Frequency modulation instruction prediction method based on nonlinear domain and related product

Abstract

The invention discloses a frequency modulation instruction prediction method based on a nonlinear domain and a related product, and belongs to the technical field of power system energy storage equipment. According to the frequency modulation instruction prediction method based on the non-linear domain, the original frequency modulation sequence is constructed, multi-round iterative decomposition is performed, the optimal prediction sequence is screened in the non-linear domain and input into the neural network for prediction, and finally the prediction results are superposed and merged, so that the non-linear characteristics of the fused salt energy storage frequency modulation instruction can be effectively captured, and the prediction precision is improved; the method has the technical effect of solving the problem that a traditional method is insufficient in nonlinear fluctuation processing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy storage devices of power systems, in particular to a frequency modulation instruction prediction method based on a nonlinear domain and related products. BACKGROUND

[0002] In the new power system, molten salt energy storage has become a core frequency modulation resource for stabilizing new energy power generation fluctuations and ensuring power grid frequency stability due to its high heat storage density and fast adjustment response. The operation of the molten salt energy storage frequency modulation system relies on accurate frequency modulation instruction prediction. By predicting the frequency modulation demand issued by the power grid in advance, the system can pre-adjust the molten salt heat charging and discharging power and optimize the operation state of the heat exchanger, thereby avoiding waste of energy storage capacity or delay of frequency modulation response caused by sudden changes in instructions, which directly affects the power grid frequency qualification rate and the economic efficiency of the energy storage system.

[0003] However, the generation of molten salt energy storage frequency modulation instructions is influenced by multiple factors such as new energy output fluctuations (e.g. wind gusts, photovoltaic shadow blocking, load changes (e.g. industrial load start-stop, residential electricity peak-valley, power grid dispatching strategy adjustment, etc.), and its time series presents significant nonlinear characteristics: the nonlinear characteristic is that the instruction amplitude change has no fixed proportional relationship, such as a sudden increase in load causing the instruction to jump from 10MW to 35MW, while subsequent load fluctuations only cause a 5-8MW instruction change, which cannot be fitted by a linear equation.

[0004] In existing molten salt energy storage frequency modulation instruction prediction technology, traditional artificial intelligence methods have low precision due to their inability to adapt to the nonlinear characteristics of the instructions. The prediction method based on VMD decomposes the original frequency modulation sequence into a series of sub-sequences, then puts these sub-sequences into a neural network for prediction, and finally obtains the final molten salt energy storage frequency modulation instruction by superimposing the prediction results. Although decomposition improves the prediction effect, it still has the defect of insufficient complex nonlinear representation, resulting in prediction accuracy that cannot meet the high requirements of the new power system for molten salt energy storage frequency modulation. SUMMARY

[0005] The purpose of the present application is to provide a frequency modulation instruction prediction method based on a nonlinear domain and related products to overcome the problem of low prediction accuracy caused by the inability of existing technology to adapt to the nonlinear characteristics of molten salt energy storage frequency modulation instructions.

[0006] The present application solves the above technical problems by the following technical solutions: The present application provides a frequency modulation instruction prediction method based on a nonlinear domain, comprising the following steps: Step one, sample the historical data of the molten salt energy storage frequency modulation instruction in chronological order to construct the original frequency modulation sequence Px; Step two, preset the total number of iterations M, and let the current iteration number m be 0; Step three, let m=m+1, randomly decompose each sub-sequence of the original frequency modulation sequence Px according to the preset decomposition layer number K to obtain a decomposition sequence S m ; convert the decomposition sequence S m into a nonlinear domain to obtain a nonlinear domain sequence P m , calculate the nonlinear parameter Z m of the nonlinear domain sequence P zm ; judge whether the current iteration number m is equal to the total iteration number M, if yes, execute step four, if no, return to step three; Step four, obtain the decomposition sequence S zmin corresponding to the smallest nonlinear parameter Z min ; Step five, input the decomposition sequence S min as the optimal prediction sequence into the trained neural network for prediction to obtain a prediction result; Step six, combine the prediction results by superposition to obtain the final molten salt energy storage frequency modulation instruction.

[0007] The further improvement of the application is that the original frequency modulation sequence Px=[X1, X2, X3,..., X N ], X1, X2, X3,..., X N are the 1st, 2nd, 3rd,..., Nth sub-sequences of the original frequency modulation sequence Px respectively; Randomly decompose each sub-sequence of the original frequency modulation sequence according to the preset decomposition layer number K to obtain a decomposition sequence S m =[S m1 ,S m2 ,S m3 ,...,S mK ], S m1 ,S m2 ,S m3 ,...,S mK are the 1st, 2nd, 3rd,..., Kth sub-sequences of the decomposition sequence S m obtained by the mth random decomposition iteration, the 1st, 2nd, 3rd,..., Kth sub-sequences S m ,S m1 ,S m2 ,...,S m3 of the decomposition sequence S mK are specifically: S m1 =[X 11 ,X 12 ,X 13 ,...,X 1N ]; S m2 =[X 21 ,X 22 ,X23 ..., X 2N ] ; S m3 = [X 31 , X 32 , X 33 ,..., X 3N ] ; ... ; S mK = [X K1 , X K2 , X K3 ,..., X KN ] ; wherein: X1= X 11 + X 21 + X 31 +... + X K1 ; X2= X 12 + X 22 + X 32 +... + X K2 ; X3= X 13 + X 23 + X 33 +... + X K3 ; ... ; X N = X 1N + X 2N + X 3N +... + X KN .

