Search

CN-122026437-A - Energy storage system control method based on multi-scale feature fusion and dynamic adjustment

CN122026437ACN 122026437 ACN122026437 ACN 122026437ACN-122026437-A

Abstract

The invention relates to an energy storage system control method based on multi-scale feature fusion and dynamic adjustment, and belongs to the technical field of energy storage regulation and control. The method comprises the steps of obtaining historical new energy power generation power and power load data, generating a net load sequence, carrying out zero-mean normalization, optimizing the number of wavelet decomposition layers based on mutual information entropy, adaptively dividing the normalized net load sequence into a high-frequency component set and a low-frequency component set, predicting the high-frequency component by adopting a weighted moving average method, predicting the low-frequency component by adopting linear regression, finally fusing prediction results of the high-frequency component and the low-frequency component, and calculating the output value of the energy storage system at the next moment. According to the invention, through data preprocessing, wavelet decomposition layer number optimization and multi-scale prediction fusion, dynamic accurate adjustment of the output of the energy storage system is realized, and the adaptability of the energy storage system to new energy fluctuation and load change is improved.

Inventors

  • LI GUOQING
  • ZHENG HUA
  • SUN YIQIAN
  • XIE LI
  • XU BO
  • SHI WEI
  • ZHANG BIN
  • LIU DAGUI
  • DING BIWEI

Assignees

  • 国网新疆电力有限公司
  • 华北电力大学

Dates

Publication Date
20260512
Application Date
20251202

Claims (8)

  1. 1. The energy storage system control method based on multi-scale feature fusion and dynamic adjustment is characterized by comprising the following steps of: Step 1, acquiring historical new energy power generation power and power load data, and obtaining a net load sequence; Step 2, data preprocessing and outlier correction, and zero-mean normalization is carried out on the net load sequence; step 3, optimizing the number of layers of the payload sequence after zero-mean normalization based on wavelet decomposition of mutual information entropy; Step 4, dividing a high-frequency component set and a low-frequency component set after wavelet decomposition according to the obtained layer number; Step 5, predicting the high-frequency component by adopting a weighted moving average method; step 6, predicting the low-frequency component by adopting linear regression; and 7, fusing the high-frequency and low-frequency component prediction results, and calculating the output value of the energy storage system at the next moment.
  2. 2. The method for controlling the energy storage system based on multi-scale feature fusion and dynamic adjustment according to claim 1, wherein the specific implementation method of step 1 is that a historical new energy generation power sequence N (T) (t=1, 2,3.., T all ,T all is the total length of historical data) and power load data L (T) are obtained, and then the net load sequence P (T) is: 。
  3. 3. The method for controlling the energy storage system based on multi-scale feature fusion and dynamic adjustment according to claim 1, wherein the specific implementation method of the step 2 is that zero-mean normalization is performed on a net load sequence, dimensional influence is eliminated, and the net load sequence after zero-mean normalization is P norm (t): ; Where u P is the arithmetic mean of the payload sequences and σ P is the standard deviation of the payload sequences.
  4. 4. The method for controlling the energy storage system based on multi-scale feature fusion and dynamic adjustment according to claim 1, wherein the specific implementation method of the step 3 is that db4 wavelet basis is selected, the initial decomposition layer number N=3 is set, wavelet decomposition is performed on a standardized sequence P norm (t), and an approximate component A N low-frequency trend and a detail component D 1 ,D 2 ,...,D N are obtained, so that P norm (t)=A N +D 1 +D 2 +...+D N is satisfied; For adjacent detail components D k and D k+1 , the mutual information entropy E (D k , D k+1 ) is calculated: ; And if E (D k , D k+1 ) is larger than a mutual information entropy threshold epsilon to represent the scale aliasing degree, carrying out repeated decomposition and calculation until E of all adjacent components is smaller than or equal to the mutual information entropy threshold epsilon to obtain the optimal layer number N * .
  5. 5. The method for controlling the energy storage system based on multi-scale feature fusion and dynamic adjustment according to claim 1, wherein the specific implementation method in the step 4 is that a high-frequency component set and a low-frequency component set after wavelet decomposition are divided according to the obtained layer number, wherein the high-frequency component set H comprises D 1 to D K , K is an integer less than or equal to N x/2 and is used for capturing short-term fluctuation features, the low-frequency component set L comprises D K+1 to D N* and A N* and is used for capturing long-term trend features, and finally the high-frequency component H (t) =D 1 +D 2 +...+D K and the low-frequency component L (t) =A N* +D K +D K+1 +...+D N* are obtained.
  6. 6. The method for controlling the energy storage system based on multi-scale feature fusion and dynamic adjustment according to claim 1, wherein the specific implementation method for predicting the high-frequency component H pred (t+1) by adopting a weighted moving average method is as follows: ; Wherein w ti represents the weight of the ti-th historical data, and the more recent data weight is obtained by fitting the initial weight value through the historical data.
  7. 7. The method for controlling the energy storage system based on multi-scale feature fusion and dynamic adjustment according to claim 1, wherein the specific implementation method for predicting the low-frequency component L pred (t+1) by adopting linear regression in the step 6 is as follows: ; Wherein θ 0 、θ 1 is the fitting coefficient of the linear regression.
  8. 8. The method for controlling the energy storage system based on multi-scale feature fusion and dynamic adjustment according to claim 1, wherein the step 7 fuses the prediction results of the high-frequency component and the low-frequency component, and calculates the output value P final (t+1) of the energy storage system in the power system at the next moment, the method is as follows: When H pred (t+1)+L pred (t+1) is not less than 0: ; When H pred (t+1)+L pred (t+1) < 0: ; Wherein P rate is the rated power of the energy storage system.

