Energy storage frequency modulation instruction prediction method and system with complementary advantages

Abstract

The invention discloses an energy storage frequency modulation instruction prediction method and system with complementary advantages, and relates to the field of energy storage frequency modulation instruction prediction.The method comprises the steps that VMD decomposition and CEEMDAN decomposition are conducted on a frequency modulation instruction to be predicted, and a first subsequence set and a second subsequence set are obtained respectively; sequence interspersing is carried out, and first judgment is carried out; carrying out sequence fragment exchange on all mutually matched subsequences based on the result of the first judgment; predicting the subsequences after sequence fragment exchange through a GRU network, and correcting a prediction result to obtain a final energy storage frequency modulation instruction prediction result; according to the method, the advantages of VMD and CEEMDAN decomposition methods are integrated, the influence of noise and initial conditions on the decomposition effect is reduced, and the stability and reliability of decomposition are improved; the prediction accuracy of the frequency modulation instruction is improved; the size of the frequency modulation instruction is predicted in advance, the super capacitor / battery acts in advance, the response time difference is reduced, and the profit value of the power plant is improved.

Description

Technical Field

[0001] The present invention relates to the field of energy storage frequency modulation instruction prediction, and in particular to an energy storage frequency modulation instruction prediction method and system with complementary advantages. Background Art

[0002] With the continuous development of the power system, the requirements for grid frequency stability are increasing. Energy storage systems play an important role in assisting the frequency regulation of thermal power units, among which hybrid energy storage (supercapacitor + lithium battery) assisting the frequency regulation of thermal power units is a common technical means. In the traditional hybrid energy storage assisted thermal power unit frequency regulation method, the difference between the frequency regulation command and the thermal power unit is transmitted to the hybrid energy storage, the low-frequency part is borne by the battery, and the high-frequency part is borne by the supercapacitor. However, this method has certain limitations. It takes time for the signal to be transmitted from the frequency regulation command to the supercapacitor or lithium battery, and it also takes time for the supercapacitor or lithium battery itself to respond, which leads to the problem of response time difference. This response time difference will have a negative impact on the frequency regulation effect, and thus affect the benefits of the power plant.

[0003] In order to solve the problem of response time difference, some methods for predicting the size of frequency modulation instructions have emerged. Traditional prediction methods usually use technologies such as variational mode decomposition (VMD) or complete empirical mode decomposition with adaptive noise (CEEMDAN) to decompose the original sequence into a series of subsequences, and then predict these subsequences separately, and finally superimpose the prediction results to obtain the final prediction value. However, these traditional decomposition methods have their own inherent shortcomings. VMD decomposition has the problems of mode aliasing and end confusion, which will affect the accuracy of the decomposition results. When the CEEMDAN algorithm processes noisy signals, the decomposition results are easily affected by noise. Since the noise is close to the signal frequency, the useful signal may be mistaken for noise, thereby reducing the decomposition effect. In addition, the CEEMDAN algorithm is greatly affected by the initial conditions, which may lead to differences in the results of different decompositions. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: how to improve the accuracy of frequency modulation instruction prediction and reduce the impact of response time difference on power plant revenue.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, an embodiment of the present invention provides a method for predicting energy storage frequency modulation instructions with complementary advantages, comprising: Perform VMD decomposition and CEEMDAN decomposition on the frequency modulation instructions to be predicted to obtain a first subsequence set and a second subsequence set respectively; Interleave the sequence based on the first subsequence set and the second subsequence set, and make a first judgment; The performing sequence interleaving based on the first subsequence set and the second subsequence set comprises: Suppose a subsequence in the first subsequence set, that is, a subsequence decomposed by VMD, is IMF i =[X i1 ,X i2 ,X i3 ,...,X ii ,...,X iN ], let the experimental group error of the sequence be W i ; Suppose a subsequence in the second subsequence set, that is, a subsequence IMF decomposed by CEEMDAN j 、 =[X j1 、 ,X j2 、 ,X j3 、 ,...,X ji 、 ,...,X jN 、 ], let the experimental group error of the sequence be W j 、 ; Interleave the two subsequences, expressed as [X i1 ,X j1 、 ,X i2 ,X j2 、 , X i3 ,X j3 、 ,...,X ii ,X ji 、 ,...,X iN ,X jN 、 ], the experimental group error generated is W ij , W ij =(W i +W j 、 ) / 2; The first determination comprises: The test group error generated by the two subsequences interlaced with each other is judged. and , then the two subsequences interlaced are considered to match each other, where avg() means taking the average value; Based on the result of the first judgment, sequence fragments are exchanged for all mutually matching subsequences; for unmatched subsequences, no sequence fragment exchange operation is performed; the mutual matching refers to the mutual matching of complete subsequences; The step of performing sequence segment exchange on all mutually matching subsequences based on the result of the first judgment includes: Suppose any pair of matching subsequences is: VMD decomposed subsequence IMF i =[X i1 ,X i2 ,X i3 ,...,X ii ,...,X iN ] and CEEMDAN decomposition of subsequence IMF j 、 =[X j1 、 ,X j2 、 ,X j3 、 ,...,X ji 、 ,...,X jN 、 ], exchange the sequence fragments of length Q for this pair of subsequences; The step of exchanging sequence segments having a length of Q comprises: The random range of the length of Q is expressed as: , where N is the length of the subsequence to be exchanged, and [ ] means rounding down; The step of performing sequence segment exchange on all mutually matching subsequences based on the result of the first judgment further includes: For each pair of mutually matching subsequences, when the length Q of the exchange sequence fragment is determined, fragments of length Q are selected from different starting positions in the subsequence for exchange operation; after each exchange operation of selecting fragments at different starting positions, the corresponding average error is calculated, and the exchange with the smallest average error is taken as the final result of sequence fragment exchange;

