Multi-Timescale Forecasting of Battery Energy Storage State-of-Charge under Frequency Containment Reserve for Normal Operation

Abstract

Forecasting the state-of-charge changes of battery energy storage, anticipated from a provision of different services, can facilitate planning of its market participation strategy and leverage the maximum potential of its energy capacity. This paper provides a performance comparison study of multiple decision-tree and data-driven machine learning methods for point forecasts of the state-of-charge of battery energy storage under frequency containment reserve for normal operation on day-, hour-, and 15-minute-ahead basis. The battery state-of-charge data for the performance evaluation were simulated with a droop curve battery model based on the historical frequency data in the northern Europe synchronous area. The results show that the data-driven methods outperform the decision-tree based methods on the 15-minute-and day-ahead time scales while demonstrating a comparable performance for the hour-ahead time scale.Index Terms-battery state-of-charge, point forecasting, decision trees, machine learning, frequency containment reserve

LSTM-based Multi-Step SOC Forecasting of Battery Energy Storage in Grid Ancillary Services

“…Several features such as timestamp, market, frequency, and modified SOC are available in the dataset. From exploratory data analysis (EDA) in [11], the BES SOC variation has a high correlation with the frequency measurement feature. With an almost perfect correlation in mutual information score (0 = no mutual information, 1 = perfect correlation), the frequency measurement becomes the essential feature that needs to be considered to reduce uncertainty and support the SOC forecasting accuracy.…”

“…We analyze and compare the performance of our developed SOC forecasting models with the existing ML methods described in [11]. The existing algorithms are included DT, RF, LightGBM, CNN, and RNN.…”

A Virtual Power Plant Solution for Aggregating Photovoltaic Systems and Other Distributed Energy Resources for Northern European Primary Frequency Reserves

“…One major use case for the virtual power plant application programming interface is energy forecasting. Several machine learning forecasting applications relevant to virtual power plants have been proposed for wind power generation [9]; photovoltaic power generation [10]; electric vehicle charging load [11]; electric vehicle battery state-of-charge [12]; battery storage state-of-charge [13]; and Heating, Ventilation and Air Conditioning (HVAC) load [14]. This article demonstrates the use of the virtual power plant application programming interface for one energy forecasting application.…”

LSTM-based Multi-Step SOC Forecasting of Battery Energy Storage in Grid Ancillary Services

“…In the case of BES in grid ancillary services, SOC forecasting models need to deal with multi-physics conditions to support forecasting accuracy. In [11], several traditional and stateof-the-art ML techniques, ranging from decision-tree-based methods (Decision-Tree (DT), Random-Forest (RF), and Light Gradient Boosting Machine (LightGBM)) to data-driven deep learning approaches (Convolutional Network Network (CNN) and Recurrent Neural Network (RNN)), have been proposed and analyzed. The developed ML models considered the uncertainties of corresponding frequency features that periodically determine the BES SOC.…”

“…Four state-of-the-art LSTM variants techniques, i.e., Vanilla-LSTM, Bidirectional-LSTM (Bi-LSTM), Vanilla-Gated recurrent unit (GRU), and Bidirectional-GRU (Bi-GRU), are proposed and analyzed to solve this problem. 3) We evaluate our developed multi-step SOC forecasting models along with the existing ML methods in [11] using real-world datasets that publicly available in [12]. We believe that the work presented in this paper will aid in realizing an efficient and meaningful tool for optimal decision-making applications of BES participation in smart grid ancillary services.…”