Energy storage assisted black start wind power prediction system and method

Abstract

The invention relates to the technical field of wind power, and discloses an energy storage assisted black-start wind power prediction system and method, and the prediction system consists of a data acquisition module, a VMD decomposition module, a prediction module, a ratio error calculation module, a change degree calculation module, a fine tuning module, a similarity calculation module and a result output module. According to the method, the original wind speed sequence is subjected to multi-level decomposition by using the variational mode decomposition VMD, a complex wind speed signal is decomposed into a plurality of intrinsic mode functions (IMF), the decomposed sub-sequence is input into the pre-trained prediction model, such as a recurrent neural network (RNN), the characteristics of time sequence data are fully utilized, the prediction accuracy is improved, and the prediction efficiency is improved. The RNN can better capture the dynamic characteristics of the wind speed change through the learning ability of the RNN for the time sequence data change, and achieves the beneficial effect of improving the wind speed prediction accuracy.

Description

Technical Field

[0001] This invention relates to the field of wind power technology, specifically to an energy storage-assisted black-start wind power prediction system and method. Background Technology

[0002] When a wind-solar-storage power generation system is used as a black-start power source, considering the charging / discharging power constraints and energy constraints of the energy storage device, during the black-start process, if the output of the wind farm and photovoltaic power station is insufficient or fluctuates drastically, overcharging and over-discharging of the energy storage may occur, leading to the inability to continue utilizing the energy storage and causing black-start failure. This necessitates evaluating the probability of continuous effective output of the wind farm based on historical wind speed data. However, traditional wind speed prediction suffers from poor prediction accuracy. To address the problems of traditional prediction, the existing technology (publication number CN 117039895) is relevant. A wind power prediction method and system for energy storage-assisted black start is described. The method includes: obtaining predicted historical wind speed components from historical wind speed data using the EEMD algorithm and a recurrent neural network model; constructing an error coefficient function based on the historical wind speed components, their weights, and the predicted historical wind speed component values; using the error coefficient function as a fitness function and employing an improved particle algorithm to obtain the target weights of each historical wind speed component when the fitness value is minimized; obtaining multiple predicted real-time wind speed components from real-time wind speed data based on the EEMD algorithm and the recurrent neural network model; and obtaining a wind speed prediction value and subsequently a wind power prediction value based on the multiple predicted real-time wind speed components and their corresponding target weights.

[0003] However, if the quality of historical wind speed data is poor—for example, if the data is missing or incorrect—it will directly affect the decomposition results and prediction accuracy. Especially under extreme weather conditions, historical data cannot fully represent future wind speed changes, leading to prediction errors. Furthermore, inaccurate wind speed predictions can further cause the energy storage system to fail to fully charge to meet the energy requirements for black start, or the energy storage system to fail to discharge in time after overcharging, resulting in start-up failure. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an energy storage-assisted black start wind power prediction system and method, which has advantages such as improving the accuracy of wind speed prediction and avoiding black start failure caused by overcharging and over-discharging of energy storage, thus solving the problems mentioned above.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an energy storage-assisted black start wind power prediction system, comprising a data acquisition module, a VMD decomposition module, a prediction module, a ratio error calculation module, a degree of change calculation module, a fine-tuning module, a similarity calculation module, and a result output module; The data acquisition module is used to acquire raw wind speed sequence data; The VMD decomposition module is used to perform variational mode decomposition (VMD) on the original wind speed sequence to obtain multiple sub-sequences IMF. The prediction module is used to input the decomposed subsequence IMF into the prediction model to obtain the predicted total aliasing degree of each subsequence; The ratio error calculation module is used to calculate the ratio error of each subsequence based on the predicted total aliasing degree and the actual total aliasing degree. The degree of change calculation module is used to calculate the degree of change of each subsequence based on the ratio error; The fine-tuning module is used to fine-tune each subsequence according to the degree of change; The similarity calculation module is used to calculate the similarity between the fine-tuned subsequence and the original subsequence; The result output module is used to determine whether the similarity is greater than 0.9 and output the final result.

