Energy storage frequency modulation control and prediction method for super capacitor coupling lithium battery

Abstract

The invention relates to the technical field of energy storage frequency modulation detection, in particular to an energy storage frequency modulation control and prediction method for coupling a super capacitor with a lithium battery, which comprises the following steps of: realizing efficient coupling control of the super capacitor and the lithium battery through an intelligent switching mechanism; by reducing signal transmission and battery response time, time difference is reduced, and response speed is increased; dynamically adjusting control parameters according to the real-time frequency modulation instruction and the battery state; through a cooperative control strategy, the super capacitor and the lithium battery complement each other in frequency modulation tasks under different frequencies. The method has the beneficial effects that the frequency modulation response speed and prediction precision are improved, the response time difference is reduced, and the power grid stability is enhanced. Meanwhile, the service life of the battery is prolonged through optimized energy distribution and charge and discharge management, the operation cost is reduced, and the economic benefits of a power plant are improved. In addition, the adaptability of the system to a complex power grid environment is enhanced, and a more reliable and efficient frequency modulation solution is provided for a modern power system.

Description

Technical Field

[0001] This invention relates to the field of energy storage frequency regulation technology, and in particular to a method for energy storage frequency regulation control and prediction of supercapacitor-coupled lithium batteries. Background Technology

[0002] In power systems, frequency regulation of thermal power units is a critical control process, ensuring stable grid operation and high-quality power supply. Traditional frequency regulation methods rely on the response capabilities of thermal power units, but this approach has several significant limitations. These limitations become more pronounced with the integration of new energy sources and increasing grid demand. Traditional hybrid energy storage systems, combining supercapacitors and lithium batteries, are used to assist thermal power units in frequency regulation. This system transmits the frequency regulation command and the output difference of the thermal power unit to the hybrid energy storage, where the low-frequency portion is handled by the battery and the high-frequency portion by the supercapacitor. However, this method suffers from several major technical drawbacks: response time delay, insufficient prediction accuracy, low system efficiency, limited economic benefits, shortened battery life, and poor environmental adaptability.

[0003] These problems indicate that existing hybrid energy storage-assisted frequency regulation systems have significant shortcomings in terms of response speed, prediction accuracy, system efficiency, economic benefits, equipment lifespan, and environmental adaptability. New technical solutions are needed to overcome these challenges in order to meet the higher requirements of modern power grids for frequency regulation. Summary of the Invention

[0004] Therefore, the technical problem to be solved by this invention is to address the issue that existing hybrid energy storage-assisted thermal power unit frequency regulation systems suffer from low grid frequency regulation efficiency and limited economic benefits due to problems such as response delay and inaccurate prediction.

[0005] The above-mentioned technical problems are solved by the following technical solution: The present invention proposes an energy storage frequency regulation control method for supercapacitor coupled lithium battery, which includes realizing efficient coupling control of supercapacitor and lithium battery through intelligent switching mechanism;

[0006] By reducing signal transmission and battery response time, the time difference is reduced, and the response speed is improved;

[0007] The control parameters are dynamically adjusted based on real-time frequency modulation commands and battery status.

[0008] Through a collaborative control strategy, supercapacitors and lithium batteries can complement each other in frequency modulation tasks at different frequencies.

[0009] In a preferred embodiment of the energy storage frequency regulation control method for supercapacitor-coupled lithium batteries described in this invention: the intelligent switching mechanism includes intelligently allocating tasks according to the characteristics of the frequency regulation command, with low-frequency frequency regulation commands being responded to by the lithium battery and high-frequency frequency regulation commands being responded to by the supercapacitor.

[0010] In a preferred embodiment of the energy storage frequency regulation control method for supercapacitor-coupled lithium batteries described in this invention: reducing signal transmission and battery response time includes optimizing the signal transmission path and improving the response mechanism of the battery system.

[0011] In a preferred embodiment of the energy storage frequency regulation control method for supercapacitor-coupled lithium batteries described in this invention: the battery status includes battery charge and battery health status.

