A super capacitor energy storage frequency modulation control method, system, device and medium

Abstract

The application discloses a kind of super capacitor energy storage frequency modulation control method, system, equipment and medium, by dividing training set and verification set to frequency modulation instruction, in combination with iterative adjustment coefficient, make the frequency modulation control instruction predicted by model more fit actual demand, improve the control precision of super capacitor energy storage frequency modulation;Dynamic adjustment mechanism ensures that the model adapts to the change of frequency modulation instruction, optimizes the response efficiency of energy storage system to grid frequency regulation, improves overall frequency modulation performance;Average absolute percentage error is used as initial prediction error index, which is convenient for quantitative evaluation of model initial prediction effect;Set reasonable error reduction target, both ensure that frequency modulation instruction prediction error is significantly improved, and avoid target too strict to lead to unable to realize, improve the efficiency and feasibility of optimization process.

Description

Technical Field

[0001] This invention belongs to the field of frequency modulation command prediction technology, specifically relating to a method, system, device and medium for frequency modulation control of supercapacitor energy storage. Background Technology

[0002] The traditional method of using hybrid energy storage (supercapacitor + lithium battery) to assist the frequency regulation of thermal power units involves transmitting the difference between the frequency regulation command and the thermal power unit to the hybrid energy storage, with the battery handling the low-frequency portion and the supercapacitor handling the high-frequency portion.

[0003] However, signal transmission (frequency modulation command transmission to the supercapacitor / lithium battery) takes time, and the supercapacitor or lithium battery itself also needs time to respond. This will cause a certain response time difference, which will further affect the K value and thus the power plant's revenue. Summary of the Invention

[0004] The purpose of this invention is to provide a supercapacitor energy storage frequency modulation control method, system, device and medium to solve the technical problems of large error and low efficiency in existing frequency modulation methods.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A method for frequency modulation control of supercapacitor energy storage includes the following steps:

[0007] The frequency modulation commands are divided into a training set and a validation set;

[0008] The training set is input into the frequency modulation command prediction model to obtain the frequency modulation command prediction value. The initial prediction error is obtained based on the frequency modulation command prediction value and the actual value in the validation set.

[0009] The frequency modulation commands in the training set are assigned adjustment coefficients in sequence, and the coefficient values ​​are iteratively adjusted until the corresponding prediction error is less than the initial prediction error. At the same time, the training set after coefficient adjustment and the corresponding prediction error index are obtained.

[0010] If the prediction error index of the training set after coefficient adjustment is greater than the preset error threshold, the dynamic adjustment coefficient is introduced into the training set after coefficient adjustment and iterated until the prediction error index of the training set after the adjustment coefficient and dynamic adjustment coefficient are less than the preset error threshold.

[0011] The frequency modulation command prediction model is retrained using the training set after introducing adjustment coefficients and dynamic adjustment coefficients to obtain a trained frequency modulation command prediction model. The frequency modulation control command is then output through the trained frequency modulation command prediction model.

[0012] Furthermore, the initial prediction error is expressed as mean absolute percentage error.

[0013] Furthermore, the formula for calculating the mean absolute percentage error is as follows:

[0014]

[0015] in, This is the actual value. is the predicted value, n is the number of data points, and MAPE is the mean absolute percentage error.

[0016] Furthermore, the steps of sequentially assigning adjustment coefficients to the frequency modulation commands in the training set, iteratively adjusting the coefficient values ​​until the corresponding prediction error is less than the initial prediction error, and simultaneously obtaining the coefficient-adjusted training set and the corresponding prediction error are as follows:

[0017] First, an adjustment coefficient is assigned to the first frequency modulation command in the training set, while the other frequency modulation commands remain unchanged, thus obtaining the first training set. The value of the adjustment coefficient is continuously adjusted until the prediction error corresponding to the first training set is less than the initial error.

[0018] Clear the adjustment coefficient of the first frequency modulation command, assign the adjustment coefficient to the second frequency modulation command in the training set, and leave the other frequency modulation commands unchanged to obtain the second training set. Continuously adjust the value of the adjustment coefficient until the prediction error corresponding to the second training set is less than the initial error.

[0019] The adjustment coefficients are sequentially assigned to the last frequency modulation command in the training set to obtain multiple training sets. The prediction error of all training sets is less than the initial prediction error, and the frequency modulation coefficients corresponding to the frequency modulation commands in the training sets are obtained.