[0006] Preferably, 70% of the subsequence IMF is used for training, the remaining 20% ​​is used for validation, and the remaining 10% is used for prediction.

[0007] Preferably, the sample size used for validation in the subsequence IMF is N, and the ratio error of the first point is... , No. A little thing The degree of change is calculated as follows:

[0008] The ratio error is: .

[0009] Preferably, the fine-tuning module fine-tunes each subsequence according to the degree of change, specifically using the following formula:

[0010] in: This is a fine-tuning factor used to adjust the first... The amplitude of each subsequence; based on the degree of change The calculation yields the result used to control the degree of fine-tuning of the subsequence. It is a non-negative number, ranging from [0, 1]. For the first The original subsequences are obtained by variational mode decomposition (VMD). The fine-tuned first Subsequences; The similarity calculation module also includes summing the fine-tuned sequence and summing the subsequences before fine-tuning, then comparing each of them, adding up the ratios of each item and dividing by the sample size.

[0011] A method for predicting black-start wind power assisted by energy storage, wherein the prediction method uses the prediction system described in claim 1, and the prediction method includes the following steps: The original wind speed sequence is decomposed using VMD, with the number of decomposition layers set to K, resulting in a corresponding number of original subsequences. These original subsequences are then input into a pre-trained prediction model to obtain the total predicted aliasing degree. Based on the total predicted aliasing degree, the degree of change corresponding to each original subsequence is obtained. The original subsequences are then fine-tuned based on their respective degrees of change to obtain new subsequences. Finally, the original subsequences are compared with the new subsequences to determine if they meet preset conditions, and the final result is selected for output. Based on the predicted total aliasing degree, the degree of change corresponding to each original subsequence is obtained, including: obtaining a ratio error based on the predicted total aliasing degree and the actual total aliasing degree; and determining the degree of change corresponding to each original subsequence based on the ratio error. The degree of change is determined by the following formula:

[0012] in: Characterizing the first The degree of change of each subsequence Characterizing the first The ratio error at each point Characterizes the size of the sample set; Determining whether a preset condition is met includes: calculating the similarity between the original subsequence and the new subsequence; determining whether the similarity is greater than 0.9; if the similarity is greater than 0.9, then outputting the new subsequence; if the similarity is not greater than 0.9, then outputting the original subsequence.

[0013] Preferably, the VMD decomposition calculation specifically includes: The goal of VMD is to decompose signals. For multiple intrinsic mode functions (IMFs) and a slow trend:

[0014] in: It is the first An inherent pattern It is the corresponding frequency. It is the number of decomposition levels. It is a regularization parameter. Indicates the first The central frequency of the IMF.

[0015] Preferably, the specific process of VMD decomposition calculation is as follows: S1.1, Initialize all and ; S1.2, Iterate until convergence;

[0016]

[0017] S1.3, Update in each iteration and The value; Indicates Fourier transform, This represents the L2 norm.

[0018] Preferably: the predicted total aliasing is calculated as follows: Suppose there are K subsequences obtained from the decomposition. Each subsequence has a corresponding predicted value. and actual value The total degree of aliasing is predicted and calculated using the following formula: Calculate the degree of aliasing for each predicted subsequence:

[0019] in: It is the sample size. It is the first The degree of aliasing of each subsequence, For at a certain point in time The actual value, These are model predictions; Indicates absolute value / error.

[0020] Preferably, the predicted total aliasing calculation further includes calculating the predicted total aliasing degree of all subsequences:

[0021] The predicted overall aliasing degree is obtained by averaging the aliasing degree of all subsequences. .

[0022] Preferably, the similarity calculation is based on the mean squared error.

[0023] The similarity is calculated to determine whether it meets the preset conditions.

[0024] in: It is the number of sample points. It is set according to the specific application scenario.