[0012] In a preferred embodiment of the energy storage frequency regulation control method for supercapacitor-coupled lithium battery described in this invention: the cooperative control strategy refers to improving the system's response speed and efficiency to grid frequency regulation by coordinating the operation of the supercapacitor and the lithium battery during the energy storage frequency regulation control process.

[0013] To solve the above-mentioned technical problems, the present invention also provides the following technical solution: a method for predicting the energy storage frequency modulation of a supercapacitor-coupled lithium battery, which includes segmenting the original frequency modulation sequence and making individual predictions to improve the overall prediction flexibility.

[0014] The parameters of the GRU network are dynamically optimized to adapt to the characteristics of different segment sequences;

[0015] Based on the prediction error of the preceding subsequence, the network parameters of the subsequent sequence are dynamically adjusted to gradually reduce the prediction error.

[0016] For each subsequence, perform iterative prediction and optimize the network parameters until the lowest prediction error is obtained;

[0017] Through multiple rounds of local optimization, the globally optimal GRU network parameters are finally determined and used for the prediction of the entire frequency modulation sequence.

[0018] Post-processing is performed on the prediction results of the GRU network to improve the prediction accuracy of unknown group values.

[0019] In a preferred embodiment of the energy storage frequency modulation prediction method for supercapacitor-coupled lithium batteries described in this invention: the parameters of the GRU network include the number of hidden layer neurons, the learning rate, and the regularization coefficient.

[0020] In a preferred embodiment of the energy storage frequency regulation prediction method for supercapacitor-coupled lithium batteries described in this invention: the network parameters of subsequent sequences are dynamically adjusted based on the prediction error of the preceding subsequence, and the model is continuously optimized by utilizing a feedback mechanism to make the prediction results closer to the actual values.

[0021] In a preferred embodiment of the energy storage frequency regulation prediction method for supercapacitor-coupled lithium batteries described in this invention: iterative prediction refers to finding the optimal network parameters through continuous iteration and optimization, and achieving the lowest prediction error of the optimal network parameters through continuous iteration and optimization of the network parameters.

[0022] In a preferred embodiment of the energy storage frequency modulation prediction method for supercapacitor-coupled lithium batteries described in this invention: post-processing includes correcting overfitting to improve the prediction accuracy of unknown values. The post-processing method is used to improve the generalization ability of the model and reduce prediction errors.

[0023] The beneficial effects of this invention are as follows: This invention improves frequency regulation response speed and prediction accuracy, reduces response time lag, and enhances grid stability. Simultaneously, optimized energy distribution and charge / discharge management extend battery life, reduce operating costs, and improve the economic efficiency of power plants. Furthermore, the system's enhanced adaptability to complex grid environments provides a more reliable and efficient frequency regulation solution for modern power systems. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Obviously, the drawings described below only relate to some embodiments of the present invention and are not intended to limit the present invention.

[0025] Figure 1 A schematic diagram of the GRU neural network unit structure for the energy storage frequency modulation control method of a supercapacitor-coupled lithium battery is shown. Detailed Implementation

[0026] To enable those skilled in the art to better understand the present invention, the present invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0027] The terminology used in this invention is that which is currently widely used in the art in consideration of the function of the invention; however, these terms may vary according to the intent of those skilled in the art, precedent, or new technology in the art. Furthermore, specific terms may be chosen by the applicant, and in such cases, their detailed meanings will be described in the detailed description of the invention. Therefore, the terms used in this specification should not be construed as simple names, but rather based on their meanings and the overall description of the invention.

[0028] As an optional embodiment, this embodiment provides an energy storage frequency regulation control method for a supercapacitor coupled to a lithium battery, including achieving efficient coupling control between the supercapacitor and the lithium battery through an intelligent switching mechanism;

[0029] By reducing signal transmission and battery response time, the time difference is reduced, and the response speed is improved;

[0030] The control parameters are dynamically adjusted based on real-time frequency modulation commands and battery status.

[0031] Through a collaborative control strategy, supercapacitors and lithium batteries can complement each other in frequency modulation tasks at different frequencies.

[0032] Furthermore, the intelligent switching mechanism includes intelligent task allocation based on the characteristics of the frequency modulation command: low-frequency frequency modulation commands are responded to by the lithium battery, while high-frequency frequency modulation commands are responded to by the supercapacitor.