[0008] The further improvement of the present application is that the decomposition sequence S m is converted to a nonlinear domain to obtain a nonlinear domain sequence P m = [[D 11 , D 12 , D 13 ,..., D 1N ], [D 21 , D 22 , D 23 ,..., D 2N ], [D 31 , D 32 , D 33 ,..., D 3N ],..., [D K1 , D K2 , D K3 ,..., D KN ]], wherein: [D 11 , D 12 , D 13 ,..., D1N ],[D 21 ,D 22 ,D 23 ,…,D 2N ],[D 31 ,D 32 ,D 33 ,…,D 3N ],…,[D K1 ,D K2 ,D K3 ,…,D KN ] are respectively the 1st, 2nd, 3rd, …, Kth subsequence of the nonlinear domain sequence P m ; D K1 ,D K2 ,D K3 ,…,D KN are respectively the 1st, 2nd, 3rd, …, Nth subsequence of the Kth subsequence [D K1 ,D K2 ,D K3 ,…,D KN ] of the nonlinear domain sequence P m , and specifically comprising the following steps: calculating the nonlinear parameter Z m of the nonlinear domain sequence P zm , and specifically comprising: Z zm =Z1+Z2+Z3+...+Z k =avg(D 11 ,D 12 ,D 13 ,…,D 1N )+avg(D 21 ,D 22 ,D 23 ,…,D 2N )+avg(D 31 ,D 32 ,D 33 ,…,D 3N )+…+avg(D K1 ,D K2 ,D K3 ,…,D KN ); wherein Z1, Z2, Z3, …, Z k are respectively the nonlinear degree of the 1st, 2nd, 3rd, …, Kth subsequence of the nonlinear domain sequence P m .

[0009] The further improvement of the present application is that the K is an even number. Step five can be replaced by the following method: decomposing the sequence Smin The subsequences are merged pairwise to obtain Group fusion sequence R1~R Calculate the nonlinear parameter Z for each fusion sequence. R1~ Z R The minimum nonlinear parameter Z Rmin The corresponding fusion sequence R min The optimal prediction sequence is input into the trained neural network for prediction, and the prediction result is obtained.

[0010] A further improvement of the present invention is that, suppose there exists a sequence A = [A1, A2, A3, ..., A...]. X ],A1,A2,A3,...,A X For the 1st, 2nd, 3rd, ..., Xth subsequences of sequence A, transform sequence A into a nonlinear domain to obtain a nonlinear domain sequence. This involves the following steps: Map all subsequences of sequence A to a Cartesian coordinate system. Based on the mapped positions of all subsequences of sequence A in the Cartesian coordinate system, calculate the rotation angle of each subsequence in sequence A and the average rotation angle of sequence A; the i-th subsequence A of sequence A... i The rotation angle is:

[0011] The average rotation angle of sequence A is:

[0012] in, For subsequence A i The rotation angle; For the subsequence A i Mapping position and subsequence A a The angle between the line mapping the position and the horizontal axis. For the subsequence A (X-1) Mapping position and subsequence A X The angle between the line mapping the position and the horizontal axis; Transform each subsequence in sequence A into the nonlinear domain as follows:

[0013] in, To subsequence A i Transform to the subsequence corresponding to the nonlinear domain; The nonlinear domain sequence P=[D1,D2,D3,…,D is obtained. X ], where D1, D2, D3, ..., D X To extract the subsequences A1, A2, A3, ..., A from sequence A XTransforming the corresponding subsequence to a nonlinear domain; The sequence A is a decomposition sequence S m Or the optimal prediction sequence of step five.

[0014] The further improvement of the application is that the included angle is an acute angle in absolute value.

[0015] The application further provides a molten salt energy storage frequency modulation instruction decomposition system, comprising: The first module is used for sampling historical data of the molten salt energy storage frequency modulation instruction in chronological order, and constructing an original frequency modulation sequence Px; The second module is used for randomly decomposing each subsequence of the original frequency modulation sequence according to a preset decomposition layer number K, to obtain a decomposition sequence S m Transforming the decomposition sequence S m to a nonlinear domain to obtain a nonlinear domain sequence P m , calculating a nonlinear parameter Z m of the nonlinear domain sequence P zm , and determining whether the current iteration number is equal to the total iteration number; The third module is used for obtaining a decomposition sequence S zmin corresponding to the smallest nonlinear parameter Z min ; The fourth module is used for inputting the decomposition sequence S min as an optimal prediction sequence to the trained neural network for prediction, to obtain a prediction result; The fifth module is used for merging the prediction results in a superimposed manner to obtain a final molten salt energy storage frequency modulation instruction.

