A frequency modulation instruction prediction method and system for a hybrid energy storage auxiliary thermal power unit

Abstract

The application discloses a kind of hybrid energy storage auxiliary frequency modulation instruction prediction method and system of thermal power generating unit, belong to energy storage frequency modulation instruction prediction technical field, method: obtain original frequency modulation instruction sequence, carry out dynamic phase compensation;Phase reference value is calculated, and nonlinear correction is carried out;First correction signal is obtained;Information entropy gradient enhancement is carried out to sampling point;First correction signal is enhanced, and second correction signal is obtained;Chaotic time series reconstruction is carried out, and third correction signal is obtained;Chaotic time series reconstruction includes: constructing time delay embedding space to map to high-dimensional phase space, calculating local Lyapunov exponent, combining the time derivative of signal and chaotic adjustment factor to carry out dynamic reconstruction;Third correction signal is input to GRU neural network, and the output frequency modulation instruction prediction result is obtained.The application can predict the size of frequency modulation instruction in advance, let super-capacitor or battery act in advance, improve the yield value, and the degree of convenience and accuracy are high, with engineering applicability.

Description

Technical Field

[0001] This invention belongs to the field of energy storage frequency regulation command prediction technology, specifically relating to a method and system for predicting frequency regulation commands of hybrid energy storage-assisted thermal power units. Background Technology

[0002] Current methods for frequency regulation of hybrid energy storage (supercapacitor + lithium battery) assisted by thermal power units involve transmitting the difference between the frequency regulation command and the thermal power unit's signal to the hybrid energy storage. The battery handles the low-frequency portion, while the supercapacitor handles the high-frequency portion. However, signal transmission (frequency regulation command transmission to the supercapacitor / lithium battery) takes time, and the supercapacitor or lithium battery itself also requires response time. This creates a response time difference, further impacting profitability and ultimately the power plant's revenue. The original frequency regulation prediction has strong nonlinearity, increasing the difficulty of neural network prediction and presenting several drawbacks, such as the capture of complex patterns and nonlinear dynamic characteristics: nonlinear time series may contain complex patterns such as abrupt changes, multimodal fluctuations, and chaotic behavior (e.g., Lorentz systems). Traditional linear models cannot adequately model these patterns, and while neural networks can theoretically approximate them, they require higher model complexity. Another challenge is model training, including the risk of overfitting: complex nonlinear relationships require deeper networks or more parameters, but overfitting is prone to occur with small sample data. Optimization difficulty: the nonconvexity of the loss function may lead to getting trapped in local optima. For example, training a GRU (Gated Response Unit)... When using a gated recurrent unit (GRU), careful selection of the learning rate and initialization strategy is necessary. Furthermore, regarding data requirements and quality, learning nonlinear patterns typically requires a large amount of data. For example, predicting a chaotic system may require tens of thousands of time steps of samples to cover different states. There are also uncertainties inherent in dynamic systems, such as the sensitivity to chaotic behavior. Even if the model fits historical data well, long-term predictions may still fail due to exponentially amplified initial condition errors. Therefore, in summary, current frequency modulation prediction methods suffer from insufficient modeling, high model optimization difficulty, high requirements for data quantity and quality, and the potential for long-term prediction failure due to exponentially amplified initial condition errors, leading to increased uncertainty in dynamic systems. Summary of the Invention

[0003] This invention provides a method and system for predicting frequency regulation commands for hybrid energy storage-assisted thermal power units. The purpose is to solve the problems of current frequency regulation prediction methods, such as insufficient modeling, high difficulty in model optimization, high requirements for data volume and quality, and the possibility of long-term prediction failure due to exponential amplification of initial condition errors, which leads to increased uncertainty in dynamic systems.

[0004] To achieve the above objectives, the present invention adopts the following technical solution:

[0005] This invention provides a method for predicting frequency regulation commands for hybrid energy storage-assisted thermal power units, comprising the following steps:

[0006] S1. Obtain the original frequency modulation command sequence, and perform dynamic phase compensation on each sampling point of the original frequency modulation command sequence; establish a phase reference system to calculate the phase reference value, generate the phase offset, and perform nonlinear correction on the phase offset; obtain the first corrected signal.