[0025] Compared with the prior art, the present invention has the following beneficial effects: This invention provides an energy storage-assisted black-start wind power prediction system. It uses Variational Mode Decomposition (VMD) to perform multi-level decomposition of the original wind speed sequence, breaking down the complex wind speed signal into multiple Intrinsic Mode Functions (IMFs). The decomposed subsequences are then input into a pre-trained prediction model, such as a Recurrent Neural Network (RNN). This fully utilizes the characteristics of time-series data to improve prediction accuracy. The RNN's ability to learn from changes in time-series data allows it to better capture the dynamic characteristics of wind speed variations. By calculating the ratio of the predicted total aliasing to the actual value, the degree of aliasing is determined. This reflects the differences in the prediction performance of each subsequence, providing a quantitative basis for fine-tuning the subsequences and improving the applicability of the prediction model. The fine-tuning module adjusts each subsequence based on the degree of change, using a fine-tuning factor. The amplitude of the subsequences is modified to better adapt each subsequence to changes in actual wind speed data, improving the personalization of the prediction. The reliability of the prediction results is ensured by calculating the similarity between the fine-tuned subsequences and the original subsequences. A similarity threshold of 0.9 is set to effectively eliminate unqualified predictions, guaranteeing the quality of the final output. By using 70% of the subsequences for training, 20% for validation, and 10% for prediction, the model's generalization ability and prediction accuracy are ensured, overfitting is prevented, and its predictive ability on unseen data is improved, achieving the beneficial effect of improving the accuracy of wind speed prediction.

[0026] This invention employs Variational Mode Decomposition (VMD) to decompose the original wind speed sequence into multiple subsequences. This accurately captures wind speed changes at different time scales, making the model more flexible in adapting to various climatic conditions and reducing charging and discharging mismatches in the energy storage system caused by wind speed prediction errors. By inputting the decomposed subsequences into a trained prediction model, such as a recurrent neural network, the system can acquire real-time wind speed trends. The system calculates the ratio error between the predicted total aliasing and the actual value to promptly assess the accuracy of the prediction, thereby determining the degree of change for each subsequence. This allows for dynamic adjustment of the energy storage system's charging and discharging strategy, avoiding overcharging and over-discharging caused by a fixed strategy. If the prediction of certain subsequences is found to be unstable, the system quickly adjusts the energy storage strategy based on feedback, reducing risks and achieving the beneficial effect of avoiding black start failures caused by overcharging and over-discharging of energy storage. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the energy storage-assisted black-start wind power prediction system in an embodiment of the present invention.

[0028] Figure 2 This is a schematic diagram of the energy storage-assisted black-start wind power prediction method in an embodiment of the present invention. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] like Figure 1 As shown, the present invention provides an energy storage-assisted black start wind power prediction system. The prediction system consists of a data acquisition module, a VMD decomposition module, a prediction module, a ratio error calculation module, a degree of change calculation module, a fine-tuning module, a similarity calculation module, and a result output module. The data acquisition module is used to acquire raw wind speed sequence data; The VMD decomposition module is used to perform variational mode decomposition (VMD) on the original wind speed sequence to obtain multiple subsequences IMF. The prediction module is used to input the decomposed subsequences into the prediction model to obtain the predicted total aliasing degree for each subsequence; The ratio error calculation module is used to calculate the ratio error of each subsequence based on the predicted total aliasing degree and the actual total aliasing degree. The variability calculation module is used to calculate the variability of each subsequence based on the ratio error; The fine-tuning module is used to fine-tune each subsequence based on the degree of variability; The similarity calculation module is used to calculate the similarity between the fine-tuned subsequence and the original subsequence; The results output module is used to determine whether the similarity is greater than 0.9 and output the final result.

[0031] Variational mode decomposition (VMD) is employed to break down the original wind speed sequence into multiple subsequences, thus accurately capturing wind speed changes at different time scales. This decomposition capability allows the model to adapt more flexibly to various climatic conditions and reduces the charging and discharging mismatch of energy storage systems caused by wind speed prediction bias.

[0032] By inputting the decomposed subsequences into a trained prediction model such as a recurrent neural network, the changing trend of wind speed can be obtained in real time. This high-precision prediction provides energy storage systems with more accurate charging and discharging instructions, thereby avoiding overcharging or over-discharging caused by inaccurate predictions.