[0016] The application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor realizes the steps of the frequency modulation instruction prediction method based on the nonlinear domain when executing the computer program.

[0017] The application further provides a computer readable storage medium, which stores a computer program, and the computer program realizes the steps of the frequency modulation instruction prediction method based on the nonlinear domain when executed by a processor.

[0018] The application further provides a computer program product comprising a computer program, and the computer program realizes the steps of the frequency modulation instruction prediction method based on the nonlinear domain when executed by a processor.

[0019] Compared with the prior art, the application has the following positive progress effects: The frequency modulation instruction prediction method based on a nonlinear domain provided by the application can effectively capture the nonlinear characteristics of the molten salt energy storage frequency modulation instruction, improve the prediction accuracy, and has the technical effect of solving the problem of insufficient processing of nonlinear fluctuations in the traditional method.

[0020] Further, by introducing a two-stage screening mechanism, the nonlinear interaction characteristics between the sub-sequences can be captured more finely after the initial decomposition and fusion re-screening. BRIEF DESCRIPTION OF DRAWINGS

[0021] The accompanying drawings are included to provide a further understanding of the application, and are incorporated in and constitute a part of this application, illustrate embodiments of the application and explain the application, and do not constitute an improper limitation on the application.

[0022] Figure 1 The flowchart of the frequency modulation instruction prediction method based on a nonlinear domain of the application.

[0023] Figure 2 The schematic diagram of mapping each original frequency modulation sub-sequence in the original frequency modulation sequence to the rectangular coordinate system of the application. DETAILED DESCRIPTION

[0024] To make the purpose, technical scheme and advantages of the embodiments of the application clearer, the technical scheme in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are some of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0025] In the description of the application, it should be understood that the terms "include" and "contain" indicate the presence of described features, whole, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, whole, steps, operations, elements, components and / or sets thereof.

[0026] It should also be understood that the terms used in the specification of the application are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in the specification and the appended claims of the application, the singular forms "a", "an" and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0027] It should be understood that although the terms first, second, third, etc. may be employed in describing the preset ranges, etc. in the embodiments of the present application, the preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from each other. For example, the first preset range can also be referred to as the second preset range, and similarly, the second preset range can also be referred to as the first preset range without departing from the scope of the embodiments of the present application.

[0028] The present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments, which are an explanation of the present application rather than a limitation.

[0029] Referring to Figure 1 , a frequency modulation instruction prediction method based on a nonlinear domain includes the following steps: Step one, sampling the historical data of the molten salt energy storage frequency modulation instruction in chronological order to construct an original frequency modulation sequence Px=[X1,X2,X3,...,X N ], X1,X2,X3,...,X N are the 1st, 2nd, 3rd,..., Nth subsequences of the original frequency modulation sequence Px respectively; Step two, preset the total iteration number M, and let the current iteration number m be 0; Step three, let m=m+1, and randomly decompose each subsequence of the original frequency modulation sequence by a preset decomposition layer number K to obtain a decomposition sequence S m =[S m1 ,S m2 ,S m3 ,...,S mK ], S m1 ,S m2 ,S m3 ,...,S mK are the 1st, 2nd, 3rd,..., Kth subsequences of the decomposition sequence S m obtained by the mth random decomposition iteration; convert the decomposition sequence S m to a nonlinear domain to obtain a nonlinear domain sequence P m , calculate the nonlinear parameter Z zm of the nonlinear domain sequence P m ; determine whether the current iteration number m is equal to the total iteration number M, if yes, execute step four, if no, return to step three; Step four, obtain the decomposition sequence S min corresponding to the smallest nonlinear parameter Z zmin ; Step five, input the decomposition sequence S min as the optimal prediction sequence to the trained neural network for prediction to obtain a prediction result; Step 6: Combine the prediction results by superposition to obtain the final molten salt energy storage frequency regulation command.

[0030] This method iteratively generates multiple decomposition schemes, each of which splits the original sequence into several subsequences. In the nonlinear domain, the geometric distribution characteristics of each subsequence are quantified into nonlinear parameters, which reflect the degree to which the decomposition scheme preserves the nonlinear features of the original sequence. Through multiple rounds of iterative comparison, the decomposition scheme with the smallest nonlinear parameter is selected and input into the neural network, enabling the prediction model to learn based on the optimal feature representation. Finally, the prediction result is used to reconstruct the complete instruction through the linear superposition of the predicted values ​​of the subsequences. This method maintains the dimensionality reduction advantage of decomposition processing while avoiding feature distortion through nonlinear domain optimization, effectively solving the problem of insufficient prediction accuracy caused by the nonlinear features of molten salt energy storage frequency modulation instructions.

