A comparative study on the prediction methods of power system equipment renewal scale based on machine learning

This paper applies different machine learning-based methods for equipment renewal scale prediction in power systems, and compares the predictability of different methods by measuring efficiency and accuracy. Focused on equipment renewal of power enterprises, this paper uses historical data to model training and dynamic prediction of the equipment renewal scale in the future, and obtains an optimal prediction model. The main contributions and innovations of this paper are as follows: (1) A machine learning-based method for predicting the equipment renewal scale of power enterprises is proposed, which can dynamically predict the future equipment renewal demand according to the operation data and equipment data of power systems, thus helping the equipment management and investment decision of power enterprises. (2) By comparing prediction performances of the machine-learning based model and the alternative methods, this paper provides a benchmark for prediction model selection in the field of power system. The comparison is conducted among ordinary least squares, vector autoregressive model, tree model and multilayer perceptron neural network model, and finds out that the neural network model has obvious advantages in predicting the accuracy of the equipment renewal scale of power enterprises. Therefore, the neural network model is more suitable for subsequent prediction.