Energy storage frequency modulation instruction prediction method and system based on space-time weaving coupling interpolation

Abstract

The present application relates to the technical field of energy storage equipment of power system, and particularly relates to a kind of energy storage frequency modulation instruction prediction method and system based on space-time weaving coupling interpolation.The method includes reconstructing the discrete original frequency modulation sequence into continuous phase signal by space-time weaving, effectively suppresses high-frequency noise and random fluctuation interference, and accurately identifies the mutation or chaotic behavior region in phase signal using chaos intensity index, to avoid prediction instability caused by step or peak.Based on phase signal and chaos intensity index, adaptive interpolation is performed on adjacent discrete sampling points of original frequency modulation sequence to generate a smooth sequence, and after the smooth sequence is input into a pre-trained GRU network for prediction, the preliminary prediction sequence is dynamically corrected by the derivative of the continuous phase signal.The present application can achieve high-precision and high-stability frequency modulation instruction prediction, support predictive scheduling optimization of energy storage system, and improve its frequency modulation efficiency, economic benefit and system stability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy storage devices in power systems, and in particular to an energy storage frequency modulation instruction prediction method and system based on space-time weaving coupling interpolation. BACKGROUND

[0002] In the power system, the energy storage system has become an ideal resource for providing frequency modulation auxiliary services due to its fast, accurate, and bidirectional power regulation capability, especially for AGC (Automatic Generation Control) instruction response. However, the traditional "passive following" response strategy has significant limitations, i.e., it is executed immediately after receiving the instruction, which cannot fully utilize the potential of energy storage. High-precision prediction of upcoming frequency modulation instructions is the key to improving the efficiency, economic benefits, and system stability of energy storage frequency modulation. Existing artificial intelligence prediction techniques, such as directly using RNN (Recurrent Neural Network), GRU (Gated Recurrent Unit), etc. to predict the original frequency modulation sequence, face the following main defects:

[0003] (1) High-frequency noise and random fluctuation interference are serious: The original frequency modulation sequence contains a large number of high-frequency, low-amplitude random fluctuations caused by measurement errors, communication noise, or local minor disturbances. These meaningless noise points will interfere with the model learning the real pattern, resulting in unnecessary minor fluctuations in the prediction results, making it difficult to capture the real, smooth trend of the dominant system frequency change;

[0004] (2) Non-uniform sampling and data sparsity: The actual frequency modulation instruction sampling frequency may be low or there may be data packet loss, timestamp problems, resulting in sparse data points and uneven time intervals. This makes the sequence discontinuous on the time axis, and the model is difficult to effectively capture the continuous dynamic characteristics of the instruction change, such as the change rate, acceleration, etc., and the prediction result may present unreasonable jumps;

[0005] (3) Mutation points (steps / peaks) cause model training difficulty and prediction instability: Instantaneous large steps or peaks (caused by large disturbances, unit switching, or control logic switching) in the original instruction sequence are a manifestation of strong non-stationarity. The model is easily over-sensitive to these rare events or produces unreasonable mutations when predicting, resulting in a decline in overall generalization ability and robustness, and poor stability of the prediction result;

[0006] (4) Masking real trends and periodicity: High-frequency noise and mutation points can mask the low-frequency trends and potential periodicity in the sequence that reflect the actual system operating state (such as load change inertia, unit response characteristics). The model needs more complex structure and more data to learn these key features, and the effect is difficult to guarantee;

[0007] (5) Increasing model complexity and computational burden: In order to cope with the above-mentioned defects, models such as deep neural networks often need to be designed to be more complex. Such as deeper, wider, the introduction of attention, etc., leading to training difficulties, time-consuming increase, online prediction delay increases, deployment difficulties in resource-limited or fast response scenarios.

[0008] Therefore, the prior art cannot provide high-precision and high-stability frequency modulation instruction prediction, which restricts the goal of state optimization, loss reduction, life extension and revenue improvement of the energy storage system through predictive scheduling. SUMMARY

[0009] The technical problem to be solved by the embodiments of the present application is to provide a kind of energy storage frequency modulation instruction prediction method and system based on space-time weaving coupling interpolation, to solve the problem of low prediction accuracy and poor stability caused by high-frequency noise, data sparsity and mutation point in directly predicting original frequency modulation sequence in prior art.

[0010] The present application discloses an energy storage frequency modulation instruction prediction method based on space-time weaving coupling interpolation, comprising:

[0011] Converting the original frequency modulation sequence collected discretely into a continuous phase signal based on space-time weaving;

[0012] Calculate the chaos intensity index of each phase signal, and identify the area where mutation or chaotic behavior occurs in the phase signal corresponding to the chaos intensity index;

[0013] According to the phase signal and chaos intensity index of the adjacent two discrete sampling points in the original frequency modulation sequence, interpolation is calculated, and the smooth sequence is obtained by interpolating between every two adjacent discrete sampling points in the original frequency modulation sequence;

[0014] The smooth sequence is input into the pre-trained GRU network for prediction to obtain a preliminary prediction sequence, and the derivative of the continuous phase signal is used to correct the preliminary prediction sequence to obtain a final prediction sequence.

