Article: Simulation and evaluation of green power consumption policies driven by spatio-temporal graph convolutional networks Journal: International Journal of Information and Communication Technology (IJICT) 2025 Vol.26 No.45 pp.34 - 51 Abstract: Green power consumption has become a key challenge in the energy transition. Existing research struggles to capture complex relationships between the spatio-temporal dynamics of the power system and policy interventions. To this end, this paper first designs a power load forecasting model based on spatio-temporal graph convolutional networks. The model dynamically adjusts the graph structure according to users' electricity consumption patterns and introduces a weighted skip connection mechanism, assigning different weights to connections at different time steps. Then, a mathematical model for optimal combinations of power consumption policies is established. Through deep reinforcement learning algorithms interacting with the environment, it solves for the optimal combination of power consumption policies that minimise economic and carbon emission costs. Experimental outcome demonstrates that the proposed method achieves a green power consumption rate of 97.16%, outperforming comparison methods, thus helping to promote efficient green power consumption. Inderscience Publishers - linking academia, business and industry through research

Title: Simulation and evaluation of green power consumption policies driven by spatio-temporal graph convolutional networks

Authors: Jie Jiao; Jiyuan Zhang; Wenshi Ren

Addresses: Economic Research Institute, State Grid Sichuan, Chengdu, 610041, China ' Economic Research Institute, State Grid Sichuan, Chengdu, 610041, China ' Economic Research Institute, State Grid Sichuan, Chengdu, 610041, China

Abstract: Green power consumption has become a key challenge in the energy transition. Existing research struggles to capture complex relationships between the spatio-temporal dynamics of the power system and policy interventions. To this end, this paper first designs a power load forecasting model based on spatio-temporal graph convolutional networks. The model dynamically adjusts the graph structure according to users' electricity consumption patterns and introduces a weighted skip connection mechanism, assigning different weights to connections at different time steps. Then, a mathematical model for optimal combinations of power consumption policies is established. Through deep reinforcement learning algorithms interacting with the environment, it solves for the optimal combination of power consumption policies that minimise economic and carbon emission costs. Experimental outcome demonstrates that the proposed method achieves a green power consumption rate of 97.16%, outperforming comparison methods, thus helping to promote efficient green power consumption.

Keywords: green power; consumption policy; spatio-temporal graph convolutional network; deep reinforcement learning algorithms; skip connections.

DOI: 10.1504/IJICT.2025.150413

International Journal of Information and Communication Technology, 2025 Vol.26 No.45, pp.34 - 51

Received: 13 Aug 2025
Accepted: 27 Sep 2025
Published online: 12 Dec 2025 *

Title: Simulation and evaluation of green power consumption policies driven by spatio-temporal graph convolutional networks

Keep up-to-date