Peak shaving strategy optimization based on load forecasting: Evidence from Anhui Provence, China

Abstract

The rapid growth of renewable energy and electricity consumption in the tertiary industry and residential sectors poses significant challenges for deep peak regulation of regional power systems. This study proposes a “Forecasting-Optimizing” approach for regional peak load optimization that integrates a machine learning-based power load forecasting and optimization model. First, the Attention-based temporal convolutional network (TCN) model is used to predict the time-point load curve in the next year. Then, considering the peak power cutting ratio, time-point distribution and duration, focusing on newly added photovoltaic (PV) installations, user-side demand response (USDR), and energy storage (ES), we built a regional peak load optimization model with the goal of minimizing the peak cutting cost. Taking Anhui Province as an example, the empirical test is carried out. The results show that at 97 % level, the peak power in Anhui Province in 2023 will be about 3 million kWh, and the optimal combination strategy of peak cutting is that the Newly added PV installed capacity is 970.6 MW, and the demand response scale during the night is 1448.2 MW; at 95 % level, the peak power will reach more than 11 million kWh, and the optimal combination strategy is that the Newly added PV installed capacity is 2793.6 MW, and the demand response scale during the night is 2413.7 MW; at 90 % level, the peak power will reach 70 million kWh, and the optimal combination strategy is that the Newly added PV installed capacity is 6464 MW, the demand response scales are respectively 1021 MW and 2900 MW during day and night, and the energy storage configuration scale is 4.06 million kWh. Furthermore, the cost and efficiency of the optimal combination peak cutting strategies are different among different peak power reduction levels. It is suggested that decision-makers analyze power load characteristics of regional power systems, and consider renewable energy penetration and USDR resources to determine the optimal peak cutting strategy that minimizes the cost. This study can provide a solution for the formulation of regional peak cutting policy.