[0015] Optionally, the original frequency modulation sequence collected discretely is converted into a continuous phase signal based on space-time weaving, comprising:

[0016] Discrete sampling of frequency modulation instruction obtains the original frequency modulation sequence, and the function expression of the original frequency modulation sequence is:

[0017]

[0018] In the formula, is the original frequency modulation sequence, is the frequency modulation instruction of the i-th discrete sampling point, and i is the serial number index of the discrete sampling point;

[0019] The original frequency modulation sequence is converted into continuous phase signals by establishing a space-time weaving function with continuous time points as independent variables, and the function expression of the space-time weaving function is:

[0020]

[0021] In the formula, is the converted phase signal, is a sine function, is a circular constant, is a time point, is the total number of discrete sampling, if indicates the main condition definition domain, otherwise indicates the definition domain that does not satisfy the main condition.

[0022] Optionally, the energy storage frequency modulation instruction prediction method further comprises converting the frequency modulation instruction into a phase signal at a special time point, comprising:

[0023] If the current time point is the same as the discrete sampling point, the frequency modulation instruction of the discrete sampling point is limited based on the space-time weaving function, and the function expression of the limit processing is:

[0024]

[0025] In the formula, is the limit function when the current time point t approaches the discrete sampling point i infinitely;

[0026] According to the result of the limit processing, the frequency modulation instruction of the discrete sampling point is determined as the phase signal corresponding to the current time point.

[0027] Optionally, the calculation obtains the chaos intensity index of each phase signal, comprising:

[0028] A chaos intensity detection function is constructed using the derivative of the continuous phase signal, and the chaos intensity index of each phase signal is calculated by the chaos intensity detection function, and the function expression of the chaos intensity detection function is:

[0029]

[0030] In the formula, is the chaos intensity index, is the third derivative of the phase signal, is an exponential function, is a damping coefficient, is the first derivative of the phase signal.

[0031] Optionally, the interpolation is calculated according to the phase signal and the chaos intensity index of the adjacent two discrete sampling points in the original frequency modulation sequence, comprising:

[0032] extracting a locally smooth phase signal from the continuous phase signal by slightly offsetting the center points of two adjacent discrete sampling points;

[0033] calculating a dynamic adjustment coefficient according to the chaotic intensity index corresponding to the discrete sampling points and a derivative thereof, a function expression of the dynamic adjustment coefficient calculation being:

[0034]

[0035] wherein, is the dynamic adjustment coefficient, is a hyperbolic tangent function, is a first derivative of the chaotic intensity index;

[0036] combining the smooth signal and the dynamic adjustment coefficient, calculating an interpolation between the discrete sampling point and the previous discrete sampling point, a function formula of the interpolation calculation being:

[0037]

[0038] wherein, is the interpolation between the discrete sampling point i and the previous discrete sampling point i-1, is a phase signal of a left offset coordinate of the interpolation interval center point, is a phase signal of a right offset coordinate of the interpolation interval center point.

[0039] Optionally, the extracting a locally smooth phase signal from the continuous phase signal by slightly offsetting the center points of two adjacent discrete sampling points comprises:

[0040] slightly offsetting the center points of two adjacent discrete sampling points to the left and calculating a coordinate of the left offset point on the time axis, a function expression of the left offset point time coordinate calculation being:

[0041]

[0042] slightly offsetting the center points of two adjacent discrete sampling points to the right and calculating a coordinate of the right offset point on the time axis, a function expression of the right offset point time coordinate calculation being:

[0043]

[0044] wherein, is a time point corresponding to the left offset of the interpolation interval center point, is a time point corresponding to the right offset of the interpolation interval center point;

[0045] The time coordinate of the left offset point and the time coordinate of the right offset point are substituted into the space-time weaving function to obtain a phase signal corresponding to the time coordinate.

[0046] Optionally, the preliminary prediction sequence is corrected using derivatives of the continuous phase signal to obtain a final prediction sequence, including:

[0047] The derivatives of the continuous phase signal are used to calculate a stability coefficient, and the function expression of the stability coefficient is:

[0048]

[0049] In the formula, is a stability coefficient, is a time scaling factor, is a time point of the point to be corrected;

[0050] The preliminary prediction sequence is corrected using the stability coefficient to obtain the final prediction sequence, and the function expression of the correction is:

[0051]

[0052] In the formula, is an FM command final prediction value of an N+i-th discrete sampling point, is an FM command preliminary prediction value of the N+i-th discrete sampling point.