[0020] All adjustment coefficients are assigned to the frequency modulation command in the training set to obtain the coefficient-adjusted training set, and the corresponding prediction error is obtained based on the coefficient-adjusted training set.

[0021] Furthermore, if the prediction error is always greater than the initial prediction error after the adjustment coefficient has been iterated, then the corresponding adjustment coefficient is assigned a value of 1.

[0022] Furthermore, the iterative formula for the dynamic adjustment coefficient is as follows:

[0023]

[0024] In the formula, This represents the dynamic adjustment coefficient; T represents the iteration number, initially set to 1; N represents the number of data points. For adjustment coefficients; is the prediction error index; gelu() represents the activation function; 0.9MAPE represents the preset error threshold; and i represents the position of the data point.

[0025] Furthermore, the preset error threshold is selected as 0.9 times the initial prediction error.

[0026] In a second aspect, the present invention provides a supercapacitor energy storage frequency regulation control system, comprising a partitioning module, a calculation module, a first iteration module, a second iteration module, and a prediction output module, wherein:

[0027] Partitioning module: Used to divide frequency modulation commands into training and validation sets;

[0028] The calculation module is used to input the training set into the frequency modulation command prediction model to obtain the frequency modulation command prediction value, and to obtain the initial prediction error based on the frequency modulation command prediction value and the actual value in the validation set.

[0029] The first iteration module is used to sequentially assign adjustment coefficients to the frequency modulation instructions in the training set, iterate the adjustment coefficient values ​​until the corresponding prediction error is less than the initial prediction error, and obtain the training set after coefficient adjustment and the corresponding prediction error index.

[0030] Second iteration module: If the prediction error index of the training set after coefficient adjustment is greater than the preset error threshold, the dynamic adjustment coefficient is introduced into the training set after coefficient adjustment and iterated until the prediction error index of the training set after the adjustment coefficient and dynamic adjustment coefficient are less than the preset error threshold.

[0031] Prediction output module: It is used to retrain the frequency modulation command prediction model using the training set after introducing adjustment coefficients and dynamic adjustment coefficients to obtain a trained frequency modulation command prediction model, and output frequency modulation control commands through the trained frequency modulation command prediction model.

[0032] Thirdly, a terminal device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method.

[0033] Fourthly, a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method.

[0034] Compared with the prior art, the present invention has the following beneficial technical effects:

[0035] This invention provides a frequency regulation control method for supercapacitor energy storage. By dividing the frequency regulation command into a training set and a validation set, and combining iterative adjustment coefficients, the frequency regulation control command predicted by the model is made to better match the actual needs, thereby improving the control accuracy of supercapacitor energy storage frequency regulation. At the same time, a dynamic adjustment mechanism is introduced to ensure that the model adapts to changes in the frequency regulation command, optimizes the response efficiency of the energy storage system to grid frequency regulation, and improves the overall frequency regulation performance.

[0036] Preferably, the mean absolute percentage error is used as the initial prediction error index, which facilitates the quantitative evaluation of the model's initial prediction performance.

[0037] Preferably, when assigning an adjustment coefficient to a frequency modulation command, the adjustment coefficients of all previous frequency modulation commands are cleared, the coefficients of a single frequency modulation command are adjusted systematically, interference between coefficients is avoided, the optimal adjustment coefficient for each command is accurately located, and iterative convergence is accelerated.

[0038] Preferably, the preset error threshold is set to 0.9 times the initial error. By setting a reasonable error reduction target, it is ensured that the prediction error of the frequency modulation command is significantly improved, while avoiding the target being too strict and thus unachievable, thereby improving the efficiency and feasibility of the optimization process. Attached Figure Description

[0039] Figure 1 This is a flowchart of a supercapacitor energy storage frequency modulation control method in an embodiment of the present invention. Detailed Implementation

[0040] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0041] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0042] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0043] The present invention will now be described in further detail with reference to the accompanying drawings:

[0044] like Figure 1As shown, a method for frequency modulation control of supercapacitor energy storage includes the following steps:

[0045] Step 1: Divide the frequency modulation commands into a training set and a validation set;

[0046] The frequency modulation (FM) commands are divided into a training set and a validation set according to a certain ratio. The specific application is to use the FM commands in the training set to predict the value of the next FM command segment, and compare the predicted value of the FM command segment with the actual value of the FM command in the corresponding validation set to determine the prediction accuracy of the FM command prediction model.