[0033] Furthermore, reducing signal transmission and battery response time includes optimizing signal transmission paths and improving the battery system's response mechanisms.

[0034] Furthermore, battery status includes battery charge level and battery health status.

[0035] Furthermore, the coordinated control strategy refers to improving the system's response speed and efficiency to grid frequency regulation by coordinating the operation of supercapacitors and lithium batteries during the energy storage frequency regulation control process.

[0036] It should be noted that the collaborative control strategy includes core elements and workflow;

[0037] The core elements of a collaborative control strategy include task allocation, real-time monitoring and dynamic adjustment, communication and information sharing, prediction and pre-response, optimization of charge and discharge cycles, and fault response and redundancy.

[0038] Task allocation involves assigning frequency modulation tasks to the most suitable energy storage device based on the characteristics of supercapacitors and lithium batteries. Supercapacitors, due to their rapid charge and discharge characteristics, are suitable for responding to high-frequency changes; while lithium batteries are better suited for handling low-frequency changes, and due to their high energy density, they are suitable for long-term discharge.

[0039] Real-time monitoring and dynamic adjustment means real-time monitoring of grid frequency changes and the status of energy storage devices, and dynamic adjustment of task allocation and charging and discharging strategies to adapt to changes in grid demand.

[0040] Communication and information sharing, namely ensuring an efficient communication mechanism between supercapacitors and lithium batteries to achieve real-time information sharing, is crucial for collaborative control.

[0041] Prediction and pre-response, that is, using prediction algorithms to predict the changing trend of the power grid frequency in advance, so that supercapacitors and lithium batteries can make pre-response in advance and reduce response delay;

[0042] Optimize charge-discharge cycles, that is, through coordinated control, optimize the charge-discharge cycles of supercapacitors and lithium batteries to extend battery life and improve energy utilization efficiency;

[0043] Fault response and redundancy mean that when one energy storage device fails, another can quickly take over its tasks to ensure the continuity and reliability of the system.

[0044] The workflow of the collaborative control strategy includes data acquisition, task allocation, execution and feedback, adjustment and optimization, prediction and pre-response, as well as system maintenance and updates;

[0045] Among them, data acquisition refers to collecting status data of power grid frequency, supercapacitors and lithium batteries, including power, health status, temperature, etc.

[0046] Task allocation, that is, determining which tasks will be performed by supercapacitors and which by lithium batteries based on changes in grid frequency and the status of energy storage devices;

[0047] Execution and feedback, that is, to execute task assignments and collect feedback on the execution results to evaluate the effectiveness of task execution;

[0048] Adjustment and optimization refer to dynamically adjusting task allocation and control parameters based on feedback results to optimize system performance;

[0049] Forecasting and pre-response involves using historical data and current trends to predict future changes in power grid frequency and make adjustments in advance.

[0050] System maintenance and updates involve regularly inspecting and maintaining energy storage devices, and updating control strategies and algorithms to adapt to new grid conditions and technological advancements.

[0051] Specifically, a thermal power plant uses this scheme for auxiliary frequency regulation. The following is a detailed application example of the coordinated control strategy in frequency regulation of a thermal power plant:

[0052] First, task allocation is performed: When grid frequency fluctuates, the control system of the thermal power plant first analyzes the fluctuation characteristics, assigning high-frequency fluctuations to the supercapacitor for response and low-frequency fluctuations to the lithium battery for processing. For example, if the grid frequency fluctuation is above 1Hz, the supercapacitor quickly releases energy to offset the fluctuation; if the fluctuation is below 0.1Hz, the lithium battery takes on the regulation task.

[0053] Furthermore, real-time monitoring and dynamic adjustments are implemented: the control system monitors the power output of the thermal power plant and the status of the energy storage system in real time. When the power output of the thermal power plant suddenly drops while the lithium battery is charging, the control system will adjust the supercapacitor to participate in frequency regulation and accelerate the charging speed of the lithium battery to ensure that the frequency regulation task is not affected.