[0053] The application further discloses a prediction system adopting the energy storage FM command prediction method based on space-time weaving coupling interpolation.

[0054] The signal conversion module is configured to convert the original FM sequence collected discretely into continuous phase signals based on space-time weaving.

[0055] The chaos intensity inspection module is configured to calculate the chaos intensity index of each phase signal and identify the area where mutation or chaos behavior occurs in the corresponding phase signal through the chaos intensity index.

[0056] The interpolation module is configured to calculate interpolation according to the phase signals and chaos intensity indexes of adjacent two discrete sampling points in the original FM sequence and obtain a smooth sequence by interpolating between every two adjacent discrete sampling points in the original FM sequence.

[0057] The prediction module is configured to input the smooth sequence into a pre-trained GRU network to obtain a preliminary prediction sequence and correct the preliminary prediction sequence using derivatives of the continuous phase signal to obtain a final prediction sequence.

[0058] The present invention also discloses a computer-readable storage medium storing a computer program thereon, characterized in that the computer program, when executed by a processor, implements the above-described method for predicting energy storage frequency modulation commands based on spatiotemporal weaving coupling interpolation.

[0059] The present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the above-mentioned energy storage frequency modulation command prediction method based on spatiotemporal weaving coupling interpolation.

[0060] Compared with the prior art, the beneficial effects of the energy storage frequency modulation command prediction method and system based on spatiotemporal braided coupling interpolation provided in this invention are as follows:

[0061] By spatiotemporally weaving, the discrete original frequency modulation sequence is reconstructed into a continuous phase signal, effectively suppressing high-frequency noise and random fluctuation interference. Furthermore, the chaos intensity index is used to accurately identify abrupt changes or chaotic behavior regions in the phase signal, avoiding prediction instability caused by steps or spikes. Based on the phase signal and the chaos intensity index, adaptive interpolation is performed on adjacent discrete sampling points of the original frequency modulation sequence to generate a smooth sequence, solving the problems of non-uniform sampling and data sparsity while realistically preserving the mid-to-low frequency trends and periodicity. After this smooth sequence is input into a pre-trained GRU network for prediction, the initial prediction sequence is dynamically corrected using the derivative of the continuous phase signal, significantly improving prediction accuracy and stability. Ultimately, this achieves high-precision, high-stability frequency modulation command prediction, supporting predictive scheduling optimization of energy storage systems and improving their frequency modulation efficiency, economic benefits, and system stability. Attached Figure Description

[0062] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. In the accompanying drawings:

[0063] Figure 1 A schematic diagram illustrating the steps of the energy storage frequency regulation command prediction method provided in an embodiment of the present invention. Detailed Implementation

[0064] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0065] This invention discloses a method for predicting energy storage frequency modulation commands based on spatiotemporal braided coupling interpolation, comprising:

[0066] S1. Based on spatiotemporal weaving, the discretely acquired original frequency modulation sequence is converted into a continuous phase signal;

[0067] S2. Calculate and obtain the chaos intensity index of each phase signal, and identify the region in the corresponding phase signal that has abrupt change or chaotic behavior through the chaos intensity index.

[0068] S3. The interpolation is calculated based on the phase signal and chaos intensity index of two adjacent discrete sampling points in the original frequency modulation sequence, and a smooth sequence is obtained by interpolating between two adjacent discrete sampling points in the original frequency modulation sequence.

[0069] S4. Input the smoothed sequence into the pre-trained GRU network to obtain the preliminary prediction sequence, and use the derivative of the continuous phase signal to correct the preliminary prediction sequence to obtain the final prediction sequence.