[0007] All subsequences that have undergone sequence segment exchange operations are input into the GRU network for prediction, and the prediction results are corrected to obtain the final energy storage frequency modulation instruction prediction results.

[0008] As a preferred solution of the energy storage frequency modulation instruction prediction method with complementary advantages, wherein: The step of inputting all subsequences that have undergone the sequence segment exchange operation into the GRU network for prediction and correcting the prediction results includes: The subsequences after sequence fragment exchange are predicted through the GRU network, and the prediction result sequence of the obtained prediction group is assumed to be C1, C2, C3, ..., C K and C1 、 , C2 、 ,C3 、 ,...,C K 、 .

[0009] As a preferred solution of the energy storage frequency modulation instruction prediction method with complementary advantages, wherein: The step of inputting all subsequences that have undergone the sequence segment exchange operation into the GRU network for prediction and correcting the prediction results also includes: Suppose the prediction result of a sequence after sequence segment exchange is: C r =[X N+1 ,X N+2 ,X N+3 ,...,X N+i ,...,X N+0.1N ]; Correct the corresponding results to: ,in, It is the length of the DNA fragment when the corresponding sequence undergoes fragment exchange.

[0010] As a preferred solution of the energy storage frequency modulation instruction prediction method with complementary advantages, wherein: The step of inputting all subsequences that have undergone the sequence segment exchange operation into the GRU network for prediction and correcting the prediction results also includes: The final energy storage frequency modulation command prediction result is expressed as 0.5 (C1+C2+C3+...+C K +C1 、 +C2 、 +C3 、 +...+C K 、 ).

