Capacitance energy storage frequency modulation control method, system, equipment and medium

Abstract

The invention discloses a capacitance energy storage frequency modulation control method, system and device and a medium, and the method comprises the steps: decomposing a frequency modulation sequence into a plurality of subsequences through VMD to extract the features of different frequency scales, enabling the features of a complex frequency modulation instruction to be clearer, obtaining the prediction error corresponding to each subsequence, measuring the similarity degree between the sequences through attraction, and obtaining the prediction error corresponding to each subsequence; an iterative merging mode of'merging-substitution-recalculation 'is adopted, sub-sequences with the maximum similarity degree are selected for merging each time, the number of the sub-sequences is gradually reduced, and it is guaranteed that sequence features are optimized in each step of merging. The finally generated combined sequence features are more concentrated, the prediction model can be better adapted, the model input quality and the prediction effect are improved, redundant information is reduced, the sequence structure input into the prediction model is optimized, the prediction precision is improved, and finally the capacitor energy storage frequency modulation control is more accurate and the response is more efficient.

Description

Technical Field

[0001] The present invention belongs to the technical field of frequency modulation instruction prediction, and in particular relates to a capacitor energy storage frequency modulation control method, system, equipment and medium. Background Art

[0002] The traditional hybrid energy storage (supercapacitor + lithium battery) method of assisting the frequency regulation of thermal power units is to transmit the difference between the frequency regulation command and the thermal power unit to the hybrid energy storage, where the low-frequency part is borne by the battery and the high-frequency part is borne by the supercapacitor.

[0003] However, the transmission of the signal (the frequency modulation command is transmitted to the supercapacitor / lithium battery) takes time, and the supercapacitor or lithium battery itself also takes time to respond. This will cause a certain response time difference, further affecting the K value, and further affecting the power plant's profits.

[0004] The traditional prediction method is to decompose the original FM sequence through VMD to generate a series of subsequences (IMF1, IMF2, ..., IMF K ), this series of subsequences are predicted by the GRU network to obtain a series of subsequence prediction values, and then the prediction values ​​of these subsequences are superimposed to obtain the final prediction result. However, the disadvantage of this method is that due to the inherent disadvantage of VMD decomposition, there may be similarities between two subsequences (for example, the decomposed IMF1 and IMF3 can actually be replaced by one sequence), which will lead to a decrease in prediction accuracy. Summary of the Invention

[0005] The purpose of the present invention is to provide a capacitor energy storage frequency modulation control method, system, device and medium to solve the technical problem of low prediction efficiency and accuracy caused by redundant subsequences in the VMD decomposition process in existing frequency modulation prediction methods.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: A method for controlling frequency modulation of capacitor energy storage comprises the following steps: Use the frequency modulation instruction VMD Decompose to obtain multiple subsequences; Inputting multiple subsequences into the frequency modulation instruction prediction model respectively to obtain frequency modulation instruction prediction values, and obtaining prediction errors based on the frequency modulation instruction prediction values ​​and actual values; Based on each subsequence and the prediction error value of each subsequence, the similarity between any two subsequences is obtained using the attraction calculation formula; Merge multiple subsequences in descending order of similarity; The combined sequence is input into the frequency modulation instruction prediction model to predict the frequency modulation instruction.

[0007] Furthermore, the VMD decomposition layer number is at least 9.

[0008] Furthermore, the initial prediction error obtained based on the frequency modulation instruction predicted value and actual value is specifically: The FM instructions in each subsequence are divided into a training set and a validation set; Inputting the training sets in the multiple subsequences into the frequency modulation instruction prediction model to obtain the frequency modulation instruction prediction value; The prediction error is obtained based on the FM instruction predicted value and the actual value in the validation set.

[0009] Furthermore, the attractiveness calculation formula is: ; Where, F(i, j) Represents a subsequence i and subsequences j The attraction between Indicates the change similarity ratio; Represents a subsequence i Middle i The element value at position Represents a subsequence j Zhongyu The element value at the corresponding position; N is the number of data points; Represents a subsequence j The prediction error, Represents a subsequence i prediction error.