[0070] Through the implementation of the above-described energy storage frequency regulation command prediction method, the discretely acquired original frequency regulation sequence is reconstructed into a continuous phase signal based on spatiotemporal weaving. This effectively filters out high-frequency random fluctuations caused by measurement errors, communication noise, or local minor disturbances, eliminating their interference with the model's learning of the real mode. The chaos intensity index is used to accurately identify abrupt changes or chaotic behavior regions in the phase signal caused by large disturbances, unit switching, or control logic switching, avoiding the model's oversensitivity to instantaneous steps or spikes and significantly improving prediction robustness and generalization ability. Chaotic adaptive interpolation is performed based on the phase signals of adjacent discrete sampling points of the original frequency regulation sequence and the chaos intensity index. This fills in data gaps caused by packet loss or timestamp issues while strictly maintaining the continuity of the time axis, enabling the model to fully capture continuous dynamic characteristics such as command change rate and acceleration. This eliminates unreasonable jumps in prediction results and fully preserves the low-to-medium frequency trends and potential periodicity reflecting load change inertia and unit response characteristics. After inputting the generated smooth sequence into a pre-trained gated recurrent unit network to obtain an initial prediction sequence, dynamic correction is performed using the first and third derivatives of the continuous phase signal to accurately suppress oscillations in the prediction sequence. This achieves deep coupling between the prediction results and the real physical process at the levels of signal change rate and abrupt acceleration. This method improves signal quality through spatiotemporal weaving and interpolation mechanisms during data preprocessing, enhances abrupt robustness and trend integrity during feature extraction, reduces the structural complexity of the gated recurrent unit network and accelerates the convergence process during prediction, and ensures output stability by coupling signal dynamics during correction. Ultimately, it forms a high-precision, high-stability frequency regulation command prediction capability, providing reliable input for predictive optimization scheduling of energy storage systems, maximizing their potential for rapid and accurate bidirectional power regulation, and simultaneously achieving the core objectives of improved frequency regulation efficiency, reduced losses, extended lifespan, increased revenue, and enhanced power system stability.

[0071] Furthermore, based on spatiotemporal weaving, the discretely acquired original frequency-modulated sequences are converted into continuous phase signals, including:

[0072] Discrete acquisition frequency modulation commands yield the original frequency modulation sequence, and the functional expression of the original frequency modulation sequence is:

[0073]

[0074] In the formula, This is the original frequency modulation sequence. This is the frequency modulation command for the i-th discrete sampling point, where i is the index of the discrete sampling point;

[0075] Using continuous time points as independent variables, a spatiotemporal weaving function is established to convert the original frequency modulation sequence into a continuous phase signal. The functional expression of the spatiotemporal weaving function is as follows:

[0076]

[0077] In the formula, The converted phase signal, It is a sine function. Pi For a point in time, The total number of discrete samples is denoted by 'if', which represents the domain of the main condition, and 'otherwise' which represents the domain of the main condition that is not satisfied.

[0078] Furthermore, the energy storage frequency regulation command prediction method also includes converting the frequency regulation command into a phase signal at specific time points, including:

[0079] If the current time point is the same as the discrete sampling point, the frequency modulation command for that discrete sampling point is subjected to limit processing based on the spatiotemporal weaving function. The function expression for the limit processing is:

[0080]

[0081] In the formula, The limit function is the function that approaches the discrete sampling point i infinitely at the current time point t.

[0082] Based on the results of the limit processing, the frequency modulation command of the discrete sampling point is determined to be the phase signal corresponding to the current time point.

[0083] Through the implementation of the above-mentioned energy storage frequency modulation command prediction method embodiment, the sinusoidal function reconstruction mechanism is precisely defined, and the discrete original frequency modulation sequence is converted into a continuous phase signal with strict mathematical meaning, which can achieve the following effects: (1) Constructing physically interpretable phase continuity in the entire signal domain: When time point t satisfies 1 / 2≤t≤N+1 / 2, the spatiotemporal weaving function through The infinite differentiability of the function generates a phase signal with a smooth infinite derivative at non-discrete sampling points, thus solving the problem of discontinuity in the time axis of traditional discrete sequences. This represents the time offset between any consecutive time point t and the position of the i-th discrete sampling point. For the special case where the current time point t coincides with the discrete sampling point i, an extreme processing method is used to ensure the phase signal of that discrete sampling point. = The accurate restoration not only eliminates the reconstruction error that may exist at integer points in conventional interpolation, but also maintains the mathematical completeness of signal reconstruction; (2) Establish the dynamic basis of noise suppression and abrupt response: the reconstructed continuous phase signal The abrupt change information implied in its third derivative can provide a dimensionally consistent input source for subsequent calculations of chaotic intensity indices. The smooth transition characteristics of this continuous signal between discrete sampling points can automatically filter out minute fluctuations caused by high-frequency noise in the original frequency-modulated sequence, such as measurement errors or communication noise. Simultaneously, through… The strict band-limited characteristics of the function in the frequency domain preserve the low-to-medium frequency fundamental components that reflect the unit's response characteristics; (3) it lays the foundation for high-dimensional feature extraction of the prediction model: continuous phase signal The time-domain analytical expression directly supports derivative operations at arbitrary time granularity, making the signal change rate represented by its first derivative and the abrupt acceleration represented by its third derivative quantifiable physical quantities. This solves the bottleneck of non-uniform sampling causing the model to fail to capture continuous dynamic characteristics. This feature can provide physically meaningful feature inputs for subsequent adaptive interpolation based on chaos intensity indices and GRU network dynamics correction, avoiding unreasonable jumps in prediction results that violate the dynamic process of the power system.