[0031] The nonlinear domain parameter-guided decomposition and optimization process can accurately capture the dynamic changes in the command sequence. Based on the optimized feature input, the neural network can generate more accurate prediction results, thereby improving the frequency regulation response speed of the energy storage system and the stability of grid frequency control. This application can effectively identify and optimize potential nonlinear interference components during the frequency regulation command decomposition process, resulting in a clearer distribution of nonlinear features in the input data received by the neural network prediction model, thus improving the prediction accuracy for scenarios of sudden changes in grid frequency regulation demand.

[0032] In a specific embodiment of the present invention, the total number of iterations M ≥ 1000.

[0033] In a specific embodiment of the present invention, the subsequence of the original frequency modulation sequence Px is >1000.

[0034] Specifically, let the original frequency modulation sequence Px=[X1,X2,X3,...,X... N ],X1,X2,X3,...,X N These are the 1st, 2nd, 3rd, ..., Nth subsequences of the original frequency modulation sequence Px; Each subsequence of the original frequency modulation sequence is randomly decomposed according to a preset number of decomposition levels K to obtain the decomposed sequence S. m =[S m1 ,S m2 ,S m3 ,...,S mK ], S m1 ,S m2 ,S m3 ,...,S mK The decomposition sequence S is obtained from the m-th random decomposition iteration. m The 1st, 2nd, 3rd, ..., Kth subsequences are decomposed into sequence S. m The 1st, 2nd, 3rd, ..., Kth subsequence S m1 ,Sm2 S m3 ..., S mK Specifically: S m1 = [X 11 , X 12 , X 13 ,..., X 1N ] ; S m2 = [X 21 , X 22 , X 23 ,..., X 2N ] ; S m3 = [X 31 , X 32 , X 33 ,..., X 3N ] ; ... S mK = [X K1 , X K2 , X K3 ,..., X KN ] ; Wherein: X1= X 11 + X 21 + X 31 +... + X K1 ; X2= X 12 + X 22 + X 32 +... + X K2 ; X3= X 13 + X 23 + X 33 +... + X K3 ; ... X N = X 1N + X 2N + X 3N +... + X KN .

[0035] The original frequency modulation sequence is split into multiple independent sub-sequence sets, and the preset decomposition layer number controls the granularity of decomposition. For example, when the decomposition layer number is set to 5, each original sub-sequence is decomposed into 5 components. In each iteration, the decomposition process randomly generates different component combinations, so that the component distribution of the same original element in different decomposition levels presents diversity.

[0036] Specifically, suppose the decomposition sequence S mTransforming into the nonlinear domain, a nonlinear domain sequence P is obtained m =[[D 11 ,D 12 ,D 13 ,…,D 1N ],[D 21 ,D 22 ,D 23 ,…,D 2N ],[D 31 ,D 32 ,D 33 ,…,D 3N ],…,[D K1 ,D K2 ,D K3 ,…,D KN ]] 11 ,D 12 ,D 13 ,…,D 1N ],[D 21 ,D 22 ,D 23 ,…,D 2N ],[D 31 ,D 32 ,D 33 ,…,D 3N ],…,[D K1 ,D K2 ,D K3 ,…,D KN ] m are the 1st, 2nd, 3rd, …, Kth subsequence of the nonlinear domain sequence P K1 ,D K2 ,D K3 ,…,D KN m are the 1st, 2nd, 3rd, …, Nth subsequence of the Kth subsequence [D K1 ,D K2 ,D K3 ,…,D KN ] of the nonlinear domain sequence P m , and specifically comprising the following steps: calculating the nonlinear parameter Z zm of the nonlinear domain sequence P zm , specifically: Z k =Z1+Z2+Z3+...+Z 11 =avg(D 12 ,D 13 ,…,D 1N )+avg(D 21 ,D 22 ,D 23 ​,…,D 2N )+avg(D 31 D 32 D 33 ,…,D 3N )+…+avg(D K1 D K2 D K3 ,…,D KN ); Where Z1, Z2, Z3, ..., Z k These are nonlinear domain sequences P m The degree of nonlinearity of the 1st, 2nd, 3rd, ..., Kth subsequences.

[0037] After the decomposed sequence is transformed into the nonlinear domain, the mapping value of each subsequence in the nonlinear domain reflects its inherent nonlinear fluctuation characteristics. By calculating the average value of each subsequence in the nonlinear domain, a quantitative index of its nonlinear intensity can be extracted. The global parameter formed by superimposing the nonlinear indices of each subsequence can be used as a standard for evaluating the decomposition quality. During multiple iterations, the system compares the nonlinear parameters corresponding to different decomposition results and selects the decomposed sequence with the smallest nonlinear parameter as the optimal input, thereby ensuring that the nonlinear characteristics of the input data have been fully optimized during subsequent neural network prediction. This method, by introducing a nonlinear domain transformation mechanism, can transform the nonlinear dynamic characteristics of the decomposed sequence into quantifiable parameter indices, enabling the decomposition process to adaptively adjust according to the nonlinear intensity of the actual data, overcoming the problem of insufficient representation of traditional decomposition methods in complex nonlinear scenarios.

