# Data-Driven Based Method for Power System Time-Varying Composite Load Modeling

## TL;DR

This paper introduces a transfer reinforcement learning method for fast, accurate identification of power system load models, enhancing stability analysis and operational efficiency.

## Contribution

It presents a novel transfer reinforcement learning approach with imitation learning and associative memory for efficient load model identification.

## Key findings

- Method converges to global optimum with probability 1
- Achieves superior convergence rate and stability in simulations
- Validated on 68-bus system with multi-test cases

## Abstract

Fast and accurate load parameters identification has great impact on the power systems operation and stability analysis. This paper proposes a novel transfer reinforcement learning based method to identify composite ZIP and induction motor (IM) load models. An imitation learning process is firstly introduced to improve the exploitation and exploration process. The transfer learning process is then employed to overcome the challenge of time consuming optimization when dealing with new tasks. An Associative memory is designed to realize demension reduction and knowledge learning and transfer between different optimization tasks. Agents can exploit the optimal knowledge from source tasks to accelerate search rate and improve solution accuracy. The greedy rule is adopted to balance global search and local search. Convergency analysis shows that the proposed method can converge to the global optimal solution with probability 1. The performance of the proposed ITQ appraoch have been validated on 68-bus system. Simulation results in multi-test cases verify that the proposed method has superior convergence rate and stability.