Brain-like autonomous control architecture of microgrid based on co-evolution of prior model and agent

Engineering notes

Engineering notes will be added by the Power for Solar editorial team.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为光伏效率、钙钛矿太阳能电池、储能技术、太阳能热利用、BIPV、并网技术等高价值论文补充中文说明。

Original abstract

A microgrid is the core carrier of efficient utilization of distributed energy. Brain-like autonomous regulation technology has been gradually applied in the field of Microgrid operation management and control. However, after 2023, relevant research shows that there are some problems in the existing regulation model, such as a generally higher than 0.5s response delay of autonomous decision-making, insufficient bionic accuracy of brain-like regulation, and unsatisfactory cooperation of agents. As a result, it is difficult to adapt to the complex working conditions after the access of distributed generation with high penetration. To deal with the above problems, this study integrates Prior Knowledge Embedding + Multi-Agent Reinforcement Learning (PKE + MARL) into the brain-like hierarchical regulatory architecture. Based on Long Short-Term Memory and Convolutional Neural Network (LSTM-CNN) co-evolution, a Microgrid brain-like autonomous regulation technology system is constructed. The innovation lies in the design of dynamic calibration and co-evolution coupling strategy, the introduction of random interference adaptive factor, the construction of two-way feedback mechanism between PKE and agent decision-making, and the optimization of brain-like hierarchical adaptation mechanism. This technology can realize the cooperative scheduling and fault autonomous response of Microgrid source load storage, effectively improve the accuracy and efficiency of regulation and control, and solve the pain points of the existing technology. In the simulation experiment, the digital twin simulation data set of the distribution network is used to test. Our findings indicate that the comprehensive performance score reaches 91.8, which is 4.9 points higher than the existing optimal combination technology. Notably, the regulation accuracy is 0.947, and the average robustness coefficient reaches up to 0.864. Besides, the practical adaptability is 0.896, along with a stable 95% confidence interval. Furthermore, the core modules have no redundancy and a clear contribution. This study provides a complete theoretical and technical support for Microgrid self-regulation, and promotes the efficient utilization of distributed energy and improves the stability of system operation, which has strong engineering application value.

Links and sources

• PDF from original source

Comments