[0038] In a specific embodiment of the present invention, K is an even number; Step five can be replaced using the following methods: For the decomposed sequence S min The subsequences are merged pairwise to obtain Group fusion sequence R1~R Calculate the nonlinear parameter Z for each fusion sequence. R1~ Z R The minimum nonlinear parameter Z Rmin The corresponding fusion sequence R min The optimal prediction sequence is input into the trained neural network for prediction, and the prediction result is obtained.

[0039] Among them, the nonlinear parameters of each fusion sequence are calculated and the nonlinear domain sequence P is calculated. m The nonlinear parameter Z zm The same method is used.

[0040] In a specific embodiment of the present invention, the number of decomposition layers K=4, and the decomposition sequence Smin =[S1,S2,S3,S4], the two-by-two fusion is performed on the sub-sequences of the decomposition sequence S min to obtain a group of fusion sequences, including: a first group:

S1+S2,S3+S4

S1+S3,S2+S4

S1+S4,S2+S3

[0041] Specifically, assuming that there is a sequence A=[A1,A2,A3,...,A X ], A1,A2,A3,...,A X are the 1st, 2nd, 3rd,..., Xth sub-sequences of sequence A, sequence A is transformed to a nonlinear domain to obtain a nonlinear domain sequence, which specifically includes the following steps:

[0042] mapping all sub-sequences of sequence A to a rectangular coordinate system, based on the mapping positions of all sub-sequences of sequence A in the rectangular coordinate system, calculating the rotation angle of each sub-sequence of sequence A and the average rotation angle of sequence A; the rotation angle of the i-th sub-sequence A i of sequence A is:

[0043] the average rotation angle of sequence A is:

[0044] wherein, is the rotation angle of sub-sequence A i ; is the included angle between the straight line passing through the mapping position of sub-sequence A i and the mapping position of sub-sequence A a and the horizontal axis, is the included angle between the straight line passing through the mapping position of sub-sequence A (X-1) and the mapping position of sub-sequence A X and the horizontal axis; transforming each sub-sequence in sequence A to a nonlinear domain, specifically:

[0045] wherein, is the transformation of sub-sequence A i ​transforming the sub-sequences in the sequence A to corresponding sub-sequences in the nonlinear domain; obtaining a sequence P=[D1, D2, D3, …, D X ] in the nonlinear domain, wherein D1, D2, D3, …, D X are sub-sequences corresponding to the sub-sequences A1, A2, A3, …, A X in the sequence A, and the sequence A is a decomposed sequence S m or the optimal prediction sequence in step five.

[0046] Referring to Figure 2 , in a specific embodiment of the present application, a sequence A=[A1, A2, A3, …, A i …, A X ] (X>1000) is constructed, wherein the subscript represents the sampling number, the sampling interval is 1 second, and the order is determined by the sampling number. Each original frequency modulation sub-sequence A1, A2, A3, …, A X in the original frequency modulation sequence is mapped to a rectangular coordinate system, wherein the vertical axis of the rectangular coordinate system represents the magnitude of the sub-sequence value, and the horizontal axis of the rectangular coordinate system represents the order of the sub-sequence in the original frequency modulation sequence Pt.

[0047] For [A1, A2, A3, …, A i …, A X ] on the coordinate axis, a corresponding trend change angle

[0048] Taking A1 as an example, the generation method of the rotation angle is as follows: A1 is connected with A2, and an angle (an acute angle is selected and the absolute value is taken) is generated between the straight line and the horizontal axis A1 is connected with A3, and an angle (an acute angle is selected and the absolute value is taken) is generated between the straight line and the horizontal axis A1 is connected with A4, and an angle (an acute angle is selected and the absolute value is taken) is generated between the straight line and the horizontal axis , …, A1 is connected with A X , and an angle (an acute angle is selected and the absolute value is taken) is generated between the straight line and the horizontal axis .

[0049] The rotation angle of A1 is defined as

[0050] Taking A2 as an example, the generation method of the rotation angle is as follows: A2 is connected with A1, and an angle (an acute angle is selected and the absolute value is taken) is generated between the straight line and the horizontal axis A2 is connected with A3, and an angle (an acute angle is selected and the absolute value is taken) is generated between the straight line and the horizontal axis ​​A2 and A4, the straight line and the horizontal axis will produce an angle (selecting acute angle and taking absolute value) A2 and A4, the straight line and the horizontal axis will produce an angle (selecting acute angle and taking absolute value) X A2 and A4, the straight line and the horizontal axis will produce an angle (selecting acute angle and taking absolute value) .

[0051] The rotation angle of A2 is defined as , the rotation angle of [A3, A4,.A i .., A X ] is sequentially extended.

[0052] After obtaining the rotation angle , the

[0053] Px=[A1, A2, A3,.A i .., A X ] can be converted into the nonlinear domain P=[D1, D2, D3,.D i .., D X ] The specific conversion method is A i For example: ( is the corresponding rotation angle of A i ).