[0047] Step 2: Input the training set into the frequency modulation command prediction model to obtain the frequency modulation command prediction value, and obtain the initial prediction error based on the frequency modulation command prediction value and the actual value in the validation set;

[0048] Specifically, the initial prediction error is expressed as the mean absolute percentage error. MAPE (Mean Absolute Percentage Error) is a statistical metric used to measure the accuracy of a forecasting model. It reflects the relative deviation of the forecast results by calculating the average absolute percentage error between the predicted and actual values. It is particularly suitable for evaluating errors in time series or numerical forecasts, and is more sensitive to small errors. The formula for calculating MAPE is:

[0049]

[0050] in, This is the actual value. is the predicted value, n is the number of data points, MAPE is the mean absolute percentage error, and i represents the location of the data point.

[0051] Step 3: Assign adjustment coefficients to the frequency modulation commands in the training set in sequence, iterate the adjustment coefficient values ​​until the corresponding prediction error is less than the initial prediction error, and obtain the training set after coefficient adjustment and the corresponding prediction error index.

[0052] The prediction error index is ;

[0053] The training set after coefficient adjustment is represented as [ , , , ,..., Using values ​​of length 0.9N as the training set to predict subsequent values ​​of length 0.1N, the resulting error is the prediction error metric. ;

[0054] The specific steps are as follows:

[0055] The steps for sequentially assigning adjustment coefficients to the frequency modulation commands in the training set, iteratively adjusting the coefficient values ​​until the corresponding prediction error is less than the initial prediction error, and obtaining the adjusted training set and the corresponding prediction error are as follows:

[0056] First, an adjustment coefficient is assigned to the first frequency modulation command in the training set, while the other frequency modulation commands remain unchanged, thus obtaining the first training set. The value of the adjustment coefficient is continuously adjusted until the prediction error corresponding to the first training set is less than the initial error.

[0057] Clear the adjustment coefficient of the first frequency modulation command, assign the adjustment coefficient to the second frequency modulation command in the training set, and leave the other frequency modulation commands unchanged to obtain the second training set. Continuously adjust the value of the adjustment coefficient until the prediction error corresponding to the second training set is less than the initial error.

[0058] The adjustment coefficients are sequentially assigned to the last frequency modulation command in the training set to obtain multiple training sets. The prediction error of all training sets is less than the initial prediction error, and the frequency modulation coefficients corresponding to the frequency modulation commands in the training sets are obtained.

[0059] All adjustment coefficients are assigned to the frequency modulation command in the training set to obtain the coefficient-adjusted training set. The corresponding prediction error is then obtained based on the coefficient-adjusted training set. .

[0060] When the corresponding prediction error is less than the initial prediction error, the iteration stops and the adjustment coefficient value is determined. If the prediction error is always greater than the initial prediction error after the adjustment coefficient iteration is completed, the corresponding adjustment coefficient is assigned a value of 1.

[0061] Step 4: If the prediction error of the training set after coefficient adjustment is greater than the preset error threshold, continue to introduce dynamic adjustment coefficients into the training set after coefficient adjustment and iterate until the prediction error index of the training set after introducing adjustment coefficients and dynamic adjustment coefficients is less than the preset error threshold.

[0062] The preset error threshold is set to 0.9 times the initial prediction error. ,like > Then the assigned coefficient needs to be dynamically adjusted;

[0063] Step 5: Retrain the frequency modulation command prediction model using the training set after introducing adjustment coefficients and dynamic adjustment coefficients to obtain the trained frequency modulation command prediction model, and output the frequency modulation control command through the trained frequency modulation command prediction model.

[0064] Specifically, a dynamic adjustment coefficient is assigned before each coefficient. The new training set is obtained as [ , , , ,..., ];

[0065] Next, we will discuss the dynamic adjustment coefficient. The iteration is performed, and the specific iteration formula is as follows:

[0066]

[0067] In the formula, This represents the dynamic adjustment coefficient; T represents the iteration number, initially set to 1; N represents the number of data points. For adjustment coefficients; The prediction error metric is gelu(); gelu() represents the activation function. 0.9MAPE Indicates the preset error threshold;

[0068] Predictions are made using the training set after introducing adjustment coefficients and dynamic adjustment coefficients, and compared with the actual values ​​on the original validation set to obtain the corresponding prediction error index. The prediction error index of the training set after introducing adjustment coefficients and dynamic adjustment coefficients is then compared with a preset error threshold. The comparison is performed, and the coefficients are iteratively and dynamically adjusted until the prediction error index of the training set after introducing the adjustment coefficients and the dynamic adjustment coefficients is less than the preset error threshold. .