[0007] S2. Based on the first correction signal, perform information entropy gradient enhancement on each sampling point; information entropy gradient enhancement includes: calculating the information entropy within the sliding window and calculating the information entropy gradient;

[0008] The first corrected signal is enhanced based on the information entropy gradient to obtain the second corrected signal;

[0009] S3. Based on the second correction signal, perform chaotic time sequence reconstruction on each sampling point to obtain the third correction signal;

[0010] Chaotic time series reconstruction includes: constructing a time delay embedding space to map to a high-dimensional phase space, calculating the local Lyapunov exponent, and combining the time derivative of the signal and the chaotic adjustment factor for dynamic reconstruction;

[0011] S4. Input the third correction signal into the GRU neural network and output the frequency regulation command prediction result to complete the frequency regulation command prediction of the hybrid energy storage assisted thermal power unit.

[0012] In some implementations, during dynamic phase compensation in S1:

[0013] Establishing a phase reference system includes: calculating the phase reference value of the current sampling point based on a preset signal window threshold and sequence period length; generating a phase offset includes: calculating the difference between the current sampling value and its phase reference value to obtain the phase offset; nonlinear phase correction includes: applying a hyperbolic tangent function to nonlinearly correct the phase offset to obtain a first corrected signal.

[0014] Furthermore, in S1, the phase reference value is calculated using the following formula:

[0015] ;

[0016] in, For the first Phase reference value at each sampling point The threshold value for the signal window. For the summation variable, The sequence period length, For historical rounding sampling points, The frequency modulation command value is the historical rounded sampling point.

[0017] Furthermore, in S1, the nonlinear phase correction is performed using the following formula:

[0018] ;

[0019] in, For the first The first corrected signal after nonlinear phase correction of each sampling point For the first The original frequency modulation command value of each sampling point It is the hyperbolic tangent function. For the first Phase offset of each sampling point For the current sampling point The standard deviation of the frequency modulation command value for a certain fixed window length.

[0020] In some implementations, in S2, the information entropy gradient enhancement specifically includes:

[0021] The information entropy of the first corrected signal is calculated within a sliding window; the information entropy gradient is calculated based on the change in information entropy and the path length; and the first corrected signal is enhanced by a modified linear unit function based on the information entropy gradient to obtain the second corrected signal.

[0022] Furthermore, in S2, the information entropy is calculated using the following formula:

[0023] ;

[0024] in, For the first Information entropy of each sampling point Calculate the window length for entropy. For the summation variable, To the first within the sliding window The normalized probability distribution corresponding to each sampling point.

[0025] Furthermore, in S2, the information entropy gradient is calculated using the following formula:

[0026] ;

[0027] in, For the first The information entropy gradient of each sampling point The time interval for gradient calculation. For the first Information entropy of each sampling point For the summation variable, and For the first The and the first The first corrected signal at each sampling point.

[0028] In some implementations, in S3, chaotic timing reconstruction specifically includes:

[0029] Constructing the time-delay embedding space: The second modified signal is mapped to a high-dimensional phase space using the time-delay embedding method to obtain a time-delay embedding vector; Calculating the local Lyapunov exponent: In the high-dimensional phase space, the local Lyapunov exponent of the current time-delay embedding vector is calculated to quantify the degree of chaos; Dynamically reconstructing the output: Combining the local Lyapunov exponent, the time derivative of the signal, and a preset chaos adjustment factor, the second modified signal is reconstructed to obtain the third modified signal.

[0030] In some implementations, the dynamic reconstruction output in S3 is expressed using the following formula:

[0031] ;

[0032] in, For the first The third corrected signal after chaotic timing reconstruction of each sampling point For the first The second corrected signal at each sampling point As a chaos regulator, For the first Local Lyapunov exponents at each sampling point For the first The time derivative of the second corrected signal at each sampling point.

