Advanced Load Cycle Generation for Electrical Energy Storage Systems Using Gradient Random Pulse Method and Information Maximizing-Recurrent Conditional Generative Adversarial Networks

• DOI: 10.20944/preprints202410.1629.v1
• Publication Date: 2024-10-22T00:24:58Z
• Authors: Steven Neupert; Jiaqi Yao; Julia Kowal
• Keywords: TK1001-1841; Production of electric energy or power. Powerplants. Central stations; Industrial electrochemistry; loadcycle analysis; simulation; load cycle design; testing; TP250-261

The paper introduces an approach to extract information from measurements and generate new load cycles for electrical energy storage systems. Load cycle analysis is performed using rainflow counting, which helps evaluate data and identify stress factors. Load cycle generation can involve clustering methods, random micro-trip methods, and machine learning techniques. The study utilises the Random Pulse Method (RPM) and presents an improved version called the Gradient Random Pulse Method (gradRPM) that allows control over stress factors such as the gradient of the State of Charge (SOC). Another method called Information Maximizing-Recurrent Conditional Generative Adversarial Network (Info-RCGAN) has been developed, and it utilises a deep learning algorithm for data-driven load profile generation with control over stress factors. The results demonstrate the effectiveness of the gradRPM and Info-RCGAN methods in generating load profiles based on the given parameters. The findings provide valuable insights into designing simulation data or testing data for electrical energy storage applications, aiding in improving and understanding system behaviour and requirements.