Power Grid Material Demand Forecasting Based on Pearson Feature Selection and Multi-Model Fusion

The demand projection of power grid materials can furnish an effective support for the management of power grid materials. Due to variations in the data distribution of individual districts and diversity of materials, a single forecasting model is incapable of accurately predicting the demand for all types of materials. Moreover, for the data-driven network model, the effect of the model has a strong correlation with the quality of its input parameters. To address these problems, this study proposes a power grid material demand forecasting model based on feature selection and multi-model fusion. The first step in this regard is the usage of Pearson coefficient in the selection of main characteristic parameters from original parameters and using them as the input of the network model. Then, stacking fusion algorithm is used to fuse multiple basic models. At last, the proposed method mentioned in this study is tested on a real dataset. The results depict that the proposed method can fully integrate the advantages of various basic models with higher accuracy and generalization ability.

Automated summary

This study proposes a power grid material demand forecasting model based on feature selection and multi-model fusion, which can effectively support the management of power grid materials with higher accuracy and generalization ability.