[0033] This invention also provides a frequency regulation command prediction system for hybrid energy storage-assisted thermal power units. The system implements the steps of the aforementioned frequency regulation command prediction method for hybrid energy storage-assisted thermal power units. The system includes a nonlinear correction module, an information entropy gradient enhancement module, a chaotic time series reconstruction module, and a frequency regulation command prediction module, wherein:

[0034] Nonlinear correction module: used to acquire the original frequency modulation command sequence, perform dynamic phase compensation on each sampling point in the original frequency modulation command sequence; establish a phase reference system to calculate the phase reference value, generate the phase offset, perform nonlinear correction on the phase offset; and obtain the first corrected signal;

[0035] Information entropy gradient enhancement module: used to enhance the information entropy gradient of each sampling point based on the first correction signal; information entropy gradient enhancement includes: calculating the information entropy within the sliding window and calculating the information entropy gradient;

[0036] The first corrected signal is enhanced based on the information entropy gradient to obtain the second corrected signal;

[0037] Chaotic timing reconstruction module: used to reconstruct the chaotic timing of each sampling point based on the second correction signal to obtain the third correction signal;

[0038] Chaotic time series reconstruction includes: constructing a time delay embedding space to map to a high-dimensional phase space, calculating the local Lyapunov exponent, and combining the time derivative of the signal and the chaotic adjustment factor for dynamic reconstruction;

[0039] Frequency modulation command prediction module: It is used to input the third correction signal into the GRU neural network, output the frequency modulation command prediction result, and complete the frequency modulation command prediction of the hybrid energy storage auxiliary thermal power unit.

[0040] Compared with the prior art, the frequency regulation command prediction method and system for hybrid energy storage assisted thermal power units of the present invention have the following beneficial effects:

[0041] This invention discloses a method for predicting frequency regulation commands in hybrid energy storage-assisted thermal power units. The method involves establishing a phase reference system and calculating phase reference values ​​to generate a phase offset. After nonlinear correction of the offset, a first correction signal is output. Next, the information entropy within a sliding window and its gradient are calculated, and a second correction signal is output based on the gradient enhancement signal. Then, a time delay embedding space is constructed, the local Lyapunov exponent is calculated, and dynamic reconstruction is performed using the signal time derivative and a chaos adjustment factor to output a third correction signal. Finally, the third correction signal is input into a GRU network to output the frequency regulation command prediction result. This invention utilizes a triple heterogeneous feature enhancement approach—phase physical characteristics, information entropy statistical characteristics, and chaotic dynamic characteristics—to decompose the strong nonlinearity, abrupt changes, and chaotic characteristics of the original frequency regulation command into modelable independent dimensions. Phase compensation extracts periodic patterns and suppresses noise interference; information entropy gradient enhancement captures disordered abrupt changes and amplifies key fluctuations; and chaotic reconstruction quantifies system stability and dynamically corrects the evolution path, fundamentally reducing the modeling difficulty of complex nonlinear modes for the GRU network.

[0042] Furthermore, this invention employs a phase reference system based on the historical mean of the window, calculates information entropy based on a sliding window probability distribution, and rapidly calculates the local Lyapunov exponent using nearest neighbor points. Preprocessing relies solely on local data streams, meeting the real-time requirement of millisecond-level response in thermal power frequency regulation systems. The phase reference system reflects the periodic characteristics of power system frequency fluctuations; the Lyapunov exponent quantifies the chaotic boundary information entropy gradient of energy dispatch to identify critical points of system state transitions, avoiding the shortcomings of purely data-driven models and improving the engineering reliability of prediction results. The final prediction results are used to guide power allocation in hybrid energy storage systems combining supercapacitors and lithium batteries. By predicting the magnitude of frequency regulation commands in advance and compensating for signal transmission and equipment response delays, supercapacitors / batteries can act ahead of time, increasing profitability.