[0033] The system assesses the accuracy of predictions in a timely manner by calculating the ratio error between the predicted total aliasing level and the actual value. This allows for the determination of the variability of each subsequence, enabling dynamic adjustment of the energy storage system's charging and discharging strategy to avoid overcharging and over-discharging caused by a fixed strategy. For example, if the prediction of certain subsequences is found to be unstable, the system quickly adjusts the energy storage strategy based on feedback to reduce risk.

[0034] The fine-tuning module adjusts the amplitude of the subsequence based on the degree of change, making the energy storage system more responsive. By adjusting the predicted input, it better adapts to real-time changes, thereby ensuring that the energy storage system does not overcharge or over-discharge when dealing with load changes.

[0035] Before outputting the final result, the system calculates the similarity between the new subsequence and the original subsequence to determine whether it meets preset conditions. This process ensures the reliability and practicality of the prediction, avoiding inaccurate predictions that could lead to overcharging or over-discharging issues during energy storage devices.

[0036] By acquiring wind speed data in real time, the system can quickly adapt to environmental changes and adjust its charging and discharging strategies accordingly. This real-time data monitoring and adjustment mechanism reduces the risks associated with the energy storage system during black start-up, ensuring that it will not fail due to inappropriate energy storage strategies in the face of unforeseen events or extreme weather conditions.

[0037] The dataset is divided into 70% training, 20% validation, and 10% prediction. This ensures the model has good generalization ability, reduces the risk of overfitting, and has a positive impact on fault tolerance. This means the system can respond more reliably to unseen data, reducing the possibility of overcharging or over-discharging in practical applications.

[0038] Specifically, 70% of the subsequence IMF is used for training, the next 20% for validation, and 10% for prediction.

[0039] Specifically, the validation set size for the subsequence is N, and the ratio error of the first point is... , No. A little thing The degree of change is calculated as follows:

[0040] The ratio error is: .

[0041] Specifically, the calculation formula for the fine-tuning module is as follows:

[0042] in: This is a fine-tuning factor used to adjust the first... The amplitude of each subsequence; based on the degree of change The calculation yields the result used to control the degree of fine-tuning of the subsequence. It is a non-negative number, ranging from [0, 1]. For the first The original subsequences are obtained by variational mode decomposition (VMD). The fine-tuned first Subsequences; The similarity calculation module also includes summing the fine-tuned sequence and summing the subsequences before fine-tuning, then comparing each of them, summing the ratios of each item, and then dividing by the sample size.

[0043] like Figure 2 As shown, this invention provides a method for predicting black-start wind power assisted by energy storage. The prediction method uses the prediction system of claim 1 and includes the following steps: Step 1: Obtain the original wind speed sequence; perform VMD decomposition on the original wind speed sequence, setting the number of decomposition layers to K, to obtain the corresponding number of original subsequences; input the corresponding number of original subsequences into a pre-trained prediction model to obtain the total predicted aliasing degree; based on the total predicted aliasing degree, obtain the degree of change corresponding to each original subsequence; fine-tune the original subsequences based on the degree of change corresponding to each original subsequence to obtain new subsequences; compare the original subsequences with the new subsequences to determine whether they meet the preset conditions, and select to output the final result; Step 2: Based on the predicted total aliasing degree, obtain the degree of change corresponding to each original subsequence, including: obtaining the ratio error between the predicted total aliasing degree and the actual total aliasing degree; and determining the degree of change corresponding to each original subsequence based on the ratio error. The degree of change is determined by the following formula:

[0044] in: Characterizing the first The degree of change of each subsequence Characterizing the first The ratio error at each point Characterizes the size of the sample set; Step 3: Determine whether the preset conditions are met, including: calculating the similarity between the original subsequence and the new subsequence; determining whether the similarity is greater than 0.9; if the similarity is greater than 0.9, output the new subsequence; if the similarity is not greater than 0.9, output the original subsequence.

[0045] Specifically, the VMD decomposition calculation in step one is as follows: The goal of VMD is to decompose signals. For multiple intrinsic mode functions (IMFs) and a slow trend:

[0046] in: It is the first An inherent pattern It is the corresponding frequency. It is the number of decomposition levels. It is a regularization parameter. Indicates the first The central frequency of the IMF.