[0054] The method can adaptively extract local features in the sequence that deviate from the overall trend, such as the area where the frequency modulation instruction jumps due to the sudden change of wind power output, by dynamically calculating the rotation angle and the average rotation angle. The difference of the rotation angle is significantly increased, and the characteristic value of the area after the nonlinear domain conversion will be exponentially amplified, making it easier for the subsequent neural network to identify such mutation patterns. By dynamically mapping the sequence to the nonlinear domain, the nonlinear fluctuations caused by the sudden change of new energy output or load step change in the instruction sequence can be explicitly separated and strengthened, providing more distinguishable input features for the neural network, which can accurately identify different amplitude change patterns, thereby improving the prediction accuracy of the molten salt energy storage frequency modulation instruction and avoiding prediction lag or misjudgment caused by feature confusion.

[0055] In some specific embodiments of the present application, the included angle is an acute angle taking absolute value.

[0056] The acute angle taking absolute value refers to converting the calculated angle to a positive value within the range of 0 to 90 degrees, which is achieved by taking the absolute value of the calculation result and limiting it within the acute angle range, which can eliminate the influence of angle direction on nonlinear parameter calculation and make the nonlinear degree under different fluctuation directions comparable.

[0057] In the nonlinear domain conversion process, when calculating the rotation angle between the mapping positions of the subsequence, by taking the absolute value of the acute angle, the nonlinear characteristics of different fluctuation trends can be uniformly represented. For example, when a subsequence mapping point appears upward or downward fluctuation relative to the previous point, the angle formed with the horizontal axis may be positive or negative, respectively. However, after taking the absolute value of the acute angle, the angles of the two fluctuation directions are converted to positive values within the range of 0 to 90 degrees. Thus, when calculating the average rotation angle, the distortion of the nonlinear characteristic representation caused by the mutual offset of positive and negative angles can be avoided, and the selection of the acute angle range can focus on significant nonlinear fluctuations and filter the noise interference caused by small angle changes.

[0058] The method can accurately quantify the nonlinear characteristics of the frequency modulation instruction sequence, avoid the interference of angle direction difference and small fluctuations on feature extraction, and thus provide reliable nonlinear parameter basis for selecting the optimal decomposition sequence.

[0059] Based on the same inventive concept, a molten salt energy storage frequency modulation instruction decomposition system comprises: A first module is configured to sample the historical data of the molten salt energy storage frequency modulation instruction in chronological order, and construct an original frequency modulation sequence. A second module is configured to randomly decompose each subsequence of the original frequency modulation sequence according to a preset decomposition layer number K, to obtain a decomposition sequence S m =[S m1 ,S m2 ,S m3 ,...,S mK ],S m1 ,S m2 ,S m3 ,...,S mK are the 1st, 2nd, 3rd, …, Kth subsequences of the decomposition sequence S m obtained by the mth random decomposition iteration; and the decomposition sequence S m is converted to a nonlinear domain to obtain a nonlinear domain sequence P m , and the nonlinear parameter Z m of the nonlinear domain sequence P zm is calculated, until the current iteration number is equal to the total iteration number. A third module is configured to obtain the decomposition sequence S zmin corresponding to the smallest nonlinear parameter Z min . A fourth module is configured to input the decomposition sequence S min as an optimal prediction sequence to a trained neural network for prediction, to obtain a prediction result. A fifth module is configured to merge the prediction results by superposition to obtain the final molten salt energy storage frequency modulation instruction.

[0060] In order to further verify the advantages of the method, the molten salt energy storage frequency modulation instructions are predicted by using the method and the VMD decomposition method respectively, and the trained neural network uses the GRU neural network, and the results are as follows: The original frequency modulation sequence is obtained from a power plant in Inner Mongolia, frequency modulation sequence 1 is collected from 2:00 to 16:00 on a certain day in January 2025, and a data point is collected every 1 second; frequency modulation sequence 2 is collected from 4:00 to 22:00 on a certain day in March 2025, and a data point is collected every 1 second.

[0061]

[0062] From the experimental results, it can be seen that the four evaluation indexes of the method are reduced, which shows that the prediction accuracy of the method is higher.

[0063] Based on the same inventive concept, the embodiments of the present application provide a computer device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the frequency modulation instruction prediction method based on a nonlinear domain when executing the computer program. The memory can include a memory, such as a high-speed random access memory, and can also include a non-volatile memory, such as at least one disk memory. The processor, network interface and memory are connected to each other through an internal bus, which can be an industry standard architecture bus, a peripheral component interconnect standard bus, an extended industry standard architecture bus, etc. The bus can be divided into an address bus, a data bus and a control bus. The memory is used to store programs, specifically, the programs can include program codes, and the program codes include computer operation instructions. The memory can include a memory and a non-volatile memory, and provide instructions and data to the processor.