[0043] This invention improves prediction accuracy and has certain engineering applicability by employing a novel data preprocessing approach that combines dynamic phase compensation, information entropy gradient enhancement, and chaotic time series reconstruction. It utilizes a triple enhancement of physical characteristics (phase), information theory (entropy), and nonlinear dynamics (chaos). Attached Figure Description

[0044] The accompanying drawings are provided to further illustrate the invention and constitute a part of this invention. The illustrative embodiments and descriptions of the invention are used to explain the invention and do not constitute an undue limitation thereof.

[0045] Figure 1 This is a flowchart illustrating a method for predicting frequency regulation commands for a hybrid energy storage-assisted thermal power unit according to the present invention. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. The components of the embodiments of the present invention generally described and shown in the accompanying drawings can be arranged and designed in various different configurations.

[0047] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to indicate that selected embodiments of the invention are based on the embodiments in this invention. All other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0048] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, elements defined by the phrase "comprising one..." do not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0049] It should be noted that the apparatus and methods disclosed in the embodiments herein can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functions, and operations of possible implementations of apparatus, methods, and computer program products according to various embodiments herein. In this respect, each block in a flowchart or block diagram may represent a module, program, or part of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the accompanying drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system to perform the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions.

[0050] In addition, the functional modules in the various embodiments of this article can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0051] like Figure 1 As shown, the present invention provides a method for predicting frequency regulation commands for hybrid energy storage-assisted thermal power units, comprising the following steps:

[0052] S1. Obtain the original frequency modulation command sequence, and perform dynamic phase compensation on each sampling point of the original frequency modulation command sequence; establish a phase reference system to calculate the phase reference value, generate the phase offset, and perform nonlinear correction on the phase offset; obtain the first corrected signal.

[0053] S2. Based on the first correction signal, perform information entropy gradient enhancement on each sampling point; information entropy gradient enhancement includes: calculating the information entropy within the sliding window and calculating the information entropy gradient;

[0054] The first corrected signal is enhanced based on the information entropy gradient to obtain the second corrected signal;

[0055] S3. Based on the second correction signal, perform chaotic time sequence reconstruction on each sampling point to obtain the third correction signal;

[0056] Chaotic time series reconstruction includes: constructing a time delay embedding space to map to a high-dimensional phase space, calculating the local Lyapunov exponent, and combining the time derivative of the signal with the chaotic adjustment factor for dynamic reconstruction;

[0057] S4. Input the third correction signal into the GRU (Gated Recurrent Unit) neural network, output the frequency regulation command prediction result, and complete the frequency regulation command prediction of the hybrid energy storage auxiliary thermal power unit.

[0058] The frequency modulation command prediction method of this invention separates the periodic components of the signal through phase compensation to eliminate low-frequency interference, enhances the ability to identify key events by strengthening abrupt change points through information entropy gradient enhancement, and quantifies the system stability boundary through chaotic reconstruction to constrain prediction divergence. It decomposes complex frequency modulation commands into independently modelable physical, statistical, and dynamic features, reducing the difficulty of GRU modeling. The preprocessing adopts local sliding window calculation to avoid global data backtracking and meets the 200ms-level response delay requirement of thermal power units. Furthermore, the output prediction results directly guide the power allocation of hybrid energy storage, which can compensate for signal transmission and equipment response delays and improve the benefits of frequency modulation.

[0059] Preferably, the phase reference system established by the present invention includes: calculating the phase reference value of the current sampling point according to a preset signal window threshold and sequence period length; generating the phase offset includes: calculating the difference between the current sampling value and its phase reference value to obtain the phase offset; nonlinear phase correction includes: using a hyperbolic tangent function to perform nonlinear correction on the phase offset to obtain a first correction signal, dynamically calibrating the signal reference through the phase reference system, and suppressing noise amplification through nonlinear correction, thereby improving the signal-to-noise ratio and reducing prediction deviations caused by periodic fluctuations.

