Multi-stage optimal allocation of power system resources using a combination of deep reinforcement learning and chance-constrained programming

Abstract With a high proportion of renewable energy integration, power system resource allocation faces the challenge of co-optimizing long-term dynamic uncertainties and probabilistic security constraints. Traditional deterministic programming and stochastic optimization methods struggle to balance economy, flexibility, and security. This paper constructs a multi-stage resource optimization allocation method that combines Deep Reinforcement Learning (DRL) and Chance Constrained Programming (CCP). To address the problem that resource allocation is difficult to adapt to the evolution of system state during long-term planning, a dynamic decision-making environment including new energy, energy storage, demand response, and network regulation resources is constructed. The Soft Actor-Critic (SAC) algorithm is used to learn the nonlinear mapping between state and configuration actions to achieve multi-stage in-distribution adaptability during rolling stages. For the problem of insufficient safety constraints in the configuration strategy, the CCP mechanism is applied to convert the new energy output error and load fluctuation into probabilistic safety constraints. Economic optimization and safety control coordination are achieved through risk feedback and Lagrange multiplier updates, and combined with the rolling deduction mechanism, the subsequent resource allocation path is continuously revised based on stage status feedback. The experiment constructs an equivalent test scenario based on the characteristics of new energy development in Jiangsu Province. The results show that, under the condition of high proportion of new energy access, the method in this paper reduces the total cost of the planning cycle by 8.5% compared with the second-best Twin Delayed Deep Deterministic Policy Gradient (TD3) method, increases the flexibility margin to 0.84, and keeps the frequency of constraint violations stably within 5%. Research shows that the method presented in this paper achieves comprehensive optimization of the economy, flexibility and security of resource allocation in uncertain new energy environments, providing an intelligent dynamic allocation decision-making method for long-term resource planning of power systems.

Authors

Institutions

Publication Details

Journal
Energy Informatics
Published
2026-11-14
DOI
https://doi.org/10.1186/s42162-026-00706-w
Primary Topic
Electric Power System Optimization
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Multi-stage optimal allocation of power system resources using a combination of deep reinforcement learning and chance-constrained programming

Xu Jianzheng, Bingjin Zhang, Yinghai Wang, Wenjuan Niu et al.
Energy Informatics
Electric Power System Optimization
article

Multi-stage optimal allocation of power system resources using a combination of deep reinforcement learning and chance-constrained programming

Xu Jianzheng, Bingjin Zhang, Yinghai Wang, Wenjuan Niu, Qingshan Wang, Yi Ge, Zheng Xu
article en

Abstract

Abstract With a high proportion of renewable energy integration, power system resource allocation faces the challenge of co-optimizing long-term dynamic uncertainties and probabilistic security constraints. Traditional deterministic programming and stochastic optimization methods struggle to balance economy, flexibility, and security. This paper constructs a multi-stage resource optimization allocation method that combines Deep Reinforcement Learning (DRL) and Chance Constrained Programming (CCP). To address the problem that resource allocation is difficult to adapt to the evolution of system state during long-term planning, a dynamic decision-making environment including new energy, energy storage, demand response, and network regulation resources is constructed. The Soft Actor-Critic (SAC) algorithm is used to learn the nonlinear mapping between state and configuration actions to achieve multi-stage in-distribution adaptability during rolling stages. For the problem of insufficient safety constraints in the configuration strategy, the CCP mechanism is applied to convert the new energy output error and load fluctuation into probabilistic safety constraints. Economic optimization and safety control coordination are achieved through risk feedback and Lagrange multiplier updates, and combined with the rolling deduction mechanism, the subsequent resource allocation path is continuously revised based on stage status feedback. The experiment constructs an equivalent test scenario based on the characteristics of new energy development in Jiangsu Province. The results show that, under the condition of high proportion of new energy access, the method in this paper reduces the total cost of the planning cycle by 8.5% compared with the second-best Twin Delayed Deep Deterministic Policy Gradient (TD3) method, increases the flexibility margin to 0.84, and keeps the frequency of constraint violations stably within 5%. Research shows that the method presented in this paper achieves comprehensive optimization of the economy, flexibility and security of resource allocation in uncertain new energy environments, providing an intelligent dynamic allocation decision-making method for long-term resource planning of power systems.

Energy Informatics
State Grid Jiangsu Electric Power (China) (CN)
Electric Power System Optimization
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon