated;

[0015] S7, the difference feature vector C is input into a classifier, the classifier contains a matrix M=[m1, m2,..., m j ], and each column vector m j represents the feature vector of a control method, the similarity is calculated by using the difference feature vector C and each column vector in the matrix M, and the highest similarity obtained is the classification result, and the hydrogen production device is controlled according to the classification result;

[0016] S8, when the input power is large, the hydrogen production device reaches the rated hydrogen production power, and the excess wind power is input into the energy storage device; if the input power is small, the energy storage device is started, the hydrogen production device is controlled to reach the rated hydrogen production power, and the influence of the fluctuation of the offshore wind turbine wind power input power on the hydrogen production device is reduced.

[0017] Preferably, the peripheral weather data D(T) includes sea waves, temperature, humidity and wind speed;

[0018] In S2, the historical peripheral weather data of the sensor and the observation station is obtained, and the data is preprocessed to ensure the quality and consistency of the data;

[0019] The historical surrounding meteorological data set is divided into a training set, a validation set and a test set, input into a machine learning model for training, and the model parameters are repeatedly iterated and optimized to minimize the prediction error.

[0020] The trained model is applied to the prediction of the surrounding meteorological data of the offshore wind turbine to generate predicted surrounding meteorological data D(T future ), and the model is updated and optimized based on these data and future surrounding meteorological data.

[0021] Preferably, in S3, the performance parameters are dynamically adjusted by introducing a machine learning algorithm to periodically update the performance curve of the offshore wind turbine to reflect changes in the actual operating state.

[0022] Preferably, in S4, the wind power output prediction power P is input into a one-dimensional convolution layer time encoder using the formula to obtain the feature vector A of the wind power output value;

[0023] Determine the convolution kernel size, select a suitable convolution kernel size K, initialize a set of weights W = [W0, W1,..., W k-1 ] and a bias term b for each convolution kernel, wherein the weights W and the bias term b are automatically adjusted through the backpropagation algorithm in the training process;

[0024] Apply the convolution operation to the offshore wind turbine wind power output prediction power P = [x1, x2,..., x t ], for each position i, calculate the convolution result at this position by the formula ;

[0025] where K is the convolution kernel length, b is a bias term, and a i is the output feature value at position i;

[0026] Repeat the above process until all possible positions i are covered, and the final [a1, a2,..., a t ] constitutes the feature vector A of the wind power output value.

[0027] Preferably, in S5, the hydrogen production power input vector is first input into a convolutional neural network based on an attention mechanism, and then input into a convolutional neural network based on a time mechanism, to obtain the feature vector B = [b1, b2,..., b t ] of the hydrogen production feature power.

[0028] Preferably, the difference feature vector C is calculated using the formula ;

[0029] where μ A , μ Brespectively, are the mean values of the feature vector A of the wind power output value and the feature vector B of the hydrogen production characteristic power, respectively, are the variances of the feature vector A of the wind power output value and the feature vector B of the hydrogen production characteristic power, and α and β are weighting factors.

[0030] Preferably, in S8, assuming that the offshore wind turbine wind power output prediction power P is greater than the rated power of the hydrogen production device, the classification result of the classifier is enabled to control the hydrogen production device to reach the rated hydrogen production power and start the charging module of the energy storage device to store the excess power in the energy storage device.

[0031] Assuming that the offshore wind turbine wind power output prediction power P is less than the required minimum power of the hydrogen production device, the classification result of the classifier is enabled to start the discharging module of the energy storage device to control the power of the hydrogen production device within the range of the minimum hydrogen production power and the rated power in combination with the wind power input.

