Photovoltaic power prediction method based on combination of inverted Transform and weighted decomposition

By combining the inverted Transformer with weighted decomposition method, the accuracy problem of existing photovoltaic power prediction methods when capturing nonlinear and volatility characteristics is solved, and more efficient photovoltaic power prediction is achieved, improving the accuracy and computing efficiency of the model.

CN120277356AActive Publication Date: 2025-07-08LANZHOU UNIVERSITY OF TECHNOLOGY
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202510337486.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-03-21
Publication Date
2025-07-08
Estimated Expiration
2045-03-21

AI Technical Summary

Technical Problem

The existing neural network-based photovoltaic power prediction methods are prone to fall into local extreme values when processing photovoltaic power generation data, resulting in inaccurate predictions and difficult to effectively capture the complex nonlinear and volatility characteristics of photovoltaic power generation data.

Method used

The inverted Transformer combined with weighted decomposition method is used to capture the complex nonlinear and volatility characteristics of photovoltaic power data through inverted Transformer architecture and probabilistic sparse attention mechanism, and the data is decomposed into trend, seasonal and random components through the weighted series decomposition module, and the attention coefficient is adjusted in combination with the sparse attention mechanism to reduce the computational complexity.

Benefits of technology

It improves the accuracy and computing efficiency of photovoltaic power prediction, can capture the correlation between multivariables more naturally, reduce the computational complexity, and improve the prediction performance of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120277356A_ABST
    Figure CN120277356A_ABST
Patent Text Reader

Abstract

The invention discloses a photovoltaic power prediction method based on combination of inverted Transform and weighted decomposition. An inverted Transform framework, a weighted series decomposition module and a probability sparse attention mechanism are included. The method comprises the following steps of: 1, capturing complex non-linear and volatility characteristics of photovoltaic power data by an inverted Transform architecture so as to improve the learning ability of a model to multivariate correlation; step 2, a weighted series decomposition module is used for extracting time characteristics and periodicity of photovoltaic data and introducing a weight conforming to an actual photovoltaic power generation output power change mode to constrain an output result of the model; 3, a probability sparse attention mechanism is combined with an inverted architecture, so that the prediction method can maintain relatively high calculation efficiency when processing long-sequence data, and the model performance is enhanced; step 4, constructing a two-stage structure for respectively processing different characteristics of the photovoltaic data so as to improve the prediction accuracy, robustness and calculation efficiency; and step 5, photovoltaic power prediction is carried out through the prediction layer.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Short-term photovoltaic power generation power prediction method based on L-Transform

    CN117034055A

  • Power load prediction method and system

    CN119026763A

  • Air target trajectory prediction system based on time sequence decomposition and reconstruction model

    CN119600854A

  • Non-Intrusive Load Decomposition Method Based on Informer Model Coding Structure

    US20220397874A1