[0084] Therefore, the method in this embodiment serves as the foundational layer of the energy storage frequency regulation command prediction method. Through a mathematically rigorous signal reconstruction mechanism, it constructs a continuous phase signal with sufficient smoothness, derivative quantization capability, and physical interpretability while preserving the numerical accuracy of the original discrete sampling points (based on limit processing). This provides fundamental technical support for eliminating high-frequency noise interference, extracting the true dynamic characteristics of the system, and suppressing the oscillation of the prediction sequence, ultimately serving the high-precision predictive scheduling of frequency regulation commands by the energy storage system.

[0085] Furthermore, the chaos intensity index for each phase signal is calculated, including:

[0086] A chaos intensity detection function is constructed using the derivatives of continuous phase signals, and the chaos intensity index for each phase signal is calculated using this function. The functional expression of the chaos intensity detection function is as follows:

[0087]

[0088] In the formula, As a measure of chaos intensity, The third derivative of the phase signal. It is an exponential function. The damping coefficient is... It is the first derivative of the phase signal.

[0089] Through the implementation of the above-described energy storage frequency regulation command prediction method embodiment, based on the third derivative of the phase signal... The absolute modulus is used to accurately quantify the acceleration of signal abrupt changes. When the power system encounters large disturbances, unit switching, or control logic switching, it directly captures the instantaneous dynamic response corresponding to the step or spike, providing a physically interpretable mathematical representation for identifying chaotic behavior regions. Secondly, the first derivative of the phase signal is introduced. The input is an exponential function, through the damping coefficient. The high-frequency noise suppression intensity is dynamically adjusted. When the signal is stable, it maintains the ability to distinguish subtle changes, while under high-dynamic conditions, it enhances noise filtering, completely eliminating meaningless fluctuations caused by measurement errors and communication noise. Furthermore, a chaos intensity index is formed using the product structure of the third and first derivatives. The adaptive coupling mechanism, while preserving the original signal's fundamental frequency band information, enables chaotic intensity detection to possess both abrupt change response sensitivity and stability under normal operating conditions, significantly improving the accuracy and robustness in identifying steps or spikes in the command sequence. Therefore, the method in this embodiment transforms the dynamic characteristics of the power system into quantifiable indicators through parameterized differential operations, eliminating non-stationary interference in the original frequency regulation sequence at the feature level. This provides a precise partitioning basis for subsequent chaotic adaptive interpolation, addressing the training divergence and prediction instability problems caused by abrupt changes in the model, while ensuring the integrity of the low-to-medium frequency trends reflecting load change inertia and unit response characteristics. This detection mechanism constitutes the core feature extraction layer at the front end of the prediction model, laying the dynamic foundation for high-precision, high-stability energy storage frequency regulation command prediction.

[0090] As mentioned above, the third derivative of the phase signal The "jump" of a phase change reflects the drastic nature of the change; the first derivative of the phase signal. That is, angular velocity; damping coefficient (A positive real number, usually taken as 0.2) is used to suppress the effects of high-frequency noise; a chaos intensity index. (Dimensionless), the larger its value, the more obvious the chaotic characteristics at that moment, and it is used to identify regions in the phase signal where abrupt changes or chaotic behavior occur.

[0091] Among them, the numerator of the chaos intensity detection function Used to detect phase abrupt changes (chaotic features), exponential part Used to suppress high-frequency noise: When the signal changes smoothly, the exponential term is close to 1, emphasizing the role of the third derivative; when the signal changes drastically, the exponential term decays, avoiding misinterpreting rapid but smooth changes as chaos.

[0092] Furthermore, the interpolation is calculated based on the phase signal and chaos intensity index of two adjacent discrete sampling points in the original frequency modulation sequence, including:

[0093] By slightly shifting the center point of two adjacent discrete sampling points, a locally smoothed phase signal can be extracted from a continuous phase signal.

[0094] The dynamic adjustment coefficient is calculated based on the chaos intensity index and its derivative corresponding to the discrete sampling points. The functional expression for calculating the dynamic adjustment coefficient is as follows:

[0095]

[0096] In the formula, For dynamic adjustment coefficients, It is the hyperbolic tangent function. The first derivative of the chaos intensity index;

[0097] Combining the smoothed signal and dynamic adjustment coefficients, the interpolation between discrete sampling points and the previous discrete sampling points is calculated. The interpolation calculation function is as follows:

[0098]

[0099] In the formula, This is the interpolation between discrete sampling point i and the previous discrete sampling point i-1. The phase signal is the coordinate shifted to the left of the center point of the interpolation interval. This is the phase signal shifted to the right by the coordinates of the center point of the interpolation interval.