[0032] The hydrogen production control system based on prediction power is used to realize the hydrogen production control method based on prediction power, comprising:

[0033] A data acquisition module is used to obtain the surrounding meteorological data D(T) of the offshore wind turbine within a predetermined time period;

[0034] An environmental simulation module is used to construct an atmosphere-ocean coupled model using the obtained surrounding meteorological data D(T) and simulate the surrounding environment, thereby obtaining the future predicted surrounding meteorological data D(T future );

[0035] A power prediction module is used to predict the offshore wind turbine wind power output prediction power P using the wind turbine performance curve according to the simulated predicted surrounding meteorological data D(T future );

[0036] A feature extraction module is used to extract the feature vector A from the offshore wind turbine wind power output prediction power P using a one-dimensional convolution layer time series encoder;

[0037] A hydrogen production data processing module is used to collect and pre-process the hydrogen production power data of the hydrogen production device within a predetermined time period, convert it into a hydrogen production power input vector, and then extract the feature vector B of the hydrogen production characteristic power from the hydrogen production power input vector using a multi-scale convolutional encoder;

[0038] A feature comparison module is used to perform multi-feature value weight difference on the feature vector A extracted from the offshore wind turbine wind power output prediction power P and the feature vector B of the hydrogen production characteristic power, thereby obtaining the difference feature vector C;

[0039] An intelligent control module is used to input the difference feature vector C into a classifier to obtain a classification result, which is used to guide the hydrogen production efficiency control strategy of the hydrogen production device.

[0040] An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the method for hydrogen production control based on predicted power when executing the program.

[0041] A computer-readable storage medium having stored thereon a computer program, the program being executable by a processor to implement the method for hydrogen production control based on predicted power.

[0042] The present application has the following beneficial technical effects:

[0043] The present application can more accurately predict the performance and output power of offshore wind turbines by simulating the environment around offshore wind turbines using an atmosphere-ocean coupled model.

[0044] The known wind turbine performance curve is used to determine the offshore wind turbine wind power output prediction power, which can reduce uncertainty and error and improve prediction accuracy.

[0045] The time series data is converted into a feature vector, which can effectively capture and analyze patterns and trends in time series data, thereby improving the prediction accuracy of future hydrogen production power.

[0046] By calculating the difference feature vector between the feature vector of the output prediction power and the hydrogen production power feature vector, the deviation between the input power demand and the actual power output of the hydrogen production device can be evaluated and identified. At the same time, the similarity calculation between the difference feature vector and the feature vector matrix in the classifier can quickly determine the optimal control strategy. This method not only improves the efficiency of control, but also dynamically adjusts the operating state of the hydrogen production device according to different meteorological conditions and wind power output volatility.

[0047] The present application can realize accurate prediction and efficient control of offshore wind turbine hydrogen production devices, improve the response speed and flexibility of the system, and ensure the stability and efficiency of the hydrogen production process. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present application will be described below in conjunction with the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0049] Embodiments:

[0050] Hydrogen production control method and system based on predicted power:

[0051] S1, obtain the surrounding meteorological data D(T) of the offshore wind turbine within a predetermined time period, wherein the surrounding meteorological data D(T) mainly includes all factors that may affect the performance of the wind turbine, such as sea waves, temperature, humidity and wind speed, etc.;

[0052] S2, using the obtained surrounding meteorological data D(T), build an atmosphere-ocean coupled model to simulate the surrounding environment of the offshore wind turbine, and obtain the predicted surrounding meteorological data D(T future );

[0053] Obtain the historical surrounding meteorological data of the sensor and the observation station, and preprocess the data to ensure the quality and consistency of the data;

[0054] Divide the historical surrounding meteorological data set into training set, validation set and test set, input to suitable physical or machine learning model for training, and optimize the model parameters through repeated iteration to minimize the prediction error;

[0055] Apply the trained model to the prediction of the surrounding meteorological data of the offshore wind turbine, generate the predicted surrounding meteorological data, and update and optimize the model based on these data and future surrounding meteorological data.

[0056] S3, according to the predicted surrounding meteorological data D(T future ), use the known wind turbine performance curve to determine the offshore wind turbine wind power output prediction power P=f[D(T future )]=[x1, x2,..., x t ], wherein x t represents the wind power output power value at time point t;

[0057] Regularly update the performance curve by introducing machine learning algorithm to dynamically adjust the performance parameters to reflect the changes of the actual running state.