[0047] Specifically, the steps for VMD decomposition calculation in step one are as follows: S1.1, Initialize all and ; S1.2, Iterate until convergence;

[0048]

[0049] S1.3, Update in each iteration and The value; Indicates Fourier transform, This represents the L2 norm.

[0050] Specifically, the total aliasing prediction in step one is calculated as follows: Suppose there are K subsequences obtained from the decomposition. Each subsequence has a corresponding predicted value. and actual value The total degree of aliasing is predicted and calculated using the following formula: Calculate the degree of aliasing for each predicted subsequence:

[0051] in: It is the sample size. It is the first The degree of aliasing of each subsequence, For at a certain point in time The actual value, These are model predictions; Indicates absolute value / error.

[0052] Specifically, the calculation of the predicted total aliasing also includes calculating the predicted total aliasing degree of all subsequences:

[0053] The predicted overall aliasing degree is obtained by averaging the aliasing degree of all subsequences. .

[0054] Specifically, in step one, the similarity is calculated using the mean squared error:

[0055] The similarity is calculated to determine whether it meets the preset conditions.

[0056] in: It is the number of sample points. It is set according to the specific application scenario.

[0057] Variational mode decomposition (VMD) is used to perform multi-level decomposition of the original wind speed sequence, breaking down the complex wind speed signal into multiple intrinsic mode functions (IMFs). This approach can extract information from different frequency components, helping to capture the characteristics of wind speed changes and thus providing more detailed input data for subsequent predictions.

[0058] The decomposed subsequences are input into a pre-trained prediction model, such as a recurrent neural network (RNN), to fully utilize the characteristics of time series data and improve prediction accuracy. The RNN's ability to learn from changes in time series data allows it to better capture the dynamic characteristics of wind speed variations.

[0059] The degree of variation is obtained by calculating the ratio error between the predicted total aliasing degree and the actual value. This reflects the differences in the prediction performance of each subsequence. This quality feedback mechanism provides a quantitative basis for fine-tuning the subsequences and improves the applicability of the prediction model.

[0060] The fine-tuning module adjusts each subsequence based on the degree of variability, using a fine-tuning factor. This allows for the modification of the amplitude of subsequences. This flexible adjustment mechanism enables each subsequence to better adapt to changes in actual wind speed data, improving the personalization of forecasts.

[0061] The reliability of the prediction results is ensured by calculating the similarity between the fine-tuned subsequence and the original subsequence. A similarity threshold, such as 0.9, is set to effectively eliminate unqualified predictions, guaranteeing the quality of the final output.

[0062] By using 70% of the subsequences for training, 20% for validation, and 10% for prediction, the model's generalization ability and prediction accuracy were ensured. A reasonable data partitioning helps prevent overfitting and improves its predictive ability on unseen data.

[0063] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A black-start wind power prediction system assisted by energy storage, characterized in that, It includes a data acquisition module, a VMD decomposition module, a prediction module, a ratio error calculation module, a degree of change calculation module, a fine-tuning module, a similarity calculation module, and a result output module; The data acquisition module is used to acquire raw wind speed sequence data; The VMD decomposition module is used to perform variational mode decomposition (VMD) on the original wind speed sequence to obtain multiple sub-sequences IMF. The prediction module is used to input the decomposed subsequence IMF into the prediction model to obtain the predicted total aliasing degree of each subsequence; The ratio error calculation module is used to calculate the ratio error of each subsequence based on the predicted total aliasing degree and the actual total aliasing degree. The degree of change calculation module is used to calculate the degree of change of each subsequence based on the ratio error; The fine-tuning module is used to fine-tune each subsequence according to the degree of change; The similarity calculation module is used to calculate the similarity between the fine-tuned subsequence and the original subsequence; The result output module is used to determine whether the similarity is greater than 0.9 and output the final result.

2. The energy storage-assisted black-start wind power prediction system according to claim 1, characterized in that: 70% of the subsequence IMF is used for training, the remaining 20% ​​is used for validation, and the remaining 10% is used for prediction.