[0011] In a second aspect, an embodiment of the present invention provides a system for predicting energy storage frequency modulation instructions with complementary advantages, including: An initial decomposition module, used for performing VMD decomposition and CEEMDAN decomposition on the frequency modulation instruction to be predicted, to obtain a first subsequence set and a second subsequence set respectively; A matching judgment module, used for performing sequence interleaving based on the first subsequence set and the second subsequence set, and performing a first judgment; The performing sequence interleaving based on the first subsequence set and the second subsequence set comprises: Suppose a subsequence in the first subsequence set, that is, a subsequence decomposed by VMD, is IMFi =[X i1 ,X i2 ,X i3 ,...,X ii ,...,X iN ], let the experimental group error of the sequence be W i ; Suppose a subsequence in the second subsequence set, that is, a subsequence IMF decomposed by CEEMDAN j 、 =[X j1 、 ,X j2 、 ,X j3 、 ,...,X ji 、 ,...,X jN 、 ], let the experimental group error of the sequence be W j 、 ; Interleave the two subsequences, expressed as [X i1 ,X j1 、 ,X i2 ,X j2 、 , X i3 ,X j3 、 ,...,X ii ,X ji 、 ,...,X iN ,X jN 、 ], the experimental group error generated is W ij , W ij =(W i +W j 、 ) / 2; The first determination comprises: The test group error generated by the two subsequences interlaced with each other is judged. and , then the two subsequences interlaced are considered to match each other, where avg() means taking the average value; A fragment exchange module, used for performing sequence fragment exchange on all mutually matching subsequences based on the result of the first judgment; for unmatched subsequences, no sequence fragment exchange operation is performed; the mutual matching refers to the mutual matching of complete subsequences; The step of performing sequence segment exchange on all mutually matching subsequences based on the result of the first judgment includes: Suppose any pair of matching subsequences is: VMD decomposed subsequence IMF i =[X i1 ,X i2 ,X i3 ,...,X ii ,...,X iN ] and CEEMDAN decomposition of subsequence IMF j 、 =[X j1 、 ,X j2 、 ,X j3 、 ,...,X ji 、 ,...,X jN 、 ], exchange the sequence fragments of length Q for this pair of subsequences; The step of exchanging sequence segments having a length of Q comprises: The random range of the length of Q is expressed as: , where N is the length of the subsequence to be exchanged, and [ ] means rounding down; The step of performing sequence segment exchange on all mutually matching subsequences based on the result of the first judgment further includes: For each pair of mutually matching subsequences, when the length Q of the exchange sequence fragment is determined, fragments of length Q are selected from different starting positions in the subsequence for exchange operation; after each exchange operation of selecting fragments at different starting positions, the corresponding average error is calculated, and the exchange with the smallest average error is taken as the final result of sequence fragment exchange; The prediction module is used to input all subsequences that have undergone sequence segment exchange operations into the GRU network for prediction, and to correct the prediction results to obtain the final energy storage frequency modulation instruction prediction results.

[0012] In a third aspect, an embodiment of the present invention provides a computing device, including: Memory and processor; The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the one or more programs are executed by the one or more processors, the one or more processors implement the energy storage frequency modulation instruction prediction method with complementary advantages as described in any embodiment of the present invention.

[0013] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, can implement the complementary energy storage frequency modulation instruction prediction method.

[0014] Beneficial effects of the present invention: Traditional VMD decomposition has problems of modal aliasing and end confusion. The present invention can effectively suppress these shortcomings of VMD decomposition through a prediction method with complementary advantages, making signal decomposition more accurate; when the CEEMDAN algorithm processes noisy signals, the decomposition results are easily affected by noise and are greatly affected by initial conditions, and there may be differences in the decomposition results of different times. The present invention combines the advantages of VMD and CEEMDAN decomposition methods, reduces the influence of noise and initial conditions on the decomposition effect, and improves the stability and reliability of decomposition; the present invention combines VMD and CEEMDAN decomposition methods, gives full play to the advantages of the two methods through sequence matching and DNA exchange operations, and comprehensively processes the frequency modulation instruction sequence, thereby improving the prediction accuracy of the frequency modulation instruction; predicts the exchanged subsequence, and corrects the prediction result, further reducing the prediction error, making the prediction result closer to the actual value; can predict the size of the frequency modulation instruction in advance, let the supercapacitor / battery act in advance, reduce the response time difference, and thus improve the revenue value of the power plant. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0016] Figure 1 It is an overall flow chart of the energy storage frequency modulation instruction prediction method with complementary advantages described in the present invention. DETAILED DESCRIPTION