[0069] In one embodiment of the present invention, a supercapacitor energy storage frequency modulation control method is provided, the purpose of which is to predict a future frequency modulation sequence of length 0.2N based on a frequency modulation command sequence of length N.

[0070] Specifically, the frequency modulation command sequence is divided into training and validation sets in a 9:1 ratio;

[0071] Let the frequency modulation command be P t =[P1,P2,P3,P4,...,P N The preceding sequence of length 0.9N is used to train [P1, P2, P3, P4, ..., P]. 0.9N A GRU neural network is used to predict values ​​of length 0.1N (the actual value is known). 0.9N+1 ,P 0.9N+2 ,P 0.9N+3 ,P 0.9N+4 ,...,P N The error between the predicted and actual values ​​at this point is calculated to be... At this point, an adjustment coefficient is assigned to P1. Adjustment coefficient The calculation formula is:

[0072] In the formula, It is located in and Random numbers between P1, P2, P3, P4, ..., P N For frequency modulation, tanh() represents the hyperbolic tangent function.

[0073] Put [W1P1,P2,P3,P4,...,P] into the list of elements. 0.9N Substituting this into a GRU neural network to predict the value of the next 0.1N length, the error between the predicted and actual values ​​is... ,like Greater than Then iterate W1 again until... Less than Until, if exhausted Still can't find MAPE2 less than Then let =1;

[0074] Next, we assign a coefficient to P2. :

[0075]

[0076] In the formula, It is a random number between these two values, P1, P2, P3, P4, ..., P N For frequency modulation, tanh() represents the hyperbolic tangent function.

[0077] Put [P1, W2P2, P3, P4, ..., P 0.9N Substituting this into a GRU neural network to predict the value of the next 0.1N length, the error between the predicted and actual values ​​is... ,like Greater than Then re-randomize ,until Less than If, after exhausting W2, it is still not found Less than Then let =1.

[0078] Following the above logic until it is generated until;

[0079] At this point, [ , , , ,..., Substituting the values ​​into a GRU neural network to predict the subsequent 0.1N values, the error between the predicted and actual values ​​is obtained as MAPE. N ;

[0080] If MAPE N No greater than 0.9 Then directly [ , , , ,..., The data is fed into a GRU neural network for prediction, yielding the final prediction result.

[0081] If MAPE N Greater than 0.9 Then, the adjustment coefficient for the assigned value needs to be dynamically adjusted. The specific adjustment method is as follows:

[0082] Assign a dynamic adjustment coefficient before each coefficient. It then became [ , , , ,..., ]

[0083] The iterative formula for dynamically adjusting the coefficients is:

[0084]

[0085] In the formula, This represents the dynamic adjustment coefficient; T represents the iteration number, initially set to 1; N represents the number of data points. For adjustment coefficients; is the prediction error index; gelu() represents the activation function; 0.9MAPE represents the preset error threshold; and i represents the data point location.

[0086] After correction, if MAPE N Greater than 0.9 Then the assigned dynamic coefficients need to be dynamically adjusted a second time, and this process is repeated until MAPE is achieved. N Less than 0.9 .

[0087] Finally, put the final sequence [ , , , ,..., , ,..., The data is fed into a GRU for prediction, and the final prediction result is obtained.

[0088] This invention introduces a dynamic adjustment mechanism to ensure that the model adapts to changes in frequency regulation commands, optimizes the response efficiency of the energy storage system to grid frequency regulation, and improves the overall frequency regulation performance.

[0089] In another embodiment of the present invention, a supercapacitor energy storage frequency regulation control system is provided, comprising a partitioning module, a calculation module, a first iteration module, a second iteration module, and a prediction output module, wherein:

[0090] Partitioning module: Used to divide frequency modulation commands into training and validation sets;

[0091] The calculation module is used to input the training set into the frequency modulation command prediction model to obtain the frequency modulation command prediction value, and to obtain the initial prediction error based on the frequency modulation command prediction value and the actual value in the validation set.

[0092] The first iteration module is used to sequentially assign adjustment coefficients to the frequency modulation instructions in the training set, iterate the adjustment coefficient values ​​until the corresponding prediction error is less than the initial prediction error, and obtain the training set after coefficient adjustment and the corresponding prediction error index.

[0093] Second iteration module: If the prediction error of the training set after coefficient adjustment is greater than the preset error threshold, the dynamic adjustment coefficient is introduced into the training set after coefficient adjustment and iterated until the prediction error index of the training set after the adjustment coefficient and the dynamic adjustment coefficient are less than the preset error threshold.