[0054] Furthermore, communication and information sharing are implemented: the energy storage system within the thermal power plant achieves real-time data sharing via industrial Ethernet. When the supercapacitor detects a sudden change in the grid frequency, the industrial Ethernet immediately transmits the information to the lithium battery and thermal power unit control systems, enabling the entire thermal power plant to quickly and collaboratively respond to grid frequency changes.

[0055] Furthermore, prediction and pre-response are implemented: using advanced prediction algorithms, the thermal power plant control system predicts upcoming peak grid loads. The control system instructs the lithium batteries to begin charging in advance and puts the lithium batteries into standby mode before the predicted peak arrives, so as to quickly release energy and reduce frequency regulation delays.

[0056] Furthermore, the charge-discharge cycle is optimized: the power plant's control system optimizes the energy storage system's charge-discharge plan based on grid load changes and electricity price information. During nighttime hours when electricity prices are low, the control system instructs the energy storage system to charge; during peak hours, energy is released according to grid demand to support frequency regulation and reduce costs.

[0057] Furthermore, fault response and redundancy are implemented: if the supercapacitor or lithium battery fails, the control system will automatically adjust the output of the thermal power unit and start the backup energy storage equipment to ensure the continuity of frequency regulation and reduce the impact on grid stability.

[0058] In summary, this invention improves frequency regulation response speed and prediction accuracy, reduces response time lag, and enhances grid stability. Simultaneously, optimized energy distribution and charge / discharge management extend battery life, reduce operating costs, and improve the economic efficiency of power plants. Furthermore, the system's enhanced adaptability to complex grid environments provides a more reliable and efficient frequency regulation solution for modern power systems.

[0059] like Figure 1 As shown, as an optional embodiment, this embodiment provides a method for predicting the frequency modulation of supercapacitor-coupled lithium batteries, including segmenting the original frequency modulation sequence and predicting it separately to improve the overall prediction flexibility.

[0060] It should be noted that the original frequency modulation sequence is set as Pt = [1, 2, 3, ..., 100]. When it is divided into ten segments, the first segment is [1, 2, 3, ..., 10]; the second segment is [11, 12, 13, ..., 20]; the third segment is [21, 22, 23, ..., 30]...; and the tenth segment is [90, 91, ..., 100].

[0061] The parameters of the GRU network are dynamically optimized to adapt to the characteristics of different segment sequences;

[0062] Based on the prediction error of the preceding subsequence, the network parameters of the subsequent sequence are dynamically adjusted to gradually reduce the prediction error.

[0063] For each subsequence, perform iterative prediction and optimize the network parameters until the lowest prediction error is obtained;

[0064] Through multiple rounds of local optimization, the globally optimal GRU network parameters are finally determined and used for the prediction of the entire frequency modulation sequence.

[0065] Post-processing is performed on the prediction results of the GRU network to improve the prediction accuracy of unknown group values.

[0066] Furthermore, the parameters of the GRU network include the number of hidden layer neurons, the learning rate, and the regularization coefficient.

[0067] It should be noted that the GRU neural network model is a type of recurrent neural network designed to address the problems of gradient vanishing and gradient explosion in long-term memory and backpropagation. Compared to LSTM, it requires fewer training parameters and is more convenient for computation.

[0068] GRU has only two gates: the update gate and the reset gate. The update gate controls the extent to which the state information from the previous moment is incorporated into the current state, i.e., old information is retained. The reset gate determines how new input information is combined with previous information, controlling the degree to which past states affect the current state.

[0069] The GRU neural network model automatically learns the values ​​of gating parameters through training data, thereby dynamically adjusting the flow of information based on the input sequence data, and more effectively capturing long-term dependencies in the sequence, such as... Figure 1 As shown, the formula for the GRU neural network model is as follows:

[0070] r i =σ(w r ·[h t-1 ,w r ,x t ])

[0071] Z t =σ(w Z ·[h t-1 ,w Z ,x t ])

[0072]

[0073] Where, r i The output of the update gate determines the hidden state h from the previous time step. t-1 The value of the update gate ranges from 0 to 1, indicating how much information will be retained in the current hidden state. A larger value means more information will be retained.