[0010] Furthermore, the change similarity ratio The calculation formula is:

[0011] Where, M Represents a subsequence i and subsequences j The number of similar values ​​between them; N -1 indicates a subsequence i and subsequences j The sequence length.

[0012] Furthermore, the step of merging the multiple subsequences in descending order of similarity is as follows: Merge the two subsequences with the greatest similarity to obtain a new subsequence; The new subsequence replaces the two merged subsequences and participates in the calculation and merging of the attraction values ​​of the remaining subsequences until all subsequences are merged.

[0013] Furthermore, the frequency modulation instruction prediction model is based on GRU Network construction.

[0014] In a second aspect, the present invention provides a capacitor energy storage frequency modulation control system, comprising a decomposition module, a first calculation module, a second calculation module, a merging module and a prediction output module, wherein: Decomposition module: used to use the frequency modulation instruction VMD Decompose to obtain multiple subsequences; A first calculation module is used to input the multiple subsequences into the frequency modulation instruction prediction model to obtain the frequency modulation instruction prediction value, and obtain the prediction error based on the frequency modulation instruction prediction value and the actual value; The second calculation module: based on each subsequence and the prediction error value of each subsequence, the similarity between any two subsequences is obtained using the attraction calculation formula; Merging module: Based on the initial prediction error, multiple subsequences are merged in descending order of similarity; Prediction output module: used to input the merged sequence into the frequency modulation instruction prediction model to predict the frequency modulation instruction.

[0015] According to a third aspect, a terminal device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0016] In a fourth aspect, a computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0017] Compared with the prior art, the present invention has the following beneficial technical effects: The present invention discloses a capacitor energy storage frequency modulation control method. The method utilizes VMD to decompose a frequency modulation sequence into multiple subsequences to extract features of different frequency scales, thereby making the features of complex frequency modulation instructions clearer and obtaining the prediction errors corresponding to each subsequence. The method utilizes attraction to measure the similarity between the sequences and then completes sequence merging, thereby reducing redundant information, optimizing the sequence structure input to the prediction model, and improving prediction accuracy. The number of merged subsequences is reduced, computational efficiency is improved, and key features are retained, thereby improving the prediction accuracy of frequency modulation instructions. Ultimately, the capacitor energy storage frequency modulation control is more accurate and the response is more efficient.

[0018] Preferably, the attractiveness calculation formula can comprehensively measure the similarity between subsequences, avoid the one-sidedness of single-dimensional judgment, and make the merging of subsequences more reasonable.

[0019] Preferably, an iterative merging approach of "merge → replace → recalculate" is employed, whereby the most similar subsequences are selected for merging each time, gradually reducing the number of subsequences to ensure that sequence features are optimized at each merging step. The resulting merged sequence features are more concentrated, better adapted to the prediction model, and improve both model input quality and prediction performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flow chart of a capacitor energy storage frequency modulation control method in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0022] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0023] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0024] Grid frequency regulation commands (such as AGC signals) can be decomposed into high-frequency fluctuation components (typically with a period of <1 second, such as load transients and sudden changes in renewable energy output) and low-frequency trend components (with a period of >1 second, such as slowly changing loads and the response of thermal power unit speed regulation systems). Supercapacitors, due to their high power density and fast response speed (millisecond-level charging and discharging), are suitable for absorbing / releasing high-frequency power components and quickly suppressing frequency fluctuations. Lithium batteries, with their high energy density but slow response speed (seconds), are suitable for absorbing low-frequency power components and providing continuous energy replenishment, thus avoiding the life loss of supercapacitors due to high-frequency reciprocating charging and discharging. However, the above frequency regulation method requires real-time collection of data such as thermal power unit output, energy storage SOC, and grid frequency. The communication delay must be less than 50ms, which places high demands on real-time communication, thus affecting the K value.