[0064] Based on the same inventive concept, the embodiments of the present application provide a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the frequency modulation instruction prediction method based on a nonlinear domain. Specifically, the computer readable storage medium includes but is not limited to, for example, a volatile memory and / or a non-volatile memory. The volatile memory can include RAM (Random Access Memory) and / or cache memory, etc. The non-volatile memory can include ROM (Read-Only Memory), hard disk, flash memory, optical disk, magnetic disk, etc.

[0065] Based on the same inventive concept, the embodiment of the present application provides a computer program product, which comprises a computer program stored on a computer readable storage medium, and the computer program comprises program instructions, which, when executed by a computer device, cause the computer device to perform the steps of the above-mentioned method for predicting frequency modulation instructions based on a nonlinear domain.

[0066] Those skilled in the art will understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM (Compact Disc Read-Only Memory), optical storage, etc.) containing computer-usable program code.

[0067] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions, which are executed via the processor of the computer or other programmable data processing apparatus, generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for carrying out the functions specified in the flowcharts and / or block diagrams.

[0068] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction means, which implement the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for carrying out the functions specified in the flowcharts and / or block diagrams.

[0069] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are performed on the computer or other programmable data processing apparatus to produce a computer-implemented process, so that the instructions executed on the computer or other programmable data processing apparatus provide a process for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for carrying out the functions specified in the flowcharts and / or block diagrams.

[0070] Finally, it should be noted that the above examples are only one or more specific forms of the present application, and their purpose is to clearly illustrate the concept, principle and application of the present application through specific examples, and are not intended to limit the scope of protection of the present application to these specific examples. In fact, the true value of the present application lies in its proposed technical ideas and innovative points, not in its forms or means of implementation.

[0071] For those skilled in the art, after reading and understanding the technical solutions of the present application, they have the ability to make various forms of changes, modifications or equivalent replacements to the specific embodiments of the application based on their own professional knowledge and skills. These changes may include but are not limited to adjusting the value range of technical parameters, optimizing the algorithm process to improve efficiency, replacing part of the technical components to achieve better compatibility or reduce cost, etc. As long as the technical solutions after these changes still maintain the technical features required by the original application, that is, still can realize the core function and effect of the present application, these changes should be considered as falling within the scope of protection of the claims of the present application.

[0072] In addition, with the continuous progress and development of technology, new technical means and methods are emerging, which also provides a broad space for the further improvement and perfection of the present application. Therefore, the scope of protection of the present application should also include those reasonable foreseeable improvements and extensions based on the existing technology, as long as these improvements and extensions do not deviate from the basic principles and core concept of the present application, they should be considered as the equivalent of the present application, and also be protected by the patent right.

Claims

1. A frequency modulation command prediction method based on nonlinear domain, characterized in that, Includes the following steps: Step 1: Sample the historical data of molten salt energy storage frequency modulation commands in chronological order to construct the original frequency modulation sequence Px; Step 2: Preset the total number of iterations M, and set the current iteration number m to 0; Step 3: Let m = m + 1, and randomly decompose each subsequence of the original frequency modulation sequence Px according to the preset decomposition level K to obtain the decomposed sequence S. m Decompose the sequence S m Transforming to the nonlinear domain, we obtain the nonlinear domain sequence P. m Calculate the nonlinear domain sequence P m The nonlinear parameter Z zm ; Determine if the current iteration number m is equal to the total iteration number M. If yes, proceed to step four. If the result is negative, return to step three; Step 4: Obtain the minimum nonlinear parameter Z zmin The corresponding decomposition sequence S min ; Step 5: Decompose the sequence S min The optimal prediction sequence is input into a trained neural network for prediction, and the prediction result is obtained. Step 6: Combine the prediction results by superposition to obtain the final molten salt energy storage frequency regulation command.

