Energy efficiency management and optimization strategies for power grid enterprises based on deep learning基于深度学习的电网企业能源效率管理与优化策略

英文 Abstract

Dispersed energy sources with highly variable output present significant challenges to power grid operators and infrastructure. The dynamic and complex structure of modern power networks makes it difficult to manage and improve energy efficiency, especially in the era of renewable integration. Imbalances in energy supply and demand can lead to system instability, including voltage drops, spikes, and outages. Additionally, ongoing power system reforms have increased the complexity of forecasting grid development and operational requirements. To address these challenges, intelligent energy management and optimization systems are essential for preventing bottlenecks and maintaining supply–demand equilibrium. Regardless of climatic variations, such systems must deliver reliable, cost-effective energy services while enabling energy sharing across the grid. This study introduces a deep learning-based framework, the spatiotemporal adaptive energy optimization network (SAEON), designed to enhance real-time energy management. SAEON integrates graph neural networks and long short-term memory to model both spatial and temporal dependencies in grid data. It includes three core components: A spatiotemporal data collection module that captures temporal trends and spatial relationships; an optimization module using deep reinforcement learning to determine optimal energy load allocation strategies; and an adaptive feedback mechanism that continuously fine-tunes model parameters during operation. Experimental results demonstrate that SAEON can significantly reduce energy costs, improve efficiency, minimize losses, and enhance load balancing. The proposed model offers a reliable and intelligent solution for energy optimization in power-grid-dependent environments.

中文翻译

输出高度波动的分布式能源给电网运营与基础设施带来了显著挑战。现代电力网络结构动态且复杂，特别是在可再生能源加速接入的背景下，能源效率管理与提升更加困难。供需失衡可能引发电压跌落、电压尖峰甚至停电；持续推进的电力系统改革也增加了电网发展和运行需求预测的复杂度。为解决这些问题，需要智能能源管理与优化系统来避免瓶颈并维持供需平衡，同时在不同气候条件下提供可靠、经济的能源服务并支持电网内能源共享。本文提出时空自适应能源优化网络 SAEON，用于提升实时能源管理能力。该框架结合图神经网络与长短期记忆网络，对电网数据的空间和时间依赖进行建模，包含时空数据采集、基于深度强化学习的负荷分配优化，以及运行中持续微调参数的自适应反馈三个核心模块。实验结果表明，SAEON 能降低能源成本和损耗、提升效率并改善负荷平衡，为依赖电网的应用环境提供可靠的智能能源优化方案。

中文概述

本文提出面向电网企业能源效率管理的时空自适应优化网络 SAEON。该框架结合图神经网络与长短期记忆网络刻画电网数据的空间和时间依赖，并利用深度强化学习优化负荷分配。原文实验表明，该方法能够降低能源成本与损耗、改善负荷平衡，并通过自适应反馈机制持续调整模型参数。

内容分类

原文与代码