[0025] The present invention is described in further detail below with reference to the accompanying drawings: like Figure 1 As shown, a capacitor energy storage frequency modulation control method includes the following steps: Step 1: Decompose the frequency modulation instruction using VMD to obtain multiple subsequences; Specifically, VMD (Variational Mode Decomposition) is a signal processing method based on the variational principle. It is used to decompose complex signals into multiple physically meaningful intrinsic mode functions. It achieves adaptive decomposition of signals through variational optimization, improving anti-interference capabilities while maintaining physical interpretability. It is suitable for the analysis of non-stationary and multi-scale signals.

[0026] Optionally, the number of VMD decomposition layers is at least 9.

[0027] Step 2: Input the multiple subsequences into the frequency modulation instruction prediction model to obtain the frequency modulation instruction prediction value, and obtain the prediction error based on the frequency modulation instruction prediction value and the actual value; Specifically, the frequency modulation instruction is decomposed using VMD to obtain multiple subsequences. Each subsequence is input into the frequency modulation instruction prediction model to obtain a prediction error. All prediction errors constitute a prediction error sequence, and the number of prediction errors is the same as the number of VMD decomposition layers.

[0028] By dividing the training set and the validation set, and calculating the error based on the predicted value and the actual value, the prediction error acquisition process is standardized to ensure that the error data truly reflects the model prediction deviation, providing a reliable basis for subsequence merging and model optimization, and improving the stability and credibility of the method. The initial prediction error based on the frequency modulation instruction predicted value and the actual value is specifically: The FM instructions in each subsequence are divided into a training set and a validation set; Inputting the training sets in the multiple subsequences into the frequency modulation instruction prediction model to obtain the frequency modulation instruction prediction value; The prediction error is obtained based on the FM instruction predicted value and the actual value in the validation set.

[0029] Optionally, the prediction error uses the mean absolute percentage error, which is calculated as:

[0030] in, is the actual value, is the predicted value, n is the number of data points, and MAPE is the mean absolute percentage error.

[0031] Step 3: Based on each subsequence and the prediction error value of each subsequence, the similarity between any two subsequences is obtained using the attraction calculation formula; The formula for calculating attractiveness is: ; Where, F(i, j) Represents a subsequence i and subsequences j The attraction between Indicates the change similarity ratio; Represents a subsequence i Middle i The element value at position Represents a subsequence j Zhongyu The element value at the corresponding position; N is the number of data points; Represents a subsequence j The prediction error, Represents a subsequence i prediction error.

[0032] Calculated attractiveness value F(i, j) The larger the value is, the greater the similarity between the two subsequences is. The attraction calculation formula can comprehensively measure the similarity between subsequences, avoid the one-sidedness of single-dimensional judgment, and make the merging of subsequences more reasonable.

[0033] Among them, the change similarity ratio The calculation formula is:

[0034] Where, M Represents a subsequence i and subsequences j The number of similar values ​​between them; N -1 indicates a subsequence i and subsequences j The sequence length.

[0035] Step 4: Merge multiple subsequences in descending order of similarity; The specific steps of merging are: Merge the two subsequences with the greatest similarity to obtain a new subsequence; The new subsequence replaces the two merged subsequences and participates in the calculation and merging of the attraction values ​​of the remaining subsequences until all subsequences are merged.

[0036] Step 5: Input the combined sequence into the frequency modulation instruction prediction model to predict the frequency modulation instruction.

[0037] Among them, the frequency modulation instruction prediction model is based on GRU Network construction.

[0038] The present invention uses VMD to decompose the frequency modulation sequence into multiple subsequences to extract features of different frequency scales, making the features of complex frequency modulation instructions clearer, and obtaining the prediction error corresponding to each subsequence. The similarity between the sequences is measured by attraction to complete the sequence merging, reduce redundant information, optimize the sequence structure input to the prediction model, and improve the prediction accuracy. The number of subsequences after merging is reduced, the calculation efficiency is improved, and the key features are retained at the same time, thereby improving the prediction accuracy of the frequency modulation instructions, and ultimately making the capacitor energy storage frequency modulation control more accurate and the response more efficient.