[0060] Furthermore, in the frequency modulation command prediction method of the present invention, the information entropy gradient enhancement specifically includes: calculating the information entropy of the first corrected signal within a sliding window; calculating the information entropy gradient based on the change in information entropy and the path length; and enhancing the first corrected signal by modifying the linear unit function based on the information entropy gradient to obtain the second corrected signal. The present invention quantifies data disorder by calculating entropy values, identifies system transient processes, amplifies only the entropy increase phase, suppresses entropy decrease noise, and improves the capture rate of key events such as frequency drops and harmonic oscillations.

[0061] Furthermore, chaotic temporal reconstruction specifically includes:

[0062] The second corrected signal is mapped to a high-dimensional phase space using a time-delay embedding method to obtain a time-delay embedding vector, thus constructing a time-delay embedding space. In the high-dimensional phase space, the local Lyapunov exponent of the current time-delay embedding vector is calculated to quantify the degree of chaos. Combining the local Lyapunov exponent, the signal's time derivative, and a preset chaos adjustment factor, the second corrected signal is reconstructed to obtain the third corrected signal. This invention extends a 1D (One-Dimensional) sequence to a 3D (Three-Dimensional) phase space and a time-delay embedding space, enabling real-time quantification of the degree of chaos. As an optional approach, such as... The system can predict instability risks when the warning time is >0, and can steadily improve the prediction results.

[0063] This invention also provides a frequency regulation command prediction system for hybrid energy storage-assisted thermal power units. The system implements the steps of the aforementioned frequency regulation command prediction method for hybrid energy storage-assisted thermal power units. The system includes a nonlinear correction module, an information entropy gradient enhancement module, a chaotic time series reconstruction module, and a frequency regulation command prediction module, wherein:

[0064] Nonlinear correction module: used to acquire the original frequency modulation command sequence, perform dynamic phase compensation on each sampling point in the original frequency modulation command sequence; establish a phase reference system to calculate the phase reference value, generate the phase offset, perform nonlinear correction on the phase offset; and obtain the first corrected signal;

[0065] Information entropy gradient enhancement module: used to enhance the information entropy gradient of each sampling point based on the first correction signal; information entropy gradient enhancement includes: calculating the information entropy within the sliding window and calculating the information entropy gradient;

[0066] The first corrected signal is enhanced based on the information entropy gradient to obtain the second corrected signal;

[0067] Chaotic timing reconstruction module: used to reconstruct the chaotic timing of each sampling point based on the second correction signal to obtain the third correction signal;

[0068] Chaotic time series reconstruction includes: constructing a time delay embedding space to map to a high-dimensional phase space, calculating the local Lyapunov exponent, and combining the time derivative of the signal and the chaotic adjustment factor for dynamic reconstruction;

[0069] Frequency modulation command prediction module: It is used to input the third correction signal into the GRU neural network, output the frequency modulation command prediction result, and complete the frequency modulation command prediction of the hybrid energy storage auxiliary thermal power unit.

[0070] The following detailed description of the frequency regulation command prediction method and system for hybrid energy storage assisted thermal power units according to the present invention will be provided through specific embodiments.

[0071] The frequency regulation command prediction method for hybrid energy storage assisted thermal power units of the present invention is performed according to the following steps:

[0072] 1. First-order correction and dynamic phase compensation;

[0073] Let the frequency modulation command be , ( , , (This indicates the number of samples, taken once per second).

[0074] 1.1 Establishing a phase reference frame:

[0075] ;

[0076] in, For the first Phase reference value at each sampling point The threshold value for the signal window. For the summation variable, The sequence period length, For historical rounding sampling points, The frequency modulation command value is the historical rounded sampling point.

[0077] W: Signal window threshold W=3, T: Sequence period length; ; The integer symbol; when Values ​​less than 0 are taken as 1. Meaning: Dynamic phase reference value, calculated by a sliding window using the average of historical periodic data, used to calibrate the current phase.

[0078] Calculation logic: For the current moment The average of the first three signal windows is taken;

[0079] 1.2 Generate phase offset;

[0080] ;

[0081] Meaning: The instantaneous offset of the current sampled value relative to the phase reference value, reflecting short-term phase fluctuations;

[0082] 1.3 Nonlinear phase correction;

[0083] ;

[0084] in, For the first The first corrected signal after nonlinear phase correction of each sampling point For the first The original frequency modulation command value of each sampling point It is the hyperbolic tangent function. For the first Phase offset of each sampling point For the current sampling point The standard deviation of the frequency modulation command value for a certain fixed window length.