[0074] Zt The output of the Reset Gate determines the hidden state h from the previous moment. t-1 The degree of influence in the current hidden state is also determined by the value of the reset gate, which is between 0 and 1. The larger the value, the greater the influence of the previous state on the current state.

[0075] It is a candidate hidden state, which is the hidden state at the current moment after considering the state at the previous moment and the current input, without being adjusted by the update gate;

[0076] σ is the sigmoid activation function, used to map input values ​​to between 0 and 1, and is suitable as a gating signal;

[0077] tanh is the hyperbolic tangent activation function, used to generate candidate hidden states, and its output value ranges from -1 to 1;

[0078] w r w Z w h These are weight matrices, corresponding to the calculation of the update gate, reset gate, and candidate hidden state, respectively. These weight matrices are multiplied by the input data and the hidden state at the previous time step to determine the gating signal and the candidate hidden state;

[0079] h t-1 It is the hidden state of the previous moment, containing information from time step t-1;

[0080] x t This is the input data at the current moment;

[0081] * represents element-wise multiplication, i.e., the Hadamard product, used to update the gate r. i Compared to the hidden state h in the previous moment t-1 Element-by-element multiplication.

[0082] It should be noted that the optimization of GRU network parameters includes,

[0083] The number of hidden layer neurons affects the degree of data fit. Too few neurons lead to increased training time and underfitting; too many neurons result in insufficient information to train all hidden layer neurons, leading to overfitting.

[0084] The initial learning rate affects the convergence of the model. If it is too small, the convergence will be slow and require a longer training time. If it is too large, it may cause the model performance to fluctuate and gradient explosion. Generally, a larger learning rate is chosen at the beginning of training and the learning rate is gradually reduced during the optimization process. This can help shorten the training time. At the same time, as training progresses, the optimal value can be reached with smaller steps, so that a more refined search can be performed at the end of training.

[0085] L2 regularization, also known as weight decay, allows you to specify a multiplier for the L2 regularizer of a network layer with learnable parameters. By adding an L2 norm penalty term to the model's loss function, it reduces the learned model parameters and is a common technique for combating overfitting. In standard machine learning models, the loss function L is typically used to measure the difference between model predictions and actual observations. L2 regularization modifies this standard loss function by adding the sum of squared weights, as shown in the following formula:

[0086]

[0087] Where L is the original loss function; λ is the regularization parameter, which controls the strength of the regularization term; the larger the value of λ, the greater the impact of regularization; w i 2 is the weight of the model; n is the total number of weights.

[0088] The mechanisms by which L2 regularization works include:

[0089] Weight reduction and L2 regularization encourage the model to learn smaller weight values, which helps reduce model complexity and thus reduces the risk of overfitting.

[0090] To improve generalization ability, L2 regularization helps to improve the model's ability to generalize on unseen data by limiting the size of the model weights;

[0091] Numerical stability: In numerical computation, smaller weight values ​​can improve the stability of the algorithm and reduce the risk of gradient explosion.

[0092] Furthermore, based on the prediction error of the preceding subsequence, the network parameters of the subsequent sequence are dynamically adjusted. By utilizing the feedback mechanism, the model is continuously optimized, making the prediction results closer to the actual values.

[0093] Furthermore, iterative prediction refers to finding the optimal network parameters through continuous iteration and optimization, and then optimizing the network parameters through continuous iteration to achieve the lowest prediction error of the optimal network parameters.

[0094] Furthermore, post-processing includes correcting overfitting to improve the prediction accuracy for unknown values. Post-processing methods are used to improve the model's generalization ability and reduce prediction errors.

[0095] In summary, this invention improves frequency regulation response speed and prediction accuracy, reduces response time lag, and enhances grid stability. Simultaneously, optimized energy distribution and charge / discharge management extend battery life, reduce operating costs, and improve the economic efficiency of power plants. Furthermore, the system's enhanced adaptability to complex grid environments provides a more reliable and efficient frequency regulation solution for modern power systems.