[0039] In another embodiment of the present invention, a method for controlling frequency modulation of capacitor energy storage is provided, wherein the frequency modulation instruction (or a subsequence) is P t =[X1,X2,X3,X4,...,X N ], the front length is 0.9N[X1,X2,X3,X4,...,X 0.9N ] sequence as input to predict the experimental group [X 0.9N+1 ,X 0.9N+2 ,X 0.9N+3 ,...,X N ] and the unknown group [X N+1 ,X N+2 ,X N+3 ,...,X N+0.1N ] values. The actual value of the test group and the predicted value generate all subsequent errors, and the present invention actually predicts the value of the unknown group.

[0040] First, the frequency modulation instruction P t =[X1,X2,X3,X4,...,X N ]Use VMD decomposition with the decomposition layer number K (K ≥ 9) to generate a series of subsequences (IMF1, IMF2, ... IMF i ..,IMF K), put them into the FM instruction prediction model built by the GRU network, and get the errors of each test group as [W1,W2,..Wi,..Wk]. The specific steps are: The FM instructions in each subsequence are divided into a training set and a validation set; Inputting the training sets in the multiple subsequences into the frequency modulation instruction prediction model to obtain the frequency modulation instruction prediction value; The prediction error is obtained based on the FM instruction predicted value and the actual value in the validation set.

[0041] The present invention proposes an attraction calculation formula to express the similarity between two subsequences. Suppose two subsequences are IMF j =[x j1 ,x j2 ,.x ji ,..x jN ],IMF i =[x i1 ,x i2 ,.x ii ,..x iN ] The corresponding experimental group errors are W j and W i Then the attractiveness calculation formula is: ; Where, F(i, j) Represents a subsequence i and subsequences j The attraction between Indicates the change similarity ratio; Represents a subsequence i Middle i The element value at position Represents a subsequence j Zhongyu The element value at the corresponding position; N is the number of data points; Represents a subsequence j The prediction error, Represents a subsequence i prediction error.

[0042] Establish IMF j The sinusoidal slope sequence of is:

[0043] Establish IMF i The sinusoidal slope sequence of is:

[0044] Obviously, the IMF j and the IMFi The length of the sequence is N-1, and there are M similar values ​​between them (similar definitions, such as IMF j The fifth value is 0.9, IMF i The fifth value is 0.85, and the absolute value of the corresponding difference is less than 0.08, so the fifth group is similar.

[0045] Change similarity ratio The calculation formula is:

[0046] Where M represents the number of similar values ​​between subsequence i and subsequence j; N-1 represents the sequence length of subsequence i and subsequence j.

[0047] (IMF1,IMF2,...IMF i ..,IMF K ) Use the attraction formula to calculate between each pair (exhaust all combinations), calculate k(k-1) / 2 results, first select the two groups of subsequences with the largest value here to merge (for example, if the attraction value between IMF1 and IMF3 is the largest, then IMF1 and IMF3 should be merged. The specific merging method is: let IMF1=[x 11 ,x 12 ,...,x 1i ,..x 1N ],IMF3=[x 31 ,x 32 ,...,x 3i ,..x 3N ], the calculation formula for the data value in the merged sequence is: ), then merge the next largest, then merge the third largest, until the final sequence is obtained.

[0048] Traditional VMD decomposition may produce subsequences with high similarity. After merging, redundant information can be eliminated. The number of subsequences after merging is reduced, which effectively reduces the amount of calculation and improves the system's rapid response capability and prediction accuracy to frequency modulation instructions.

[0049] Finally, the final merged sequence is put into GRU for prediction.