[0085] The standard deviation of the first 20 values, for example =30, find the answer. The standard deviation, if <20, then calculate before One standard deviation.

[0086] Meaning: The corrected phase value, with the correction magnitude limited by the nonlinear function tanh (Hyperbolic Tangent Function);

[0087] Normalize the offset to Multiplying by the standard deviation to restore the actual dimensions allows the present invention to suppress minor noise interference while retaining a significant phase shift.

[0088] 2. Second-order correction and information entropy gradient enhancement;

[0089] ;

[0090] in, For the first Information entropy of each sampling point Calculate the window length for entropy. For the summation variable, To the first within the sliding window The normalized probability distribution corresponding to each sampling point. Meaning: Information entropy within the sliding window, quantifying the degree of disorder in time-series data. Entropy calculation window length ; By using data within the window The frequency is obtained by statistical analysis after binning (equal width binning);

[0091] The calculation logic of this invention is as follows:

[0092] The higher the entropy value, the more disordered the data within the window; the lower the entropy value, the more regular the data.

[0093] ;

[0094] in, For the first The information entropy gradient of each sampling point The time interval for gradient calculation. For the first Information entropy of each sampling point For the summation variable, and For the first The and the first The first corrected signal at each sampling point. Meaning: The gradient direction of information entropy, normalized by combining entropy change and path length;

[0095] The time interval for gradient calculation =3.

[0096] Denominator: the square root of the path length, i.e. The Euclidean distance of data changes within seconds is used to eliminate the influence of dimensions.

[0097] Physical meaning: A positive gradient indicates an increase in entropy (an increase in disorder), while a negative gradient indicates a decrease in entropy (an increase in order).

[0098] like , , If any term in the expression is less than zero, then that term becomes .

[0099] 2.1 Generate an enhanced signal;

[0100] ;

[0101] ReLU (Rectified Linear Unit): Rectifies the linear unit, retaining only the positive gradient; Output the original value, otherwise output 0. The learnable enhancement factor is 0.1.

[0102] This invention amplifies signal changes during entropy increase (sudden increase in disorder) to highlight key changes.

[0103] 3. Third-order correction and chaotic temporal sequence reconstruction;

[0104] 3.1 Constructing a latency embedding space;

[0105] ;

[0106] Delay embedding vectors map one-dimensional time series data to a high-dimensional phase space to capture chaotic properties.

[0107] The optimal time delay is determined by the cross-autocorrelation function; Embedding dimension, selected using the spurious nearest neighbor method; any index value less than 0 becomes... ;

[0108] 3.2 Calculate the local Lyapunov exponent;

[0109] ;

[0110] Meaning: Local Lyapunov exponent, quantifies the divergence rate (degree of chaos) of neighboring orbits in phase space.

[0111] : The nearest neighbor in historical data (minimum Euclidean distance); K: prediction step size K=2;

[0112] Time step 1 second;

[0113] >0 indicates local chaos. <0 indicates local stability.

[0114] 3.3 Dynamically reconstructed output;

[0115] ;

[0116] in, For the first The third corrected signal after chaotic timing reconstruction of each sampling point For the first The second corrected signal at each sampling point As a chaos regulator, For the first Local Lyapunov exponents at each sampling point For the first The time derivative of the second corrected signal at each sampling point. Meaning: Based on the signal enhanced by the entropy gradient, preset hyperparameters control the intensity of the chaotic correction. . Calculated using five-point differences.

[0117] Function: To capture the instantaneous trend of signal changes;

[0118] .

[0119] 4. Prediction;

[0120] Original frequency modulation sequence After being enhanced by physical characteristics (phase), information theory (entropy), and nonlinear dynamics (chaos), it becomes a sequence. It is fed into the GRU network for prediction.