[0094] Prediction output module: It is used to retrain the frequency modulation command prediction model using the training set after introducing adjustment coefficients and dynamic adjustment coefficients to obtain a trained frequency modulation command prediction model, and output frequency modulation control commands through the trained frequency modulation command prediction model.

[0095] The partitioning module divides the frequency modulation command into training and validation sets at a ratio of 9:1; the calculation module uses the mean absolute percentage error (MASE) for the initial prediction error. The MASE is used to reflect the relative deviation of the prediction result by calculating the average absolute percentage error between the predicted value and the actual value.

[0096] The specific steps of the iteration module to assign adjustment coefficients to the frequency modulation instructions in the training set in sequence, and to iterate the adjustment coefficient values ​​until the corresponding prediction error is less than the initial prediction error, are as follows:

[0097] First, an adjustment coefficient is assigned to the first frequency modulation command in the training set. Then, adjustment coefficients are assigned sequentially until the last frequency modulation command in the training set. The adjustment coefficient values ​​are iterated until the corresponding prediction error is less than the initial prediction error.

[0098] When an adjustment factor is assigned to a frequency modulation command, the adjustment factors of all frequency modulation commands preceding that command are cleared.

[0099] All frequency modulation commands in the training set are assigned adjustment coefficients, resulting in the coefficient-adjusted training set and the corresponding prediction error, i.e., the prediction error index. .

[0100] When assigning an adjustment coefficient to a frequency modulation command, the adjustment coefficients of all previous frequency modulation commands are cleared. The coefficients of a single frequency modulation command are adjusted systematically to avoid interference between coefficients, accurately locate the optimal adjustment coefficient for each command, and accelerate iterative convergence.

[0101] When the corresponding prediction error is less than the initial prediction error, the iteration stops and the adjustment coefficient value is determined. If the prediction error is always greater than the initial prediction error after the adjustment coefficient iteration is completed, the corresponding adjustment coefficient is assigned a value of 1.

[0102] If the prediction error of the training set after one iteration is still greater than the preset error threshold, the dynamic adjustment coefficient of the second iteration module is introduced to continue the iteration until the prediction error index is less than the preset error threshold.

[0103] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0104] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations 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, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0105] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0106] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0107] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit its scope of protection. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading the present invention, they can still make various changes, modifications or equivalent substitutions to the specific implementation 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 frequency modulation control of supercapacitor energy storage, characterized in that, Includes the following steps: The frequency modulation commands are divided into a training set and a validation set; The training set is input into the frequency modulation command prediction model to obtain the frequency modulation command prediction value. The initial prediction error is obtained based on the frequency modulation command prediction value and the actual value in the validation set. The frequency modulation commands in the training set are assigned adjustment coefficients in sequence, and the coefficient values ​​are iteratively adjusted until the corresponding prediction error is less than the initial prediction error. At the same time, the training set after coefficient adjustment and the corresponding prediction error index are obtained. If the prediction error index of the training set after coefficient adjustment is greater than the preset error threshold, the dynamic adjustment coefficient is introduced into the training set after coefficient adjustment and iterated until the prediction error index of the training set after the adjustment coefficient and dynamic adjustment coefficient are less than the preset error threshold. The frequency modulation command prediction model is retrained using the training set after introducing adjustment coefficients and dynamic adjustment coefficients to obtain a trained frequency modulation command prediction model. The frequency modulation control command is then output through the trained frequency modulation command prediction model. The steps for sequentially assigning adjustment coefficients to the frequency modulation commands in the training set, iteratively adjusting the coefficient values ​​until the corresponding prediction error is less than the initial prediction error, and simultaneously obtaining the coefficient-adjusted training set and the corresponding prediction error are as follows: First, an adjustment coefficient is assigned to the first frequency modulation command in the training set, while the other frequency modulation commands remain unchanged, thus obtaining the first training set. The value of the adjustment coefficient is continuously adjusted until the prediction error corresponding to the first training set is less than the initial error. Clear the adjustment coefficient of the first frequency modulation command, assign the adjustment coefficient to the second frequency modulation command in the training set, and leave the other frequency modulation commands unchanged to obtain the second training set. Continuously adjust the value of the adjustment coefficient until the prediction error corresponding to the second training set is less than the initial error. The adjustment coefficients are sequentially assigned to the last frequency modulation command in the training set to obtain multiple training sets. The prediction error of all training sets is less than the initial prediction error, and the frequency modulation coefficients corresponding to the frequency modulation commands in the training sets are obtained. All adjustment coefficients are assigned to the frequency modulation command in the training set to obtain the coefficient-adjusted training set, and the corresponding prediction error is obtained based on the coefficient-adjusted training set. If the prediction error is always greater than the initial prediction error after the adjustment coefficient has been iterated, then the corresponding adjustment coefficient is assigned a value of 1. The iterative formula for the dynamic adjustment coefficient is: In the formula, This represents the dynamic adjustment coefficient; T represents the iteration number, initially set to 1; N represents the number of data points. For adjustment coefficients; is the prediction error index; gelu() represents the activation function; 0.9MAPE represents the preset error threshold; and i represents the position of the data point.