[0017] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0018] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0019] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0020] Example 1, reference Figure 1 , which is the first embodiment of the present invention, and which provides a method for predicting energy storage frequency modulation instructions with complementary advantages, comprising: S1: Perform VMD decomposition and CEEMDAN decomposition on the frequency modulation instruction to be predicted to obtain a first subsequence set and a second subsequence set respectively; S2: performing sequence interleaving based on the first subsequence set and the second subsequence set, and performing a first judgment; S3: Based on the result of the first judgment, sequence fragments are exchanged for all mutually matching subsequences; S4: The subsequence after the sequence fragment exchange is predicted through the GRU network, and the prediction result is corrected to obtain the final energy storage frequency modulation instruction prediction result.

[0021] It should be noted that, through steps S1-S4, VMD decomposition and CEEMDAN decomposition are first applied to the frequency modulation instructions to be predicted to generate the corresponding subsequence set. Then, these subsequences are interspersed and the matching is judged, and then the sequence fragments are exchanged for the matching subsequences. Finally, the GRU network is used to predict the exchanged subsequences, and the prediction results are corrected to obtain the final energy storage frequency modulation instruction prediction results. This process combines a variety of signal decomposition, processing and prediction methods, effectively giving play to the advantages of different decomposition methods, and can more accurately predict energy storage frequency modulation instructions, providing strong support for frequency control and revenue improvement of related power systems.

[0022] Example 2, reference Figure 1 , which is an embodiment of the present invention, provides a method for predicting energy storage frequency modulation instructions with complementary advantages based on the previous embodiment, including: In this embodiment, in the above step S1, the frequency modulation instruction to be predicted is subjected to VMD decomposition and CEEMDAN decomposition to obtain a first subsequence set and a second subsequence set, respectively, including: Assume the frequency modulation instruction is P t , after VMD decomposition, it becomes IMF1,IMF2,IMF3,IMF i ,...,IMF K , which is the first subsequence set; the same frequency modulation instruction becomes IMF1 after CEEMDAN decomposition 、 ,IMF2、 ,IMF3 、 ,IMF i ,...,IMF K 、 , that is, the second subsequence set.

[0023] In this embodiment, the performing sequence interleaving based on the first subsequence set and the second subsequence set in the above step S2 includes: Suppose a subsequence in the first subsequence set, that is, a subsequence decomposed by VMD, is IMF i =[X i1 ,X i2 ,X i3 ,...,X ii ,...,X iN ], let the experimental group error of the sequence be W i .

[0024] Suppose a subsequence in the second subsequence set, that is, a subsequence IMF decomposed by CEEMDAN j 、 =[X j1 、 ,X j2 、 ,X j3 、 ,...,X ji 、 ,...,X jN 、 ], let the experimental group error of the sequence be W j 、 .

[0025] Interleave the two subsequences, expressed as [X i1 ,X j1 、 , X i2 ,X j2 、 , X i3 ,X j3 、 ,...,X ii ,X ji 、 ,...,X iN ,X jN 、 ], the experimental group error generated is W ij , W ij =(W i +W j 、 ) / 2.

[0026] It should be noted that, assuming the frequency modulation instruction is P t=[X1,X2,X3,...,X i ,...,X N ], N is the length of the initial frequency modulation instruction sequence, and the first 0.9N length series [X1, X2, X3, ..., X 0.9N ] as input to predict the experimental group [X 0.9N+1 ,X 0.9N+2 ,X 0.9N+3 ,...,X N ] and the unknown group [X N+1 ,X N+2 ,X N+3 ,...,X N+0.1N ] value. Among them, the actual value and predicted value of the test group generate all the errors (MAPE) of this embodiment. What this embodiment actually predicts is the value of the unknown group. MAPE is the average value of the absolute error, expressed as: ; In this embodiment, the first judgment in step S2 includes: like and , then the two subsequences that are interlaced are considered to match each other, where avg() means taking the average value.

[0027] It should be noted that based on a detailed analysis of sequence characteristics and error indicators, complementary subsequence pairs can be accurately found.

[0028] In this embodiment, in the above step S3, based on the result of the first judgment, exchanging sequence fragments of all mutually matching subsequences includes: The mutual matching refers to that the complete subsequences match each other.