[0121] In summary, the present invention provides a frequency regulation command prediction method and system for hybrid energy storage-assisted thermal power units. By leveraging the advantages of multimodal feature fusion and cross-domain feature coupling, it integrates three perspectives: signal processing (DPC, Dynamic Phase Compensation), information theory (IEGB, Information Entropy Gradient Boost), and nonlinear dynamics (CTR, Chaotic Time-series Reconstruction). Simultaneously, it captures the physical characteristics of periodic phase regularity (reference frame construction in DPC), the statistical characteristics of information density distribution (entropy gradient perception in IEGB), and the characteristics of chaotic attractor structure (Lyapunov exponential feedback in CTR), thereby achieving modeling from surface fluctuations to deep dynamic laws. This invention addresses spatiotemporal decoupling by separating long-term trends from short-term abrupt changes through phase compensation (DPC) and entropy gradient (IEGB), and extending a one-dimensional sequence into a high-dimensional phase space trajectory using time delay embedding (CTR), thus avoiding the dependence of traditional time series models on local correlations. It adapts to the latency requirements of industrial scenarios by employing moving averages in DPC, window entropy calculation in IEGB, local Lyapunov exponents in CTR, and ensemble sliding windows. Furthermore, this invention encodes phase and chaos into feature transformation rules, overcoming the limitations of purely data-driven approaches. It reveals the system state through information entropy gradients and meets the real-time response requirements of thermal power frequency regulation systems through Lyapunov exponents, demonstrating significant engineering application value.

[0122] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Anyone skilled in the art can readily implement the present invention according to the description and above. Any modifications, alterations, or equivalent variations made using the technical content disclosed above are equivalent embodiments of the present invention. Furthermore, any modifications, alterations, or variations made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the present invention.

Claims

1. A method for predicting frequency regulation commands for hybrid energy storage-assisted thermal power units, characterized in that, Includes the following steps: S1. Obtain the original frequency modulation command sequence and perform dynamic phase compensation for each sampling point in the original frequency modulation command sequence; establish a phase reference system to calculate the phase reference value, generate the phase offset, and perform nonlinear correction on the phase offset; The first correction signal is obtained; S2. Based on the first correction signal, perform information entropy gradient enhancement on each sampling point; Information entropy gradient enhancement includes: calculating the information entropy within the sliding window and calculating the information entropy gradient; The first corrected signal is enhanced based on the information entropy gradient to obtain the second corrected signal; S3. Based on the second correction signal, perform chaotic time sequence reconstruction on each sampling point to obtain the third correction signal; Chaotic time series reconstruction includes: constructing a time delay embedding space to map to a high-dimensional phase space, calculating the local Lyapunov exponent, and combining the time derivative of the signal and the chaotic adjustment factor for dynamic reconstruction; S4. Input the third correction signal into the GRU neural network and output the frequency regulation command prediction result to complete the frequency regulation command prediction of the hybrid energy storage assisted thermal power unit.

2. The frequency regulation command prediction method for hybrid energy storage assisted thermal power units according to claim 1, characterized in that, In S1, during dynamic phase compensation: Establishing a phase reference system includes: calculating the phase reference value of the current sampling point based on a preset signal window threshold and sequence period length; generating a phase offset includes: calculating the difference between the current sampling value and its phase reference value to obtain the phase offset; nonlinear phase correction includes: applying a hyperbolic tangent function to nonlinearly correct the phase offset to obtain a first corrected signal.

3. The frequency regulation command prediction method for hybrid energy storage assisted thermal power units according to claim 2, characterized in that, In S1, the phase reference value is calculated using the following formula: ； in, For the first Phase reference value at each sampling point The threshold value for the signal window. For the summation variable, The sequence period length, For historical rounding sampling points, The frequency modulation command value is the historical rounded sampling point.

