Open Access

ARTICLE

A Physico-Data Hybrid Driven Approach to Digital Twin Modeling and Coordinated Optimization of Generation-Grid-Load-Storage Systems

Xiaowen Liang¹, Xiaomin Sun¹, Chunfang Liu², Hongyao Wu¹, Haonan Xie¹, Fengming Zhang², Linpo Zhong², Xiaojing Wang^3,*
1 Zhaoqing Power Supply Bureau of Guangdong Power Grid Co., Ltd., Zhaoqing, China
2 Foshan Electric Power Design Institute Co., Ltd., Foshan, China
3 College of Electrical Engineering, Zhejiang University, Hangzhou, China
* Corresponding Author: Xiaojing Wang. Email:
(This article belongs to the Special Issue: AI for Next Generation Flexible, Reliable, Resilient and Sustainable Energy Systems)

Energy Engineering https://doi.org/10.32604/ee.2026.079361

Received 20 January 2026; Accepted 17 April 2026; Published online 22 May 2026

Abstract

As the penetration of renewable energy in power systems continues to increase, traditional simulation and modeling methods encounter challenges, including low accuracy and high computational costs, when addressing multi-timescale dynamics and complex control. To overcome these challenges, a hybrid physics- and data-driven digital twin modeling approach for source-grid-load-storage systems is proposed. The method introduces a two-stage correction model for the digital twin: first, an initial mechanistic model is established based on factory data of the equipment; second, the control system structure and parameters are refined by comparing actual operational data. Additionally, an adaptive long short-term memory network is employed, integrating physical models and virtual data through a bidirectional feedback mechanism. The adaptive mechanism dynamically adjusts model parameters, constructing a hybrid physics- and data-driven digital twin model for the power system’s generation-grid-load-storage framework, thereby mitigating overfitting. Furthermore, an improved firefly algorithm is adopted to solve the cost-minimization collaborative operation problem of generation-grid-load-storage systems, enhancing the utilization rate of renewable energy. Simulation results demonstrate that the proposed method significantly improves the fitting capability and real-time updating performance of the digital twin model. The enhanced algorithm also strengthens the collaborative operation of generation-grid-load-storage systems, exhibiting superior accuracy and robustness compared to traditional methods.

---

Keywords

Generation-grid-load-storage; physics-data hybrid driving; improved firefly algorithm; digital twin; adaptive LSTM

---