[0029] Suppose any pair of matching subsequences is: VMD decomposed subsequence IMF i =[X i1 ,X i2 ,X i3 ,...,X ii ,...,X iN ] and CEEMDAN decomposition of subsequence IMF j 、 =[X j1 、 ,X j2 、 ,X j3 、 ,X ji 、 ,...,X jN 、 ], and exchange the sequence fragments of length Q for this pair of subsequences.

[0030] The random range of the length of Q is expressed as: , where N is the length of the subsequence to be exchanged, and [ ] means rounding down.

[0031] For example, let the subsequence IMF of VMD decomposition be i =[1,2,3,4,...,9], subsequence IMF decomposed by CEEMDAN j 、 =[10,11,12,13,...,18]; Randomly determine Q=5, then the result of the first exchange is IMF i =[10,11,12,13,14,6,7,8,9],IMF j 、 =[1,2,3,4,5,15,16,17,18]; Calculate the test group errors W1 and W1 respectively. 、 and the corresponding error mean.

[0032] The result of the second exchange was that the IMF i =[1,11,12,13,14,15,7,8,9],IMF j 、 =[10,2,3,4,5,6,16,17,18]; Calculate the test group error and the corresponding error average respectively.

[0033] And so on, exhausting all possibilities; It should be noted that for each pair of mutually matching subsequences, when the length Q of the exchange sequence fragment is determined, fragments of length Q are selected from different starting positions in the subsequence for exchange operations. After each exchange operation of selecting fragments at different starting positions, the corresponding error average is calculated, and the exchange with the smallest error average is taken as the final result of the sequence fragment exchange.

[0034] It should be noted that the sequence fragments of the matched subsequences are exchanged so that the subsequences generated by different decomposition methods can learn from each other's advantageous information, further optimize the characteristic expression of the subsequences, reduce the errors caused by the limitations of a single decomposition method, and provide high-quality data input for the subsequent prediction process.

[0035] In this embodiment, in the above step S4, the subsequence after the sequence fragment exchange is predicted by the GRU network, and the prediction result is corrected, including: All the subsequences after sequence fragment exchange, i.e. DNA fragment exchange, are put into the GRU network for prediction. The prediction result sequence of the prediction group is C1, C2, C3, ..., C K and C1 、 , C2 、 ,C3 、 ,...,C K 、 .

[0036] Suppose the predicted result of a sequence after DNA exchange is C r =[X N+1 ,X N+2 ,X N+3 ,...,X N+i ,...,X N+0.1N ], and correct the corresponding result to: ,in, is the length of the DNA fragment corresponding to this sequence.

[0037] The final energy storage frequency modulation command prediction result is expressed as 0.5 (C1+C2+C3+...+C K +C1 、 +C2 、 +C3 、 +...+C K 、 ).

[0038] It should be noted that the GRU network is used to predict the subsequence after the sequence fragments are exchanged. As an advanced recurrent neural network, the GRU network has powerful time series data processing capabilities and can capture the complex dynamic changes in the frequency modulation instruction sequence. In the correction link of the prediction results, the characteristics of the subsequence and the influence of the exchange operation are fully considered to further eliminate possible errors, so that the prediction results are more accurately close to the actual energy storage frequency modulation instructions.

[0039] Embodiment 3, the above is a schematic scheme of the energy storage frequency modulation instruction prediction method with complementary advantages in this embodiment. It should be noted that the technical scheme of the energy storage frequency modulation instruction prediction system with complementary advantages and the technical scheme of the energy storage frequency modulation instruction prediction method with complementary advantages belong to the same concept, and the details of the technical scheme of the energy storage frequency modulation instruction prediction system with complementary advantages in this embodiment that are not described in detail can all be referred to the description of the technical scheme of the energy storage frequency modulation instruction prediction method with complementary advantages.