[0096] As an optional embodiment, this embodiment provides a method for predicting the frequency modulation of a supercapacitor-coupled lithium battery, including, to further verify the advantages of this solution, using the method of this invention to perform actual numerical calculations to predict the frequency modulation sequence, the specific content of which is as follows:

[0097] First, Pt1 is fed into the GRU network for prediction, obtaining the error W1 of the experimental group. Then, based on the error, the initial number of hidden neurons n is adjusted. o Initial learning rate L O Initial regularization coefficient C O The correction is made, and the specific formula is as follows:

[0098]

[0099] L1=L0+rand(-sigmoid(sin(W1)),|gelu(sin(W1))|)

[0100]

[0101] Where [W1e] represents the integer part of W1e;

[0102] rand means generating a random number within a specified range;

[0103] The sigmoid is an activation function used to map input values ​​to a range between 0 and 1, making it suitable as a gating signal.

[0104] gelu is the Gaussian Error Linear Unit activation function, a smooth nonlinear activation function used to increase the nonlinear expressive power of the model;

[0105] sin is a sine function used to introduce periodic changes, which may help the model capture periodic features in the data;

[0106] e is the base of the natural logarithm, approximately equal to 2.71828, used to calculate the exponent;

[0107] t is the number of iterations, used to control the decay of the regularization coefficient. As the number of iterations increases, the regularization coefficient gradually decreases, and its initial value is 1.

[0108] n1 is the updated number of hidden neurons, based on the initial number of hidden neurons;

[0109] n0 is used for adjustment, and the adjustment amount is determined by a random function. The values ​​range from [W1e], where w1 is the error W1 of the experimental group;

[0110] L1 is the updated initial learning rate, which is adjusted based on the initial learning rate L0. The adjustment amount is determined by a random function that takes a value between -sigmoid(sin(W1)) and |gelu(sin(W1))|.

[0111] C1 is the updated initial regularization coefficient, adjusted based on the initial regularization coefficient C0. The adjustment amount is C0·e -t / 100 Determine, where t is the number of loops, with an initial value of 1.

[0112] After all parameters of GRU are updated, Pt2 is fed into the GRU network for prediction to obtain the error W2 of the experimental group. If W2≥W1, n1L1 is re-randomized and C1 (t=t+1) is updated until W2<W1.

[0113] When W2 < W1, Pt3 is fed into the GRU network for prediction, and the error W3 of the experimental group is obtained. At this time, n1L1C1 is corrected according to the error.

[0114] Repeat the above steps until Ptn, and then obtain the final GRU parameters, nr LrCr.

[0115] Furthermore, the original frequency modulation sequence Pt is predicted using a GRU network with final parameters, and the prediction result for the unknown group is [X]. N+1 X N+2 X N+3 , ..., X N+n ].

[0116] Furthermore, the prediction results need to be corrected because the continuous adjustment of the neural network may lead to overfitting. Therefore, the prediction results need to be corrected, that is, for a certain value X in the unknown group. N+i The specific formula for correction is as follows:

[0117]

[0118] Among them, X` N+i This is the original predicted value, that is, the prediction result before the correction;

[0119] X N+i This is the corrected predicted value, where N represents the total number of known data and i is the index of the unknown data;

[0120] max(W1, W2, ..., Wn) is the maximum value of the weight W, where W1, W2, ..., Wn are the weights of a certain layer or a certain set of weights in the neural network;

[0121] avg(W1, W2, ..., Wn) is the average of the weights W, that is, the average of all weights.

[0122] To further verify the advantages of the present invention, the present invention uses the direct prediction method of GRU and the prediction method of the present invention to predict and calculate the frequency modulation sequence of the actual operation of a power plant. The results are shown in the table below. Frequency modulation sequence 1 is collected from the 24-hour instruction data of a power plant, and frequency modulation sequence 2 is collected from the instruction data of a power plant within one week.

[0123]

[0124] Table 1 shows the instruction data for a certain power plant under two time lengths. The calculation formulas and definitions of the four evaluation indicators are as follows:

[0125]

[0126] Table 2. Instruction Data Calculation Formulas and Definitions

[0127] Among them, the values ​​of all four evaluation indicators in this scheme are smaller than the results of the direct prediction by GRU, that is, the prediction error is smaller than the results of the direct prediction by GRU. This indicates that this scheme provides more accurate and reliable predictions, which helps to improve the operating efficiency and economic benefits of power plants.