2. The method for frequency modulation control of supercapacitor energy storage according to claim 1, characterized in that, The initial prediction error is expressed as mean absolute percentage error.

3. The method for frequency modulation control of supercapacitor energy storage according to claim 2, characterized in that, The formula for calculating the mean absolute percentage error is: in, This is the actual value. is the predicted value, n is the number of data points, and MAPE is the mean absolute percentage error.

4. The method for frequency modulation control of supercapacitor energy storage according to claim 1, characterized in that, The preset error threshold is selected as 0.9 times the initial prediction error.

5. A supercapacitor energy storage frequency regulation control system, characterized in that, A supercapacitor energy storage frequency modulation control method according to any one of claims 1-4 includes a partitioning module, a calculation module, a first iteration module, a second iteration module, and a prediction output module, wherein: Partitioning module: Used to divide frequency modulation commands into training and validation sets; The calculation module is used to input the training set into the frequency modulation command prediction model to obtain the frequency modulation command prediction value, and to obtain the initial prediction error based on the frequency modulation command prediction value and the actual value in the validation set. The first iteration module is used to sequentially assign adjustment coefficients to the frequency modulation instructions in the training set, iterate the adjustment coefficient values ​​until the corresponding prediction error is less than the initial prediction error, and obtain the training set after coefficient adjustment and the corresponding prediction error index. Second iteration module: If the prediction error index corresponding to the training set after coefficient adjustment is greater than the preset error threshold, the dynamic adjustment coefficient is introduced into the training set after coefficient adjustment and iterated until the prediction error index corresponding to the training set after the adjustment coefficient and dynamic adjustment coefficient are less than the preset error threshold. Prediction output module: used to retrain the frequency modulation command prediction model using the training set after introducing adjustment coefficients and dynamic adjustment coefficients to obtain a trained frequency modulation command prediction model, and output frequency modulation control commands through the trained frequency modulation command prediction model; The steps for sequentially assigning adjustment coefficients to the frequency modulation commands in the training set, iteratively adjusting the coefficient values ​​until the corresponding prediction error is less than the initial prediction error, and simultaneously obtaining the coefficient-adjusted training set and the corresponding prediction error are as follows: First, an adjustment coefficient is assigned to the first frequency modulation command in the training set, while the other frequency modulation commands remain unchanged, thus obtaining the first training set. The value of the adjustment coefficient is continuously adjusted until the prediction error corresponding to the first training set is less than the initial error. Clear the adjustment coefficient of the first frequency modulation command, assign the adjustment coefficient to the second frequency modulation command in the training set, and leave the other frequency modulation commands unchanged to obtain the second training set. Continuously adjust the value of the adjustment coefficient until the prediction error corresponding to the second training set is less than the initial error. The adjustment coefficients are sequentially assigned to the last frequency modulation command in the training set to obtain multiple training sets. The prediction error of all training sets is less than the initial prediction error, and the frequency modulation coefficients corresponding to the frequency modulation commands in the training sets are obtained. All adjustment coefficients are assigned to the frequency modulation command in the training set to obtain the coefficient-adjusted training set, and the corresponding prediction error is obtained based on the coefficient-adjusted training set. If the prediction error is always greater than the initial prediction error after the adjustment coefficient has been iterated, then the corresponding adjustment coefficient is assigned a value of 1. The iterative formula for the dynamic adjustment coefficient is: In the formula, This represents the dynamic adjustment coefficient; T represents the iteration number, initially set to 1; N represents the number of data points. For adjustment coefficients; is the prediction error index; gelu() represents the activation function; 0.9MAPE represents the preset error threshold; and i represents the position of the data point.

6. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-4.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-4.

Images