3. The energy storage-assisted black-start wind power prediction system according to claim 2, characterized in that: The sample size used for validation in the subsequence IMF is N, and the ratio error of the first point is... , No. A little thing The degree of change is calculated as follows: The ratio error is: .

4. The energy storage-assisted black-start wind power prediction system according to claim 1, characterized in that: The fine-tuning module fine-tunes each subsequence according to the degree of change, specifically using the following formula: in: This is a fine-tuning factor used to adjust the first... The amplitude of each subsequence; based on the degree of change The calculation yields the result used to control the degree of fine-tuning of the subsequence. It is a non-negative number, ranging from [0, 1]. For the first The original subsequences are obtained by variational mode decomposition (VMD). The fine-tuned first Subsequences; The similarity calculation module also includes summing the fine-tuned sequence and summing the subsequences before fine-tuning, then comparing each of them, adding up the ratios of each item and dividing by the sample size.

5. A method for predicting black-start wind power with energy storage assistance, characterized in that, The prediction method uses the prediction system of claim 1 to make predictions, and the prediction method includes the following steps: The original wind speed sequence is decomposed using VMD, and the number of decomposition layers is set to K to obtain the corresponding number of original subsequences. The original subsequences of the corresponding decomposition layer are input into the pre-trained prediction model to obtain the total predicted aliasing degree. Based on the predicted total aliasing degree, the degree of change corresponding to each original subsequence is obtained; the original subsequences are then fine-tuned based on the degree of change corresponding to each original subsequence to obtain new subsequences. The original subsequence is compared with the new subsequence to determine whether it meets the preset conditions, and the final result is selected and output.

6. The energy storage-assisted black start wind power prediction system and method according to claim 5, characterized in that: Based on the predicted total aliasing degree, the degree of change corresponding to each original subsequence is obtained, including: obtaining a ratio error based on the predicted total aliasing degree and the actual total aliasing degree; and determining the degree of change corresponding to each original subsequence based on the ratio error. The degree of change is determined by the following formula: in: Characterizing the first The degree of change of each subsequence Characterizing the first The ratio error at each point Characterizes the size of the sample set; Determining whether a preset condition is met includes: calculating the similarity between the original subsequence and the new subsequence; determining whether the similarity is greater than 0.9; if the similarity is greater than 0.9, then outputting the new subsequence; if the similarity is not greater than 0.9, then outputting the original subsequence.

7. The energy storage-assisted black start wind power prediction system and method according to claim 6, characterized in that: The VMD decomposition calculation specifically includes: The goal of VMD is to decompose signals. For multiple intrinsic mode functions (IMFs) and a slow trend: in: It is the first An inherent pattern It is the corresponding frequency. It is the number of decomposition levels. It is a regularization parameter. Indicates the first The central frequency of each IMF; Specifically, the following steps are included: S1.1, Initialize all and ; S1.2, Iterate until convergence; S1.3, Update in each iteration and The value; Indicates Fourier transform, This represents the L2 norm.

8. The energy storage-assisted black start wind power prediction system and method according to claim 7, characterized in that: The predicted total aliasing is calculated as follows: Suppose there are K subsequences obtained from the decomposition. Each subsequence has a corresponding predicted value. and actual value The total degree of aliasing is predicted and calculated using the following formula: Calculate the degree of aliasing for each predicted subsequence: in: It is the sample size. It is the first The degree of aliasing of each subsequence, For at a certain point in time The actual value, These are model predictions; Indicates absolute value / error.

9. The energy storage-assisted black start wind power prediction system and method according to claim 8, characterized in that: The predicted total aliasing calculation also includes calculating the predicted total aliasing degree of all subsequences: The predicted overall aliasing degree is obtained by averaging the aliasing degree of all subsequences. .

10. The energy storage-assisted black start wind power prediction system and method according to claim 5, characterized in that: The similarity calculation is based on the mean squared error. The similarity is calculated to determine whether it meets the preset conditions. in: It is the number of sample points. It is set according to the specific application scenario.

Images