Description

Energy storage system control method based on multi-scale feature fusion and dynamic adjustment Technical Field The invention belongs to the technical field of energy storage regulation and control, and particularly relates to an energy storage system control method based on multi-scale feature fusion and dynamic regulation. Background Along with the acceleration of the global energy transformation process, the permeability of renewable energy sources such as wind energy, solar energy and the like in an electric power system presents a rapid growth situation. However, this large scale access to green energy also presents new technical challenges in that new energy generation has significant intermittent, stochastic and fluctuating features, with output variations spanning multiple time scales, regular variations from instantaneous fluctuations in the seconds, minutes to the hours, days, and even seasonal long-term trends. This multi-time scale feature makes the payload sequence of the power system extremely complex, and conventional single time scale based prediction and control methods are difficult to effectively cope with. In particular, high frequency fluctuations may be caused by sudden weather changes (e.g., cloud cover, gust changes), while low frequency trends are related to macroscopic factors such as season changes, climate patterns, etc. The energy storage system is used as a key regulating means for balancing power generation and power consumption and maintaining stable operation of a power grid, and the accuracy of a control strategy of the energy storage system is directly related to the safety and the economy of the whole power system. The current widely applied energy storage control method, such as a charge-discharge strategy based on a fixed threshold rule or a simple moving average prediction model, is often not fully suitable for the multi-scale and non-stable load characteristics, and easily causes the problems of control lag or excessive adjustment and the like. Therefore, there is a need to develop a novel energy storage control method capable of intelligently identifying and fusing multi-scale features. The technical scheme in the current power system energy storage control field can be divided into the following (1) traditional time sequence prediction methods, such as an autoregressive integral moving average (ARIMA) model and an exponential smoothing method, and has the main advantages of simple structure, high calculation efficiency and easiness in engineering realization. However, such methods are inherently based on linear assumptions, and have limited ability to model new energy output sequences that exhibit strong nonlinear, non-stationary characteristics, especially with difficulty in effectively capturing bursty fluctuations and multi-time scale features. (2) The control strategy based on rules, such as fixed threshold charge-discharge control, is simple to implement and quick in response, but lacks prospective judgment on the running state of the system, and is easy to cause poor control effect or frequent actions of equipment under complex running conditions. (3) With economical efficiency or technical performance as a target, an intelligent algorithm is adopted to solve based on constraint conditions, such as a neural network, a support vector machine and the like, and the intelligent algorithm has strong nonlinear fitting capability, but usually requires a large amount of training data, and has the problems of poor model interpretation, complex parameter adjustment and the like. Disclosure of Invention The invention aims to overcome the defects of the prior art and provides an energy storage system control method based on multi-scale feature fusion and dynamic adjustment, wherein a standardized net load sequence is obtained after historical new energy power generation and load data are subjected to data preprocessing, the number of wavelet decomposition layers is optimized based on mutual information entropy, the net load sequence is adaptively divided into high-frequency and low-frequency components, the high-frequency and low-frequency components are respectively predicted by adopting weighted moving average and linear regression, and finally the output of the energy storage system is dynamically adjusted by fusing high-frequency and low-frequency prediction results. The invention solves the technical problems by adopting the following technical scheme: an energy storage system control method based on multi-scale feature fusion and dynamic adjustment comprises the following steps: Step 1, acquiring historical new energy power generation power and power load data, and obtaining a net load sequence; Step 2, data preprocessing and outlier correction, and zero-mean normalization is carried out on the net load sequence; step 3, optimizing the number of layers of the payload sequence after zero-mean normalization based on wavelet decomposition of mutual information entropy;