Modeling the condition of lithium ion batteries using the extreme learning machine

Abstract

Recent years have seen increased interest in the use of off-grid solutions for electrification of rural areas. Off-grid electrification (such as solar home systems and micro-grids) are particularly applicable to the rural African context, where little infrastructure exists and in many regions grid extension is prohibitively expensive. To be economically viable, these systems must maximize the power delivered while ensuring the health of energy storage devices. Batteries in particular are a key weakness and typically the first major component to fail. In this paper we present an improved and simplified method for simulating the state of charge (SoC) and state of health (SoH) of lithium-ion batteries. SoC and SoH are predicted using the Extreme Learning Machine (ELM) algorithm. ELM is a state of the art single layer, feed-forward neural network that is characterized by its good generalized performance and fast learning speed. Real-life battery data from the NASA-AMES dataset provides the benchmark for evaluation of the ELM model.