[0050] This invention calculates the attractiveness of subsequences after VMD decomposition and merges similar subsequences. This overcomes the problem of reduced prediction accuracy caused by similar subsequences in traditional VMD decomposition, thereby improving the accuracy of frequency modulation command prediction. Predicting the magnitude of frequency modulation commands in advance allows for early activation of supercapacitors / batteries, reducing the impact of response time differences on the K value, thereby increasing the K value and boosting power plant revenue.

[0051] For example, suppose the frequency modulation instruction sequence is P t =[X1,X2,X3,X4,...,X N ], where N is the total length of the sequence. Taking N=1000 as an example, we take the first 0.9N=900 data points [X1,X2,...,X 900 ] as the input layer training data of the GRU network.

[0052] The experimental group was [X 901 ,X 902 ,...,X 1000 ], a total of 100 data points, used for error checking after model training.

[0053] The unknown group is [X 1001 ,X 1002 ,...,X 1100 ], a total of 100 data points, which are the final prediction targets of the model.

[0054] Take K=12 (satisfying K≥9), input the frequency modulation instruction sequence P t , generate 12 subsequences IMF1,IMF2,...IMF through VMD algorithm 12 .

[0055] Subsequence representation: Let IMF j =[x j1 ,x j2 ,.x ji ,..x jN ], IMF i =[x i1 ,x i2 ,.x ii ,..x iN ], where N=1000 (the same length as the original sequence).

[0056] Experimental group error: IMF through GRU network j After training, calculate the experimental group [X 901 ,X 902 ,...,X 1000 ] is the root mean square error (RMSE), , similarly, W i .

[0057] Calculating IMF j The sinusoidal slope sequence S j =[S j1 , S j2 ,…, S j(N-1) ], where S jK =arctan(Δt S j(k+1)-S jk )× , Δt=1s (time interval).

[0058] The same goes for the IMF i The sinusoidal slope sequence S i : Define the similarity judgment threshold δ=0.08, traverse S j and S i The corresponding position element in ∣S jK -S iK If ∣≤δ, it is determined to be a similar point.

[0059] Number of similarities , where II is the indicator function.

[0060] calculate , the value range is [0,1], and the larger the value, the more similar the slope changes are.

[0061] Assume that the experimental group errors of IMF1 and IMF3 are W1=0.5MW and W3=0.55MW respectively.

[0062] Calculated M=750, N-1=999, then =999750≈0.751; the attraction is: .

[0063] Calculate 212×11=66 groups of attraction values ​​for the 12 subsequences to form a 12×12 attraction matrix Fmatrix.

[0064] The maximum value in the matrix is ​​F(1,3)=15.02, the second largest value is F(2,5)=12.35, and the third largest value is F(1,5)=10.78.

[0065] First merger: Select IMF1 and IMF3 corresponding to F(1,3) and merge them using the weighted average method: ; Specifically: ; Secondary merging: Process the second largest value F(2,5) and merge IMF2 and IMF5 using the same method as above.

[0066] Merge three times: Merge IMF1-3 (already merged) and IMF5 corresponding to F(1,5). At this time, the error of the merged sequence needs to be recalculated: W1-3-5=31(W12+W32+W52).

[0067] There are initially 12 subsequences, and after 3 mergers, there are 12-3=9 sequences remaining, reducing data redundancy by 33.3%.

[0068]

[0069] In yet another embodiment of the present invention, a capacitor energy storage frequency modulation control system is provided, comprising a decomposition module, a first calculation module, a second calculation module, a merging module, and a prediction output module, wherein: Decomposition module: used to decompose the frequency modulation instruction into multiple subsequences using VMD; A first calculation module is used to input the multiple subsequences into the frequency modulation instruction prediction model to obtain the frequency modulation instruction prediction value, and obtain the prediction error based on the frequency modulation instruction prediction value and the actual value; The second calculation module: based on each subsequence and the prediction error value of each subsequence, the similarity between any two subsequences is obtained using the attraction calculation formula; Merging module: Based on the initial prediction error, multiple subsequences are merged in descending order of similarity; Prediction output module: used to input the merged sequence into the frequency modulation instruction prediction model to predict the frequency modulation instruction.