[0058] S4, the wind power output power value is encoded by a one-dimensional convolution layer time encoder using the formula To obtain the feature vector A=[a1, a2,...a t ] of the wind power output power value; Specifically as follows:

[0059] Determine the size of the convolution kernel, select a suitable convolution kernel size K, initialize a set of weights W=[W0, W1,..., W k-1 ] and a bias term b for each convolution kernel, wherein the weights W and the bias term b are automatically adjusted by the back propagation algorithm in the training process;

[0060] Apply the convolution operation to the input prediction power sequence P=[x1, x2,..., x t ], for each position i, through the formula calculating the convolution result at the position;

[0061] where K is the length of the convolution kernel, b is a bias term, a i is the output feature value at position i;

[0062] The above process is repeated until all possible positions i are covered, and the final [a1, a2,... a t ] constitutes the feature vector A of the output predicted power.

[0063] In addition to the above process, in some cases, in order to reduce the dimension of the feature vector or capture higher-level features, a pooling operation (such as max pooling or average pooling) can be further applied to process the feature vector.

[0064] S5, obtaining the hydrogen production power of the hydrogen production device at multiple time points within a predetermined time period, and preprocessing the collected data;

[0065] S6, arranging the hydrogen production power of the hydrogen production device within the predetermined time period into a hydrogen production power input vector in the time dimension,

[0066] S7, inputting the hydrogen production power input vector first into a convolution neural network based on attention mechanism, and then into a convolution neural network based on time mechanism, to obtain a feature vector B = [b1, b2,... b t ] of the hydrogen production power.

[0067] Arranging the hydrogen production efficiency of the hydrogen production device at multiple time points within the predetermined time period into a hydrogen production power input vector in the time dimension;

[0068] Based on the sequence encoder of the trained network model (Clip model can be used), the hydrogen production power input vector is subjected to multi-scale convolution coding to obtain a feature vector B of the hydrogen production power.

[0069] S8, calculating the difference feature vector C = [c1, c2,... c t ] between the feature vector A of the wind power output power value and the feature vector B of the hydrogen production power using the formula

[0070] where μ A , μ B are the average values of the feature vector A of the wind power output power value and the feature vector B of the hydrogen production power, respectively, are the variances of the feature vector A of the wind power output power value and the feature vector B of the hydrogen production power, respectively, and α and β are weighting factors.

[0071] Calculate the average values μ​A , μ B and variance

[0072] According to experience and specific problem Set weight parameters a, b;

[0073] Based on the formula Calculate the element c i In the difference feature vector C of each time point i

[0074] The calculated c i Value of each time point i is combined into a vector C, that is, the difference feature vector C = [c1, c2, …, c t ].

[0075] S9, the difference feature vector C is input into the classifier, and the classifier contains a matrix M = [m1, m2, …, m j ], each column vector m j Represents the feature vector of a control method, the similarity is calculated by using the difference feature vector C and each column vector in the matrix M, and the highest similarity obtained is the classification result, which is used to control the control strategy of the hydrogen production device, so as to control the hydrogen production efficiency;

[0076] S10, if the input power is large, the hydrogen production device reaches the rated hydrogen production power, and the excess wind power is input to the energy storage device, if the input power is small, the energy storage device is started, and the hydrogen production device is controlled to reach the rated hydrogen production power, and the influence of wind power input power fluctuation on the hydrogen production device is reduced.

[0077] Assuming that the predicted wind power input power is greater than the rated power of the hydrogen production device, the classification result of the classifier is enabled, the hydrogen production device is controlled to reach the rated hydrogen production power, and the charging module of the energy storage device is started, and the excess power is stored in the energy storage device;

[0078] Assuming that the predicted wind power input power is less than the required minimum power of the hydrogen production device, the classification result of the classifier is enabled, the discharging module of the energy storage device is started, and the power of the hydrogen production device is controlled to be within the minimum hydrogen production power and the rated power range.

