Forecasting Model of Electricity Sales Market Indicators With Distributed New Energy Access

Introduction

China has actively responded to the carbon peak and carbon neutral carbon emission reduction goals to accelerate the green transformation of energy. Under the dual-carbon background, the traditional power system will usher in comprehensive transformation and upgrades (Singla et al., 2021). Among them, distributed power generation offers advantages like cleanliness, efficiency, and local balancing in accordance with the development of China’s future power grid (David et al., 2021). Under the combined effect of objective resource conditions and policy promotion, distributed power generation in all regions of China has shown rapid development.

Distributed power can be freely installed at the distribution network terminals. Through the spontaneous self-use of users, its remaining power can be returned to the distribution network to achieve efficient energy consumption and utilization (López et al., 2022). Still, due to the dynamic energy balance, once the power supply is surplus, it will intensify the phenomenon of abandoning distributed new energy. Therefore, when selecting the grid-connection scheme of new energy according to the grid’s acceptance capacity of wind and photovoltaic power (WPV) in various regions, accurate power prediction is indispensable (Lu et al., 2022). In the new power grid, the massive access of clean energy power generation, such as WPV, will increase the pressure of power dispatching (Bian et al., 2022). Due to the characteristics of distributed new energy, its power generation form has obvious volatility, intermittence, and unpredictability, creating new challenges to the accurate prediction of power consumption in regional power systems.

Regional electricity consumption is characterized by multiple factors, increased uncertainties, and complex changes due to the improvement of consumer terminals’ demand of the new power system and continuous access of new energy sources. The accurate prediction of electricity quantity can provide a reliable basis for the planning and construction of a power grid, optimal dispatching, and optimal load distribution. This poses a challenge to the accurate prediction of regional electricity (Luo et al., 2022). The research results of electricity forecasting are mainly divided into traditional prediction and artificial intelligence prediction (AIP). Among them, AIP methods include tree integration algorithm, support vector machine, and neural network algorithm (Dab et al., 2022; Kalhori et al., 2022; Tan, 2022). Both traditional methods and AIP methods do not fully consider the deep mining and utilization of data under the modern powerful Internet of things (IoT), which limits the highest accuracy of the prediction algorithm.

Based on the above analysis and aiming to accurately predict the electricity consumption, an indicator prediction model under the cloud-side collaborative architecture is proposed. Compared with traditional models:

• 1.

  It aims to improve the processing efficiency of the massive distribution of new energy access in the electricity sales market. Thus, the proposed model builds a prediction system based on the cloud-side collaborative architecture. It completes the data classification processing by deploying K-means clustering algorithm on the edge side to improve data quality.

• 2.

  The LSTM model lacks the analyzing ability of data space characteristics. Thus, the proposed model uses the ConvLSTM model for data learning and the Adaboost integration strategy to weigh and combine the ConvLSTM base learners. This greatly improves the universality of the model.