[0070] The present invention uses VMD decomposition and patented technology to achieve multi-scale feature extraction of frequency modulation instructions. Combined with attraction-driven subsequence merging, it breaks through the bottleneck of data redundancy in traditional VMD decomposition and provides key technical support for improving the frequency modulation efficiency of thermal power units.

[0071] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0072] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0073] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0074] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit its scope of protection. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that after reading the present invention, those skilled in the art may still make various changes, modifications or equivalent substitutions to the specific implementation methods of the invention, but these changes, modifications or equivalent substitutions are all within the scope of protection of the pending claims of the invention.

Claims

1. A method for controlling frequency modulation of capacitor energy storage, characterized in that: The following steps are involved: Decompose the frequency modulation instruction using VMD to obtain multiple subsequences; Inputting multiple subsequences into the frequency modulation instruction prediction model respectively to obtain frequency modulation instruction prediction values, and obtaining prediction errors based on the frequency modulation instruction prediction values ​​and actual values; Based on each subsequence and the prediction error value of each subsequence, the similarity between any two subsequences is obtained using the attraction calculation formula; Merge multiple subsequences in descending order of similarity; The combined sequence is input into the frequency modulation instruction prediction model to predict the frequency modulation instruction.

2. A capacitor energy storage frequency modulation control method according to claim 1, characterized in that: The VMD decomposition layer number is at least 9.

3. A capacitor energy storage frequency modulation control method according to claim 1, characterized in that: The prediction error based on the frequency modulation instruction prediction value and actual value is specifically: The FM instructions in each subsequence are divided into a training set and a validation set; Inputting the training sets in the multiple subsequences into the frequency modulation instruction prediction model to obtain the frequency modulation instruction prediction value; The prediction error is obtained based on the FM instruction predicted value and the actual value in the validation set.

4. A capacitor energy storage frequency modulation control method according to claim 1, characterized in that: The attractiveness calculation formula is: ； Where, F(i, j) Represents a subsequence i and subsequences j The attraction between Indicates the change similarity ratio; Represents a subsequence i Middle i The element value at position Represents a subsequence j Zhongyu The element value at the corresponding position; N is the number of data points; Represents a subsequence j of Prediction error , represents the prediction error of subsequence i.

5. A capacitor energy storage frequency modulation control method according to claim 4, characterized in that: The change similarity ratio The calculation formula is: Where, M Represents a subsequence i and subsequences j The number of similar values ​​between them; N -1 indicates a subsequence i and subsequences j The sequence length.

6. A capacitor energy storage frequency modulation control method according to claim 1, characterized in that: The steps of merging the multiple subsequences in descending order of similarity are as follows: Merge the two subsequences with the greatest similarity to obtain a new subsequence; The new subsequence replaces the two merged subsequences and participates in the calculation and merging of the attraction values ​​of the remaining subsequences until all subsequences are merged.

7. The method for controlling frequency modulation of capacitor energy storage according to claim 1, characterized in that: The frequency modulation instruction prediction model is based on GRU Network construction.

8. A capacitor energy storage frequency modulation control system, characterized in that: A capacitor energy storage frequency modulation control method according to any one of claims 1 to 7, comprising a decomposition module, a first calculation module, a second calculation module, a merging module and a prediction output module, wherein: Decomposition module: used to use the frequency modulation instruction VMD Decompose to obtain multiple subsequences; A first calculation module is used to input multiple subsequences into the frequency modulation instruction prediction model to obtain frequency modulation instruction prediction values, and obtain prediction errors based on the frequency modulation instruction prediction values ​​and actual values; The second calculation module: based on each subsequence and the prediction error value of each subsequence, the similarity between any two subsequences is obtained using the attraction calculation formula; Merge module: merge multiple subsequences in descending order of similarity; Prediction output module: used to input the merged sequence into the frequency modulation instruction prediction model to predict the frequency modulation instruction.

9. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images