[0079] A hydrogen production device intelligent control system based on offshore wind turbine predicted power, comprising:

[0080] A data acquisition module is used to acquire the surrounding meteorological data D(T) of the offshore wind turbine in a predetermined time period, which mainly includes all factors that may affect the performance of the wind turbine, such as sea waves, temperature, humidity and wind speed, etc.

[0081] an environmental simulation module configured to construct an atmosphere-ocean coupling model using the acquired surrounding meteorological data D(T) of the offshore wind turbine and to simulate the surrounding environment to obtain future predicted surrounding meteorological data D(T future );

[0082] a power prediction module configured to use a wind turbine performance curve to predict offshore wind turbine wind power output predicted power P according to the simulated future predicted surrounding meteorological data D(T future );

[0083] a feature extraction module configured to extract a feature vector A from the offshore wind turbine wind power output predicted power P using a one-dimensional convolution layer time series encoder;

[0084] a hydrogen production data processing module configured to collect and pre-process hydrogen production power data of a hydrogen production device within a predetermined time period, convert the hydrogen production power data into a hydrogen production power input vector, and then extract a hydrogen production feature power feature vector B from the hydrogen production power input vector using a multi-scale convolutional encoder;

[0085] a feature comparison module configured to perform multi-feature value weight difference on the extracted feature vector A in the offshore wind turbine wind power output predicted power P and the hydrogen production feature power feature vector B to obtain a difference feature vector C;

[0086] an intelligent control module configured to input the difference feature vector C into a classifier to obtain a classification result, which is used to guide a hydrogen production efficiency control strategy of the hydrogen production device; when the input power exceeds the rated power of the hydrogen production device, the excess energy is stored; and when the input power is insufficient, energy is supplemented from an energy storage device to maintain stable operation of the hydrogen production device.

[0087] The above-described embodiments only express specific implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the present application. 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 protection of the present application.

Claims

1. A hydrogen production control method based on predicted power, characterized in that, Includes the following steps: S1. Obtain meteorological data surrounding the offshore wind turbine within a predetermined time period. ; S2. Utilize the acquired surrounding meteorological data An atmosphere-ocean coupled model was constructed to simulate the environment surrounding offshore wind turbines and obtain predicted surrounding meteorological data. ; S3. Based on the measured surrounding meteorological data The predicted output power of offshore wind turbines is determined using known wind turbine performance curves. ,in Indicates a point in time The wind power output value; S4. Extract the feature vector of wind power output value ; S5. Obtain the hydrogen production power of the hydrogen production unit within a predetermined time period, arrange it into a hydrogen production power input vector according to the time dimension, and extract the feature vector of the hydrogen production power. ; S6. Calculate the difference eigenvector between the eigenvector A of wind power output power and the eigenvector B of hydrogen production power. ; S7. Input the difference feature vector C into the classifier, which contains a matrix. Each column vector The feature vector representing a control method is used to calculate the similarity between the difference feature vector C and each column vector in matrix M. The one with the highest similarity is the classification result, and the hydrogen production device is controlled according to the classification result. S8. If the input power is large, the hydrogen production unit will reach its rated hydrogen production power, and the excess wind power will be input to the energy storage device. If the input power is small, the energy storage device will be activated to control the hydrogen production unit to reach its rated hydrogen production power, thereby reducing the impact of the fluctuation of offshore wind power input on the hydrogen production unit.

2. The hydrogen production control method based on predicted power according to claim 1, characterized in that, The surrounding meteorological data Including ocean waves, temperature, humidity, and wind speed; In S2, historical meteorological data from the surrounding area of ​​the sensors and observation stations are acquired, and the data is preprocessed to ensure data quality and consistency. Historical meteorological datasets are divided into training, validation, and test sets, which are then input into a machine learning model for training. The model parameters are then iteratively optimized to minimize prediction errors. The trained model is applied to predict meteorological data around offshore wind turbines, generating predicted surrounding meteorological data. The model will be updated and optimized based on this data and future surrounding meteorological data.