[0040] This embodiment also provides a system for predicting energy storage frequency modulation instructions based on complementary advantages, including: An initial decomposition module, used for performing VMD decomposition and CEEMDAN decomposition on the frequency modulation instruction to be predicted, to obtain a first subsequence set and a second subsequence set respectively; A matching judgment module, used for performing sequence interleaving based on the first subsequence set and the second subsequence set, and performing a first judgment; The performing sequence interleaving based on the first subsequence set and the second subsequence set comprises: Suppose a subsequence in the first subsequence set, that is, a subsequence decomposed by VMD, is IMF i =[X i1 ,X i2 ,X i3 ,...,X ii ,...,X iN ], let the experimental group error of the sequence be W i ; Suppose a subsequence in the second subsequence set, that is, a subsequence IMF decomposed by CEEMDAN j 、 =[X j1 、 ,X j2 、 ,X j3 、 ,...,X ji 、 ,...,X jN 、 ], let the experimental group error of the sequence be W j 、 ; Interleave the two subsequences, expressed as [X i1 ,X j1 、 ,X i2 ,X j2 、 , X i3 ,X j3 、 ,...,X ii ,X ji 、 ,...,X iN ,X jN 、 ], the experimental group error generated is W ij , W ij =(W i +W j 、 ) / 2; The first determination comprises: The test group error generated by the two subsequences interlaced with each other is judged. and , then the two subsequences interlaced are considered to match each other, where avg() means taking the average value; A fragment exchange module, used for performing sequence fragment exchange on all mutually matching subsequences based on the result of the first judgment; for unmatched subsequences, no sequence fragment exchange operation is performed; the mutual matching refers to the mutual matching of complete subsequences; The step of performing sequence segment exchange on all mutually matching subsequences based on the result of the first judgment includes: Suppose any pair of matching subsequences is: VMD decomposed subsequence IMF i =[X i1 ,X i2 ,X i3 ,...,X ii ,...,X iN ] and CEEMDAN decomposition of subsequence IMF j 、 =[X j1 、 ,X j2 、 ,X j3 、 ,...,X ji 、 ,...,X jN 、 ], exchange the sequence fragments of length Q for this pair of subsequences; The step of exchanging sequence segments having a length of Q comprises: The random range of the length of Q is expressed as: , where N is the length of the subsequence to be exchanged, and [ ] means rounding down; The step of performing sequence segment exchange on all mutually matching subsequences based on the result of the first judgment further includes: For each pair of mutually matching subsequences, when the length Q of the exchange sequence fragment is determined, fragments of length Q are selected from different starting positions in the subsequence for exchange operation; after each exchange operation of selecting fragments at different starting positions, the corresponding average error is calculated, and the exchange with the smallest average error is taken as the final result of sequence fragment exchange; The prediction module is used to input all subsequences that have undergone sequence segment exchange operations into the GRU network for prediction, and to correct the prediction results to obtain the final energy storage frequency modulation instruction prediction results.

[0041] This embodiment further provides a computing device, which is applicable to the case of a method for predicting energy storage frequency modulation instructions with complementary advantages, including: Memory and processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the energy storage frequency modulation instruction prediction method with complementary advantages as proposed in the above embodiment.

[0042] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the energy storage frequency modulation instruction prediction method with complementary advantages proposed in the above embodiment is implemented.

[0043] The storage medium proposed in this embodiment and the energy storage frequency modulation instruction prediction method with complementary advantages proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0044] Example 4, referring to Table 1-Table 2, is an embodiment of the present invention, which provides a method for predicting energy storage frequency modulation instructions with complementary advantages. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0045] The present invention respectively adopts the method of the present invention, the prediction method of VMD-GRU and the prediction method of CEEMDAN-GRU to predict the frequency modulation sequence, and the results are as follows.

[0046] Table 1 Comparison of evaluation indicators

[0047] Table 2 Four evaluation indicators

[0048] Where N represents the sample size, and Represent the actual value and predicted value at time n respectively.