2. The frequency modulation command prediction method based on nonlinear domain according to claim 1, characterized in that, let... The original frequency modulation sequence Px=[X1,X2,X3,...,X] N ],X1,X2,X3,...,X N These are the 1st, 2nd, 3rd, ..., Nth subsequences of the original frequency modulation sequence Px; Each subsequence of the original frequency modulation sequence is randomly decomposed according to a preset number of decomposition levels K to obtain the decomposed sequence S. m =[S m1 ,S m2 ,S m3 ,...,S mK ], S m1 ,S m2 ,S m3 ,...,S mK The decomposition sequence S is obtained from the m-th random decomposition iteration. m The 1st, 2nd, 3rd, ..., Kth subsequences are decomposed into sequence S. m The 1st, 2nd, 3rd, ..., Kth subsequence S m1 ,S m2 ,S m3 ,...,S mK Specifically: S m1 =[X 11 ,X 12 ,X 13 ,...,X 1N ]; S m2 =[X 21 ,X 22 ,X 23 ,...,X 2N ]; S m3 =[X 31 ,X 32 ,X 33 ,...,X 3N ]; ...； S mK =[X K1 ,X K2 ,X K3 ,...,X KN ]; in: X1=X 11 +X 21 +X 31 +…+X K1 ; X2=X 12 +X 22 +X 32 +…+X K2 ; X3=X 13 +X 23 +X 33 +…+X K3 ; …； X N = X 1N +X 2N +X 3N +…+X KN 。 3. The frequency modulation command prediction method based on nonlinear domain according to claim 2, characterized in that, Suppose that the decomposed sequence S m Transforming to the nonlinear domain, we obtain the nonlinear domain sequence P. m =[[D 11 D 12 D 13 ,…,D 1N ],[D 21 D 22 D 23 ,…,D 2N ],[D 31 D 32 D 33 ,…,D 3N ],…,[D K1 D K2 D K3 ,…,D KN ]], where: [D 11 D 12 D 13 ,…,D 1N ],[D 21 D 22 D 23 ,…,D 2N ],[D 31 D 32 D 33 ,…,D 3N ],…,[D K1 D K2 D K3 ,…,D KN [P] represents the nonlinear domain sequence P. m The 1st, 2nd, 3rd, ..., Kth subsequence; D K1 D K2 D K3 ,…,D KN These are nonlinear domain sequences P m The Kth subsequence [D K1 D K2 D K3 ,…,D KN The 1st, 2nd, 3rd, ..., Nth subsequence of ] specifically includes the following steps: Calculate the nonlinear domain sequence P m The nonlinear parameter Z zm Specifically: Z zm =Z1+Z2+Z3+...+Z k =avg(D 11 ,D 12 ,D 13 ,…,D 1N )+avg(D 21 ,D 22 ,D 23 ,…,D 2N )+avg(D 31 ,D 32 ,D 33 ,…,D 3N )+…+avg(D K1 ,D K2 ,D K3 ,…,D KN ); Where Z1, Z2, Z3, ..., Z k These are nonlinear domain sequences P m The degree of nonlinearity of the 1st, 2nd, 3rd, ..., Kth subsequences.

4. The frequency modulation command prediction method based on the nonlinear domain according to claim 3, characterized in that, The K is an even number; Step five can be replaced using the following methods: For the decomposed sequence S min The subsequences are merged pairwise to obtain Group fusion sequence R1~R Calculate the nonlinear parameter Z for each fusion sequence. R1~ Z R The minimum nonlinear parameter Z Rmin The corresponding fusion sequence R min The optimal prediction sequence is input into the trained neural network for prediction, and the prediction result is obtained.

5. A frequency modulation command prediction method based on nonlinear domain according to any one of claims 1 to 4, characterized in that, let... There exists a sequence A = [A1, A2, A3, ..., A... X ],A1,A2,A3,...,A X For the 1st, 2nd, 3rd, ..., Xth subsequences of sequence A, transform sequence A into a nonlinear domain to obtain a nonlinear domain sequence. This involves the following steps: Map all subsequences of sequence A to a Cartesian coordinate system. Based on the mapped positions of all subsequences of sequence A in the Cartesian coordinate system, calculate the rotation angle of each subsequence in sequence A and the average rotation angle of sequence A; the i-th subsequence A of sequence A... i The rotation angle is: The average rotation angle of sequence A is: in, For subsequence A i The rotation angle; For the subsequence A i Mapping position and subsequence A a The angle between the line mapping the position and the horizontal axis. For the subsequence A (X-1) Mapping position and subsequence A X The angle between the line mapping the position and the horizontal axis; Transform each subsequence in sequence A into the nonlinear domain as follows: in, To subsequence A i Transform to the subsequence corresponding to the nonlinear domain; The nonlinear domain sequence P=[D1,D2,D3,…,D is obtained. X ], where D1, D2, D3, ..., D X To extract the subsequences A1, A2, A3, ..., A from sequence A X Transform to the subsequence corresponding to the nonlinear domain; The sequence A is a decomposed sequence S. m Or the optimal prediction sequence described in step five.

6. The frequency modulation command prediction method based on the nonlinear domain according to claim 5, characterized in that, The included angle is an acute angle whose absolute value is taken.

7. A molten salt energy storage frequency modulation command decomposition system, characterized in that, include: The first module is used to sample historical data of molten salt energy storage frequency modulation commands in chronological order to construct the original frequency modulation sequence Px; The second module is used to randomly decompose each subsequence of the original frequency modulation sequence according to a preset number of decomposition levels K, to obtain the decomposed sequence S. m Decompose the sequence S m Transforming to the nonlinear domain, we obtain the nonlinear domain sequence P. m Calculate the nonlinear domain sequence P m The nonlinear parameter Z zm Continue until the current iteration count equals the total number of iterations; The third module is used to obtain the minimum nonlinear parameter Z. zmin The corresponding decomposition sequence S min ; The fourth module is used to decompose the sequence S. min The optimal prediction sequence is input into a trained neural network for prediction, and the prediction result is obtained. The fifth module is used to combine the prediction results by superposition to obtain the final molten salt energy storage frequency regulation command.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the frequency modulation command prediction method based on any one of claims 1 to 4.

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 steps of the frequency modulation command prediction method based on any one of claims 1 to 4.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the frequency modulation command prediction method based on any one of claims 1 to 4.

Images