3. The hydrogen production control method based on predicted power according to claim 1, characterized in that, In S3, machine learning algorithms are introduced to dynamically adjust performance parameters and periodically update the performance curves of offshore wind turbines to reflect changes in actual operating conditions.

4. The hydrogen production control method based on predicted power according to claim 1, characterized in that, In S4, the predicted wind power output will be... A temporal encoder using a one-dimensional convolutional layer utilizes the formula To obtain the eigenvector of wind power output power value ; Determine the kernel size, select an appropriate kernel size K, and initialize a set of weights for each kernel. and a bias term The weight and bias terms It is automatically adjusted through the backpropagation algorithm during the training process; Applying convolution operations to predict the output power of offshore wind turbines For each position Through formula Calculate the convolution result at this position; in, It is the kernel length. It is a bias term. In position Output feature values; Repeat the above process until all possible locations are covered. The final result This constitutes the characteristic vector A of the wind power output power value.

5. The hydrogen production control method based on predicted power according to claim 4, characterized in that, In S5, the hydrogen production power input vector is first fed into a convolutional neural network based on an attention mechanism, and then into a convolutional neural network based on a time mechanism, thereby obtaining the feature vector of hydrogen production feature power. .

6. The hydrogen production control method based on predicted power according to claim 5, characterized in that, Using formula Calculate the difference eigenvector C; in, , These are the eigenvectors of the wind power output value. Eigenvectors of hydrogen production characteristic power The average value, , These are the eigenvectors of the wind power output value. Eigenvectors of hydrogen production characteristic power variance and It is a weighting factor.

7. The hydrogen production control method based on predicted power according to claim 1, characterized in that, In S8, assuming that the predicted output power P of the offshore wind turbine is greater than the rated power of the hydrogen production device, the classification result of the classifier is used to control the hydrogen production device to reach the rated hydrogen production power and start the charging module of the energy storage device to store the excess electrical energy in the energy storage device. If the predicted output power P of the offshore wind turbine is less than the minimum power required by the hydrogen production unit, then the classification result of the classifier is used to start the discharge module of the energy storage device and control the power of the hydrogen production unit within the range of minimum hydrogen production power and rated power, in conjunction with the wind power input power.

8. A hydrogen production control system based on predicted power, used to implement the hydrogen production control method based on predicted power as described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to acquire meteorological data around the offshore wind turbines within a predetermined time period. ; The environmental simulation module utilizes acquired surrounding meteorological data. An atmosphere-ocean coupled model is constructed, and the surrounding environment is simulated to obtain future forecasts of surrounding meteorological data. ; The power prediction module is used to predict surrounding meteorological data based on simulations. The wind turbine performance curve is used to predict the predicted wind power output P of offshore wind turbines; The feature extraction module is used to extract feature vector A from the predicted output power P of offshore wind turbines using a temporal encoder with a one-dimensional convolutional layer; The hydrogen production data processing module is used to collect and preprocess the hydrogen production power data of the hydrogen production device within a predetermined time period, and after converting it into a hydrogen production power input vector, a multi-scale convolutional encoder is used to extract the feature vector B of the hydrogen production feature power from the hydrogen production power input vector. The feature comparison module is used to extract feature vector A from the predicted output power P of offshore wind turbines and feature vector B from the hydrogen production feature power, and perform multi-feature value weighted difference to obtain the difference feature vector C. The intelligent control module is used to input the differential feature vector C into the classifier to obtain the classification result, which is used to guide the hydrogen production efficiency control strategy of the hydrogen production unit.

9. An electronic 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 program, it implements the hydrogen production control method based on predicted power as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the hydrogen production control method based on predicted power as described in any one of claims 1 to 7.

Images