[0100] Furthermore, by slightly shifting the center points of two adjacent discrete sampling points, a locally smoothed signal is extracted from the continuous phase signal, including:

[0101] The center points of two adjacent discrete sampling points are slightly shifted to the left, and the coordinates of the left shift point on the time axis are calculated. The function expression for calculating the time coordinates of the left shift point is as follows:

[0102]

[0103] The center points of two adjacent discrete sampling points are slightly shifted to the right, and the coordinates of the right shift point on the time axis are calculated. The function expression for calculating the time coordinates of the right shift point is as follows:

[0104]

[0105] In the formula, The time point corresponding to the leftward offset of the center point of the interpolation interval. This is the time point corresponding to the rightward offset of the center point of the interpolation interval;

[0106] Substitute the time coordinates of the left offset point and the right offset point into the spatiotemporal weaving function to obtain the phase signal of the corresponding time coordinate.

[0107] Through the implementation of the above-described energy storage frequency regulation command prediction method embodiment, the chaos intensity index is used. and its derivative Constructed dynamic adjustment coefficient Adaptive control for interpolation behavior: when the chaos intensity index When a large perturbation-induced mutation or chaotic behavior is detected, the hyperbolic tangent function is used to map the result... Approaching 2, the driving interpolation function enhances interpolation of discrete sampling points. The response intensity; under steady operating conditions, it makes... Approaching 0, it degenerates into local smoothing based on continuous phase signals. By precisely shifting the center points of adjacent discrete sampling points by ±0.02 units (i-1+i) / 2, the time coordinates of the left offset point t⁻=i-0.52 and the right offset point t⁺=i-0.48 are generated. This allows extraction from the continuous phase signal reconstructed by the spatiotemporal weaving function. and This serves as a noise immunity benchmark. Ultimately, a dual-mechanism coupling is achieved in the interpolation function. That is, the first term in the interpolation function is... and The mean calculation retains the low-to-medium frequency trend that reflects the unit's response characteristics; the second term is based on... Dynamically weighted discrete sampling point interpolation enhances the tracking ability for steps or spikes in chaotic regions. Therefore, the method in this embodiment simultaneously achieves four core effects: filling gaps in non-uniform sampling data, suppressing high-frequency noise interference, maintaining time axis continuity, and preserving system dynamic characteristics, providing high-quality input features with physical interpretability for pre-trained GRU networks.

[0108] Furthermore, the preliminary prediction sequence is corrected using the derivative of the continuous phase signal to obtain the final prediction sequence, including:

[0109] The stability coefficient is obtained by calculating the derivative of the continuous-phase signal. The functional expression for the stability coefficient is:

[0110]

[0111] In the formula, For stability coefficient, This is the time scaling factor. The time point to be corrected;

[0112] The final predicted sequence is obtained by correcting the initial predicted sequence using a stability coefficient. The corrected function expression is as follows:

[0113]

[0114] In the formula, This represents the final predicted value of the frequency modulation command at the (N+i)th discrete sampling point. This is the initial predicted value of the frequency modulation command for the (N+i)th discrete sampling point.

[0115] Through the implementation of the above-described energy storage frequency regulation command prediction method embodiment, an adaptive stability coefficient is constructed using the derivative of the continuous phase signal to achieve accurate prediction sequence correction. That is, it is based on the first derivative of the continuous phase signal at the time point t=N+i to be corrected. (Signal rate of change) and third derivative (Sudden acceleration), combined with the time scaling factor S to construct the stability coefficient This coefficient increases when the signal is stable to enhance the correction of the predicted sequence, and approaches zero under abrupt changes to maintain the abrupt change characteristics of the initial predicted sequence. This is achieved through a correction function. The stability coefficient is compared with the discrete sample values ​​in the initial prediction sequence. Dynamic coupling eliminates oscillations in the prediction sequence caused by high-frequency noise while preserving the accuracy of steps or spikes caused by large disturbances. This correction mechanism ensures that the final prediction sequence significantly improves the stability and physical consistency of energy storage frequency regulation command predictions while fully maintaining the ability of the original prediction results to capture the dynamic response characteristics of the system. This provides energy storage systems with highly reliable prediction inputs that are dynamically adaptable.

[0116] To further verify the advantages of this invention, the method of this embodiment and the direct prediction method using a GRU network were respectively used to predict frequency modulation sequences. Four evaluation metrics—MAE (Mean Absolute Error), SSE (Sum of Squares Error), RMSE (Root Mean Square Error), and MAPE (Mean Absolute Percentage Error)—were used to evaluate the prediction results. The selection of these evaluation metrics can be found in the evaluation metric table.

[0117] Table 1 Evaluation Indicators

[0118]

[0119] As shown in Table 1, among the four evaluation metrics, N represents the sample size. and These represent the actual value and the predicted value at time n, respectively.

