Non-Intrusive Power Load Identification for High-Penetration Renewable Energy Grids

Abstract

With the popularity of new energy grids with high penetration rate, classic non-intrusive power load identification algorithms such as hidden Markov model (HMM) need to face the uncertainty caused by new energy generation. It will cause the load state like active power to continue to change, and new state transitions appear during operation, resulting in the lack of robustness of state identification and power decomposition. Aiming to solve this problem, this study proposes and constructs a Gaussian mixture model–binary parameter hidden Markov model (GMM-BPHMM) which takes into account the randomness of new energy power supply, clusters the load status based on active power and steady-state current to reduce the possibility of volatile clustering results from the new energy grid under a high penetration rate, improves the Viterbi algorithm to take into account the updating HMM parameters to achieve the optimal prediction of the load state, considers the random volatility of load power brought by new energy grids, constructs a power calculation optimization model, and realizes the power decomposition of the load based on the principle of maximum likelihood estimation. Finally, on the basis of the public data set AMPds2, the study generated simulation data based on the new energy generation model and verified the method, and the test case verified the effectiveness of the method.