4. The frequency regulation command prediction method for hybrid energy storage assisted thermal power units according to claim 2, characterized in that, In S1, the nonlinear phase correction is performed using the following formula: ； in, For the first The first corrected signal after nonlinear phase correction of each sampling point For the first The original frequency modulation command value of each sampling point It is the hyperbolic tangent function. For the first Phase offset of each sampling point For the current sampling point The standard deviation of the frequency modulation command value for a certain fixed window length.

5. The frequency regulation command prediction method for hybrid energy storage assisted thermal power units according to claim 1, characterized in that, In S2, the information entropy gradient enhancement specifically includes: The information entropy of the first corrected signal is calculated within a sliding window; the information entropy gradient is calculated based on the change in information entropy and the path length; and the first corrected signal is enhanced by a modified linear unit function based on the information entropy gradient to obtain the second corrected signal.

6. The frequency regulation command prediction method for hybrid energy storage assisted thermal power units according to claim 5, characterized in that, In S2, the information entropy is calculated using the following formula: ； in, For the first Information entropy of each sampling point Calculate the window length for entropy. For the summation variable, To the first within the sliding window The normalized probability distribution corresponding to each sampling point.

7. The frequency regulation command prediction method for hybrid energy storage assisted thermal power units according to claim 5, characterized in that, In S2, the information entropy gradient is calculated using the following formula: ； in, For the first The information entropy gradient of each sampling point The time interval for gradient calculation. For the first Information entropy of each sampling point For the summation variable, and For the first The and the first The first corrected signal at each sampling point.

8. The frequency regulation command prediction method for hybrid energy storage assisted thermal power units according to claim 1, characterized in that, In S3, the chaotic time series reconstruction specifically includes: Constructing the time-delay embedding space: The second modified signal is mapped to a high-dimensional phase space using the time-delay embedding method to obtain a time-delay embedding vector; Calculating the local Lyapunov exponent: In the high-dimensional phase space, the local Lyapunov exponent of the current time-delay embedding vector is calculated to quantify the degree of chaos; Dynamically reconstructing the output: Combining the local Lyapunov exponent, the time derivative of the signal, and a preset chaos adjustment factor, the second modified signal is reconstructed to obtain the third modified signal.

9. The frequency regulation command prediction method for hybrid energy storage assisted thermal power units according to claim 1, characterized in that, In S3, the dynamic reconstruction output uses the following formula: ； in, For the first The third corrected signal after chaotic timing reconstruction of each sampling point For the first The second corrected signal at each sampling point As a chaos regulator, For the first Local Lyapunov exponents at each sampling point For the first The time derivative of the second corrected signal at each sampling point.

10. A frequency regulation command prediction system for a hybrid energy storage-assisted thermal power unit, characterized in that, The system is used to implement the steps of the frequency regulation command prediction method for hybrid energy storage-assisted thermal power units according to any one of claims 1-9. The system includes a nonlinear correction module, an information entropy gradient enhancement module, a chaotic time series reconstruction module, and a frequency regulation command prediction module, wherein: Nonlinear correction module: used to acquire the original frequency modulation command sequence, perform dynamic phase compensation on each sampling point in the original frequency modulation command sequence; establish a phase reference system to calculate the phase reference value, generate the phase offset, perform nonlinear correction on the phase offset; and obtain the first corrected signal; Information entropy gradient enhancement module: used to enhance the information entropy gradient of each sampling point based on the first correction signal; information entropy gradient enhancement includes: calculating the information entropy within the sliding window and calculating the information entropy gradient; The first corrected signal is enhanced based on the information entropy gradient to obtain the second corrected signal; Chaotic timing reconstruction module: used to reconstruct the chaotic timing of each sampling point based on the second correction signal to obtain the third correction signal; Chaotic time series reconstruction includes: constructing a time delay embedding space to map to a high-dimensional phase space, calculating the local Lyapunov exponent, and combining the time derivative of the signal and the chaotic adjustment factor for dynamic reconstruction; Frequency modulation command prediction module: It is used to input the third correction signal into the GRU neural network, output the frequency modulation command prediction result, and complete the frequency modulation command prediction of the hybrid energy storage auxiliary thermal power unit.

Images