[0120] The results of evaluating the above predictions using four evaluation metrics are shown in the FM sequence prediction table:

[0121] Table 2 Frequency Modulation Sequence Prediction Table

[0122]

[0123] As shown in Table 2, frequency modulation sequence 1 comes from frequency modulation data of a power plant in Hulunbuir from 0:00 to 20:00 on December 1, 2024; frequency modulation sequence 2 comes from frequency modulation data of a power plant in Hulunbuir from 2:00 to 22:00 on December 2, 2024. The evaluation results in Table 2 show that all four evaluation indicators have decreased, indicating that the decomposition method proposed in this invention can reduce the nonlinearity of the original sequence compared to the traditional GRU prediction method, further improving the prediction accuracy and effectively helping power plants improve their frequency modulation response capability, thereby further increasing the power plant's revenue.

[0124] This invention also discloses a prediction system that employs the aforementioned energy storage frequency modulation command prediction method based on spatiotemporal weaving coupling interpolation. The prediction system includes:

[0125] The signal conversion module is used to convert discretely acquired raw frequency modulation sequences into continuous phase signals based on spatiotemporal weaving.

[0126] The chaos intensity test module is used to calculate and obtain the chaos intensity index of each phase signal, and to identify the region in the corresponding phase signal that has abrupt change or chaotic behavior through the chaos intensity index.

[0127] The interpolation module is used to calculate the interpolation based on the phase signal and chaos intensity index of two adjacent discrete sampling points in the original frequency modulation sequence, and to obtain a smooth sequence by interpolating between two adjacent discrete sampling points in the original frequency modulation sequence.

[0128] The prediction module is used to input the smooth sequence into the pre-trained GRU network to obtain the preliminary prediction sequence, and then use the derivative of the continuous phase signal to correct the preliminary prediction sequence to obtain the final prediction sequence.

[0129] The present invention also discloses a computer-readable storage medium storing a computer program thereon, characterized in that the computer program, when executed by a processor, implements the above-described method for predicting energy storage frequency modulation commands based on spatiotemporal weaving coupling interpolation.

[0130] The present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the above-mentioned energy storage frequency modulation command prediction method based on spatiotemporal weaving coupling interpolation.

[0131] This invention is described based on flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to specific embodiments. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of 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, special-purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing device, generate instructions for implementing the flowcharts 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.

[0132] 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.

[0133] 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.

[0134] It should be understood that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Those skilled in the art can modify the technical solutions described in the above embodiments, or make equivalent substitutions for some of the technical features; and all such modifications and substitutions should fall within the protection scope of the present invention.

Claims

1. A method for predicting frequency modulation commands for energy storage based on spatiotemporal braided coupling interpolation, characterized in that, The energy storage frequency regulation command prediction method includes: Spatiotemporal weaving is used to convert discretely acquired original frequency modulation sequences into continuous phase signals; The chaos intensity index of each phase signal is calculated and obtained, and the region in the corresponding phase signal that has abrupt change or chaotic behavior is identified by the chaos intensity index. The interpolation is calculated based on the phase signal and the chaos intensity index of two adjacent discrete sampling points in the original frequency modulation sequence, and a smooth sequence is obtained by interpolating between two adjacent discrete sampling points in the original frequency modulation sequence. The smoothed sequence is input into a pre-trained GRU network to obtain a preliminary prediction sequence, and the derivative of the continuous phase signal is used to correct the preliminary prediction sequence to obtain the final prediction sequence.

2. The energy storage frequency modulation command prediction method based on spatiotemporal weaving coupling interpolation according to claim 1, characterized in that, The method of converting discretely acquired original frequency-modulated sequences into continuous phase signals based on spatiotemporal weaving includes: The original frequency modulation sequence is obtained by discrete acquisition frequency modulation command, and the functional expression of the original frequency modulation sequence is: In the formula, This is the original frequency modulation sequence. This is the frequency modulation command for the i-th discrete sampling point, where i is the index of the discrete sampling point; Using continuous time points as independent variables, a spatiotemporal weaving function is established to convert the original frequency modulation sequence into a continuous phase signal. The functional expression of the spatiotemporal weaving function is as follows: In the formula, The converted phase signal, It is a sine function. Pi For a point in time, The total number of discrete samples is denoted by 'if', which represents the domain of the main condition, and 'otherwise' which represents the domain of the main condition that is not satisfied.

3. The energy storage frequency modulation command prediction method based on spatiotemporal weaving coupling interpolation according to claim 2, characterized in that, The energy storage frequency regulation command prediction method further includes converting the frequency regulation command into a phase signal at a specific time point, including: If the current time point is the same as the discrete sampling point, the frequency modulation command of the discrete sampling point is subjected to limit processing based on the spatiotemporal weaving function. The function expression for the limit processing is: In the formula, The limit function is the function that approaches the discrete sampling point i infinitely at the current time point t. Based on the results of the limit processing, the frequency modulation command of the discrete sampling point is determined to be the phase signal corresponding to the current time point.

