Inter-hours rolling scheduling of behind-the-meter storage operating systems using electricity price forecasting based on deep convolutional neural network

Abstract

With the growth of renewable energy utilities, it is necessary to optimize system scheduling to reduce operation cost and increase profit. An effective approach depends on electricity price forecasting becomes important in energy management and control of behind-the-meter storage system. In this paper, for the first time, an inter-hours rolling horizon strategy is proposed to offer a more effective scheduling strategy based on high-resolution five-minute data. In addition, to avoid great uncertainties in price fluctuation, we present a novel model based on deep learning for hour-period and multi-step price forecasting. Moreover, in order to detect spikes in thecurrent day, we design a convolutional neural network with an end-to-end manner to detect price spikes and capture severe price variations in market profiles, which remarkably improve the scheduling of behind-the-meter storage system and operation profits. Sufficient experiments introduced the real-time available market dataset from Ontario to evaluate the performances of our proposed models. Related spike predictions were also used to optimize operation scheduling of a behind-the-meter storage system, demonstrating the outcomes of our proposed inter-hours rolling horizon strategy from an economic perspective. Compared with the state-of-the-art model (ARX), our proposed strategy based on enhanced spike detection achieved 8.7% improvement on operating cost with more predominant robustness and efficiency.