[0049] It can be seen that when predicting the frequency modulation sequence, the complementary energy storage frequency modulation instruction prediction method proposed in the present invention is significantly better than the prediction methods of VMD-GRU and CEEMDAN-GRU in various evaluation indicators. Specifically, in terms of MAE (mean absolute prediction error), SSE (intra-group sum of squares or residual sum of squares), RMSE (root mean square prediction error) and MAPE (average value of absolute error), the values ​​obtained by the method of the present invention are much lower than those of the other two comparison methods. This fully demonstrates that the method of the present invention can greatly reduce the prediction error, significantly improve the accuracy and reliability of the prediction, and more accurately approach the actual frequency modulation instruction situation, thereby providing a more accurate and effective basis for the frequency modulation control of the power system, thereby improving the revenue of the power plant and the stability of the power system, which strongly verifies that the method of the present invention has obvious advantages and good application effects.

[0050] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for predicting energy storage frequency modulation instructions with complementary advantages, characterized in that: include: Perform VMD decomposition and CEEMDAN decomposition on the frequency modulation instructions to be predicted to obtain a first subsequence set and a second subsequence set respectively; Interleave the sequence based on the first subsequence set and the second subsequence set, and make a first judgment; The performing sequence interleaving based on the first subsequence set and the second subsequence set comprises: Suppose a subsequence in the first subsequence set, that is, a subsequence decomposed by VMD, is IMF i =[X i1 ,X i2 ,X i3 ,...,X ii ,...,X iN ], let the experimental group error of the sequence be W i ; Suppose a subsequence in the second subsequence set, that is, a subsequence IMF decomposed by CEEMDAN j 、 =[X j1 、 ,X j2 、 ,X j3 、 ,...,X ji 、 ,...,X jN 、 ], let the experimental group error of the sequence be W j 、 ; Interleave the two subsequences, expressed as [X i1 ,X j1 、 ,X i2 ,X j2 、 , X i3 ,X j3 、 ,...,X ii ,X ji 、 ,...,X iN ,X jN 、 ], the experimental group error generated is W ij , W ij =(W i +W j 、 ) / 2; The first determination comprises: The test group error generated by the two subsequences interlaced with each other is judged. and , then the two subsequences interlaced are considered to match each other, where avg() means taking the average value; Based on the result of the first judgment, sequence fragments are exchanged for all mutually matching subsequences; for unmatched subsequences, no sequence fragment exchange operation is performed; the mutual matching refers to the mutual matching of complete subsequences; The step of performing sequence segment exchange on all mutually matching subsequences based on the result of the first judgment includes: Suppose any pair of matching subsequences is: VMD decomposed subsequence IMF i =[X i1 ,X i2 ,X i3 ,...,X ii ,...,X iN ] and CEEMDAN decomposition of subsequence IMF j 、 =[X j1 、 ,X j2 、 ,X j3 、 ,...,X ji 、 ,...,X jN 、 ], exchange the sequence fragments of length Q for this pair of subsequences; The step of exchanging sequence segments having a length of Q comprises: The random range of the length of Q is expressed as: , where N is the length of the subsequence to be exchanged, and [ ] means rounding down; The step of performing sequence segment exchange on all mutually matching subsequences based on the result of the first judgment further includes: For each pair of mutually matching subsequences, when the length Q of the exchange sequence fragment is determined, fragments of length Q are selected from different starting positions in the subsequence for exchange operation; after each exchange operation of selecting fragments at different starting positions, the corresponding average error is calculated, and the exchange with the smallest average error is taken as the final result of sequence fragment exchange; All subsequences that have undergone sequence segment exchange operations are input into the GRU network for prediction, and the prediction results are corrected to obtain the final energy storage frequency modulation instruction prediction results.

2. A method for predicting energy storage frequency modulation instructions with complementary advantages as claimed in claim 1, characterized in that: The step of inputting all subsequences that have undergone the sequence segment exchange operation into the GRU network for prediction and correcting the prediction results includes: The subsequences after sequence fragment exchange are predicted through the GRU network, and the prediction result sequence of the obtained prediction group is assumed to be C1, C2, C3, ..., C K and C1 、 , C2 、 ,C3 、 ,...,C K 、 .