4. The energy storage frequency modulation command prediction method based on spatiotemporal weaving coupling interpolation according to claim 2, characterized in that, The calculation of the chaos intensity index for each phase signal includes: A chaos intensity detection function is constructed using the derivatives of the continuous phase signals, and a chaos intensity index for each phase signal is calculated using this function. The functional expression of the chaos intensity detection function is as follows: In the formula, As a measure of chaos intensity, The third derivative of the phase signal. It is an exponential function. The damping coefficient is... It is the first derivative of the phase signal.

5. The energy storage frequency modulation command prediction method based on spatiotemporal weaving coupling interpolation according to claim 4, characterized in that, The interpolation calculated based on the phase signal and chaos intensity index of two adjacent discrete sampling points in the original frequency modulation sequence includes: By slightly shifting the center points of two adjacent discrete sampling points, a locally smoothed phase signal is extracted from the continuous phase signal. The dynamic adjustment coefficient is calculated based on the chaos intensity index and its derivative corresponding to the discrete sampling points. The functional expression for calculating the dynamic adjustment coefficient is as follows: In the formula, For dynamic adjustment coefficients, It is the hyperbolic tangent function. The first derivative of the chaos intensity index; Combining the locally smoothed phase signal and the dynamic adjustment coefficient, the interpolation between the discrete sampling point and the previous discrete sampling point is calculated. The function formula for calculating the interpolation is: In the formula, This is the interpolation between discrete sampling point i and the previous discrete sampling point i-1. The phase signal is the coordinate shifted to the left of the center point of the interpolation interval. This is the phase signal shifted to the right by the coordinates of the center point of the interpolation interval.

6. The energy storage frequency modulation command prediction method based on spatiotemporal weaving coupling interpolation according to claim 5, characterized in that, The step of extracting a locally smoothed phase signal from a continuous phase signal by slightly shifting the center points of two adjacent discrete sampling points includes: The center points of two adjacent discrete sampling points are slightly shifted to the left, and the coordinates of the left shift point on the time axis are calculated. The function expression for calculating the time coordinates of the left shift point is as follows: The center points of two adjacent discrete sampling points are slightly shifted to the right, and the coordinates of the right shift point on the time axis are calculated. The function expression for calculating the time coordinates of the right shift point is: In the formula, The time point corresponding to the leftward offset of the center point of the interpolation interval. This is the time point corresponding to the rightward offset of the center point of the interpolation interval; Substitute the time coordinates of the left offset point and the right offset point into the spatiotemporal weaving function to obtain the phase signal of the corresponding time coordinate.

7. The energy storage frequency modulation command prediction method based on spatiotemporal weaving coupling interpolation according to claim 4, characterized in that, The step of correcting the preliminary prediction sequence using the derivatives of the continuous phase signals to obtain the final prediction sequence includes: The stability coefficient is calculated using the derivative of the continuous phase signal, and the functional expression of the stability coefficient is: In the formula, For stability coefficient, This is the time scaling factor. The time point to be corrected; The final predicted sequence is obtained by correcting the initial predicted sequence using a stability coefficient. The corrected function expression is as follows: In the formula, This represents the final predicted value of the frequency modulation command at the (N+i)th discrete sampling point. This is the initial predicted value of the frequency modulation command for the (N+i)th discrete sampling point.

8. A prediction system, employing the energy storage frequency modulation command prediction method based on spatiotemporal weaving coupling interpolation as described in any one of claims 1-7, characterized in that, The prediction system includes: The signal conversion module is used to convert discretely acquired raw frequency modulation sequences into continuous phase signals based on spatiotemporal weaving. The chaos intensity detection module is used to calculate and obtain the chaos intensity index of each phase signal, and to identify the region in the corresponding phase signal that has abrupt change or chaotic behavior through the chaos intensity index. The interpolation module is used to calculate the interpolation based on the phase signal and the chaos intensity index of two adjacent discrete sampling points in the original frequency modulation sequence, and to obtain a smooth sequence by interpolating between two adjacent discrete sampling points in the original frequency modulation sequence. The prediction module is used to input the smoothed sequence into a pre-trained GRU network to obtain a preliminary prediction sequence, and to correct the preliminary prediction sequence using the derivative of the continuous phase signal to obtain the final prediction sequence.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the energy storage frequency modulation command prediction method based on spatiotemporal braided coupling interpolation as described in any one of claims 1-7.

10. A computer 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 energy storage frequency modulation command prediction method based on spatiotemporal braided coupling interpolation as described in any one of claims 1-7.

Images