Hybrid energy storage layered response method and system based on multi-modal space-time fusion network

By combining one-dimensional convolutional neural networks, long short-term memory networks, and Transformer models, along with dynamic time warping and cross-modal attention mechanisms, the problem of time axis offset between grid load and photovoltaic data is solved, achieving high-precision prediction and full life-cycle cost optimization, and improving the collaborative scheduling capability of hybrid energy storage systems.

CN120914853APending Publication Date: 2025-11-07STATE GRID HUNAN ELECTRIC POWER COMPANY LIMITED +2
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202511056729.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-07-30
Publication Date
2025-11-07

AI Technical Summary

Technical Problem

Traditional prediction models struggle to capture the short-term fluctuations and long-term cyclical patterns of grid load and photovoltaic data, leading to time axis shifts during cross-modal feature fusion and increasing prediction errors. Existing energy storage control strategies do not fully consider battery life degradation and total lifecycle costs. Independent modeling methods cannot capture the complex spatiotemporal relationships between load, photovoltaics, and energy storage, thus limiting the potential for multi-energy synergistic optimization.

Method used

We employ a one-dimensional convolutional neural network and a long short-term memory network to extract short-term time-series features, and combine them with a Transformer model to extract long-term time-series features. We then use a dynamic time warping algorithm and a cross-modal attention mechanism to perform time series alignment and feature fusion, construct a hybrid energy storage hierarchical response strategy, and optimize energy storage scheduling to maximize the benefits throughout the entire life cycle.

Benefits of technology

It significantly improves the cross-scale prediction accuracy of grid load and photovoltaic power generation, realizes real-time economic optimization of energy storage systems, enhances the multi-energy coordinated dispatch capability, and reduces operating costs and prediction errors.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120914853A_ABST
    Figure CN120914853A_ABST
Patent Text Reader

Abstract

The invention discloses a hybrid energy storage hierarchical response method and system based on a multi-modal space-time fusion network, and the method comprises the steps: extracting short-term time sequence features of short-term data through employing a 1D CNN + LSTM, and extracting long-term time sequence features of long-term historical data through employing Transform; aligning time sequence characteristics with different period lengths; fusing the aligned short-term time sequence characteristics and long-term time sequence characteristics by adopting a dynamic weighting strategy based on a cross-modal attention mechanism; performing power grid load demand prediction and photovoltaic power generation power real-time prediction on the hybrid energy storage system; a real-time power gap is calculated, marginal cost of each energy storage unit in the hybrid energy storage system is quantified in real time, a hybrid energy storage layered response mechanism is established, an energy storage technology with the lowest marginal cost is preferentially called, and full-life-cycle revenue maximization is achieved.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Cited By

  • Power generation strategy adjustment method and device based on multi-modal electricity price prediction result, equipment and medium

    CN121120306A