[0128] In summary, this invention improves frequency regulation response speed and prediction accuracy, reduces response time lag, and enhances grid stability. Simultaneously, optimized energy distribution and charge / discharge management extend battery life, reduce operating costs, and improve the economic efficiency of power plants. Furthermore, the system's enhanced adaptability to complex grid environments provides a more reliable and efficient frequency regulation solution for modern power systems.

[0129] Finally, it should be noted that the methods and devices described in detail above are merely embodiments, and those skilled in the art can modify these embodiments in different ways as long as they do not depart from the scope of the present invention.

Claims

1. A method for energy storage frequency regulation control of a supercapacitor-coupled lithium battery, characterized in that: include, Through an intelligent switching mechanism, efficient coupling control of supercapacitors and lithium batteries is achieved; By reducing signal transmission and battery response time, the time difference is reduced, and the response speed is improved; The control parameters are dynamically adjusted based on real-time frequency modulation commands and battery status. Through a collaborative control strategy, supercapacitors and lithium batteries can complement each other in frequency modulation tasks at different frequencies.

2. The energy storage frequency regulation control method for a supercapacitor-coupled lithium battery as described in claim 1, characterized in that: The intelligent switching mechanism includes intelligently allocating tasks according to the characteristics of the frequency modulation command. Low-frequency frequency modulation commands are responded to by the lithium battery, while high-frequency frequency modulation commands are responded to by the supercapacitor.

3. The energy storage frequency regulation control method for a supercapacitor-coupled lithium battery as described in claim 2, characterized in that: The reduction in signal transmission and battery response time includes optimizing signal transmission paths and improving the battery system's response mechanism.

4. The energy storage frequency regulation control method for a supercapacitor-coupled lithium battery as described in claim 3, characterized in that: The battery status includes battery charge and battery health status.

5. The energy storage frequency regulation control method for a supercapacitor-coupled lithium battery as described in claim 4, characterized in that: The aforementioned coordinated control strategy refers to improving the system's response speed and efficiency to grid frequency regulation by coordinating the operation of supercapacitors and lithium batteries during the energy storage frequency regulation control process.

6. A method for predicting the frequency regulation of energy storage in a supercapacitor-coupled lithium battery, based on the above-mentioned method for controlling the frequency regulation of energy storage in a supercapacitor-coupled lithium battery, characterized in that: include, The original frequency modulation sequence is segmented and predicted separately to improve the flexibility of the overall prediction. The parameters of the GRU network are dynamically optimized to adapt to the characteristics of different segment sequences; Based on the prediction error of the preceding subsequence, the network parameters of the subsequent sequence are dynamically adjusted to gradually reduce the prediction error. For each subsequence, perform iterative prediction and optimize the network parameters until the lowest prediction error is obtained; Through multiple rounds of local optimization, the globally optimal GRU network parameters are finally determined and used for the prediction of the entire frequency modulation sequence. Post-processing is performed on the prediction results of the GRU network to improve the prediction accuracy of unknown group values.

7. The energy storage frequency regulation prediction method for supercapacitor-coupled lithium batteries as described in claim 6, characterized in that: The parameters of the GRU network include the number of hidden layer neurons, the learning rate, and the regularization coefficient.

8. The energy storage frequency regulation prediction method for supercapacitor-coupled lithium batteries as described in claim 7, characterized in that: The network parameters of subsequent sequences are dynamically adjusted based on the prediction error of the preceding subsequences. By utilizing a feedback mechanism, the model is continuously optimized to make the prediction results closer to the actual values.

9. The method for predicting the energy storage frequency regulation of a supercapacitor-coupled lithium battery as described in claim 8, characterized in that: The iterative prediction refers to finding the optimal network parameters through continuous iteration and optimization, and then optimizing the network parameters through continuous iteration to achieve the lowest prediction error of the optimal network parameters.

10. The energy storage frequency regulation prediction method for supercapacitor-coupled lithium batteries as described in claim 9, characterized in that: The post-processing includes correcting overfitting to improve the prediction accuracy of unknown values. The post-processing method is used to improve the generalization ability of the model and reduce prediction errors.

Images