3. A method for predicting energy storage frequency modulation instructions with complementary advantages as claimed in claim 2, characterized in that: The step of inputting all subsequences that have undergone the sequence segment exchange operation into the GRU network for prediction and correcting the prediction results also includes: Suppose the prediction result of a sequence after sequence segment exchange is: C r =[X N+1 ,X N+2 ,X N+3 ,...,X N+i ,...,X N+0.1N ]; Correct the corresponding results to: ,in, It is the length of the DNA fragment when the corresponding sequence undergoes fragment exchange.

4. The energy storage frequency modulation instruction prediction method with complementary advantages as claimed in claim 3 is characterized in that: The step of inputting all subsequences that have undergone the sequence segment exchange operation into the GRU network for prediction and correcting the prediction results also includes: The final energy storage frequency modulation command prediction result is expressed as 0.5 (C1+C2+C3+...+C K +C1 、 +C2 、 +C3 、 +...+C K 、 ).

5. A complementary energy storage frequency modulation instruction prediction system, using any method as claimed in claim 1 to claim 4, characterized in that: include: An initial decomposition module, used for performing VMD decomposition and CEEMDAN decomposition on the frequency modulation instruction to be predicted, to obtain a first subsequence set and a second subsequence set respectively; A matching judgment module, used for performing sequence interleaving based on the first subsequence set and the second subsequence set, and performing a first judgment; The performing sequence interleaving based on the first subsequence set and the second subsequence set comprises: Suppose a subsequence in the first subsequence set, that is, a subsequence decomposed by VMD, is IMF i =[X i1 ,X i2 ,X i3 ,...,X ii ,...,X iN ], let the experimental group error of the sequence be W i ; Suppose a subsequence in the second subsequence set, that is, a subsequence IMF decomposed by CEEMDAN j 、 =[X j1 、 ,X j2 、 ,X j3 、 ,...,X ji 、 ,...,X jN 、 ], let the experimental group error of the sequence be W j 、 ; Interleave the two subsequences, expressed as [X i1 ,X j1 、 ,X i2 ,X j2 、 , X i3 ,X j3 、 ,...,X ii ,X ji 、 ,...,X iN ,X jN 、 ], the experimental group error generated is W ij , W ij =(W i +W j 、 ) / 2; The first determination comprises: The test group error generated by the two subsequences interlaced with each other is judged. and , then the two subsequences interlaced are considered to match each other, where avg() means taking the average value; A fragment exchange module, used for performing sequence fragment exchange on all mutually matching subsequences based on the result of the first judgment; for unmatched subsequences, no sequence fragment exchange operation is performed; the mutual matching refers to the mutual matching of complete subsequences; The step of performing sequence segment exchange on all mutually matching subsequences based on the result of the first judgment includes: Suppose any pair of matching subsequences is: VMD decomposed subsequence IMF i =[X i1 ,X i2 ,X i3 ,...,X ii ,...,X iN ] and CEEMDAN decomposition of subsequence IMF j 、 =[X j1 、 ,X j2 、 ,X j3 、 ,...,X ji 、 ,...,X jN 、 ], exchange the sequence fragments of length Q for this pair of subsequences; The step of exchanging sequence segments having a length of Q comprises: The random range of the length of Q is expressed as: , where N is the length of the subsequence to be exchanged, and [ ] means rounding down; The step of performing sequence segment exchange on all mutually matching subsequences based on the result of the first judgment further includes: For each pair of mutually matching subsequences, when the length Q of the exchange sequence fragment is determined, fragments of length Q are selected from different starting positions in the subsequence for exchange operation; after each exchange operation of selecting fragments at different starting positions, the corresponding average error is calculated, and the exchange with the smallest average error is taken as the final result of sequence fragment exchange; The prediction module is used to input all subsequences that have undergone sequence segment exchange operations into the GRU network for prediction, and to correct the prediction results to obtain the final energy storage frequency modulation instruction prediction results.

6. A computing device comprising: Memory and processor; The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the method described in any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions, when executed by a processor, implement the steps of the method according to any one of claims 1 to 4.

Images