Accurate state of charge prediction for lithium-ion batteries in electric vehicles using deep learning and dimensionality reduction

One of the most crucial and pricey parts of electric automobiles is the battery. The state of charge of lithium-ion batteries, which are primarily found in electric vehicles (EV’s), is essential to their ongoing functioning. To guarantee precise battery balancing and accurate assessment of the vehicle’s remaining driving range, a robust state of charge prediction model is needed. The exact charge level attests to the dependability and safety of electric cars. Nonetheless, state of charge is influenced by a variety of factors that cannot be easily quantified. This work mainly focuses on designing the state of charge of lithium-ion batteries in electric vehicles using a novel deep learning model and a dimensionality reduction mechanism. Initially, the current, voltage, and temperature data are collected from the openly available dataset. After that, normalization is performed on the collected data to standardize the dataset. Moore–Penrose pseudo-inverse (MPPI)-based principal component analysis is then implemented to diminish the dimensions of the standardized data. These dimensions reduced dataset are finally fed into the optimal parameter selection-based long short-term memory model, which predicts the lithium-ion batteries’ state of charge. The experiments are carried out employing the Panasonic 18650PF lithium-ion battery dataset. Simulation findings demonstrate that the suggested algorithms can accurately predict the lithium-ion batteries’ states in a short period of time with minimal error when contrasted with existing techniques.