This article presents a novel approach to battery thermal management control in electric vehicles (EVs), focusing on the establishment of a power loss model that incorporates temperature and aging effects on internal resistance, thereby enabling accurate estimation of battery power loss for optimized battery thermal management systems (BTMS). In addition, this article introduces a BTMS design capable of both heating and cooling, aiming to maintain optimal battery temperature and enhance battery efficiency and longevity. The proposed methodology includes an offline optimization layer to improve battery longevity and BTMS energy efficiency and an online control layer to maintain a safe battery temperature operation. The adaptability of this BTMS design for real-time applications in various climates is achieved by integrating discharge rate (c-rate) information from the drive cycle. This results in a two-level, driving-aware BTMS control system tailored to varying driving patterns specific to commuter applications. Consequently, this research significantly advances EV battery thermal management by addressing key challenges such as reducing power loss estimation error by up to 28%, optimizing temperature regulation, improving power efficiency by up to 7 kJ for different drive cycles, and enhancing battery aging by more than 3% per life cycle, while ensuring adaptability to various driving patterns for commuters.

<div class="section abstract"><div class="htmlview paragraph">Battery terminal voltage modelling is crucial for various applications, including electric vehicles, renewable energy systems, and portable electronics. Terminal voltage models are used to determine how a battery will respond under load and can be used to calculate run-time, power capability, and heat generation and as a component of state estimation approaches, such as for state of charge. Previous studies have shown better voltage modelling accuracy for long short-term memory (LSTM) recurrent neural networks than other traditional methods (e.g., equivalent circuit and electrochemical models). This study presents two new approaches – sequence training and data shuffling – to improve LSTM battery voltage models further, making them an even better candidate for the high-accuracy modelling of lithium-ion batteries. Because the LSTM memory captures information from past time steps, it must typically be trained using one series of continuous data. Instead, the proposed sequence training approach feeds a fixed window of prior data (e.g., 100 seconds) into the LSTM at each time step to initialize the memory states properly and then only uses the output at the current time step. With this method, the LSTM just requires the prior data window to be continuous, thereby allowing the handling of discontinuities. This also means that during the training process, the data can be shuffled randomly, enabling mini-batches to speed up the training significantly. When these approaches were applied, LSTM voltage estimation error was reduced by 22%, from 28.5 mV to 22.3 mV RMS error over four drive cycles and temperatures from -20 to 25°C.</div></div>

Nowadays, lithium-ion batteries are widely used in electric vehicles (EVs). For safer and more efficient battery operation, a battery management system (BMS) that estimates the state of charge (SOC), state of health (SOH) and power capability (state of power, SOP) is required. Accurately estimating SOP is particularly challenging due to its susceptibility to various factors, including temperature, SOC, and aging. To gain insight into how various factors affect power capability, this study conducts an SOP measurement test throughout a battery aging test at two different temperatures. The SOP measurement technique utilized is developed from our previous patented work. Analysis of the impact of aging, SOC, SOH, and temperature on SOP dynamics are conducted. Furthermore, our study scales the cell-level results to a plug-in hybrid electric vehicle, offering practical insights into SOP estimation algorithms for usable energy and driving range scenarios. Given the absence of comparable data online, these experimental results serve as a valuable resource for developing an accurate power capability estimation algorithm.

Accurate state-of-health (SOH) estimation is critical for lithium-ion batteries' safe and reliable operation. These batteries are widely used for commercial products, including smartphones, laptops, and electric vehicles. In this paper, we develop a convolutional neural network (CNN) based battery SOH estimation model trained to estimate SOH from constant current charge and discharge data. Aging data from four cells, each charged with a different fifteen-minute fast-charging current profile, is used to train and test the SOH estimation model. The model's accuracy is demonstrated by training with data from one fast-charging aging case and tested using the other three cases, which age at a considerably different rate. The results show that the method is quite robust when the tested cells have more than 80% SOH, with error typically within $\pm \mathbf{2}{\%}$ and not exceeding $\pm \mathbf{3}{\%}$ . However, the proposed method has limitations when trying to predict battery health below 80% or when trying to predict battery health from curves with different C-rates. The datasets and the code for the algorithm in this paper are available to download.

Lithium-ion batteries are widely used in electric vehicle powertrain systems. As batteries age, their state of health (SOH), indicated by their usable capacity and power capability, decreases. For reliable battery operation, accurate estimation and prediction of SOH are essential. This paper proposes an algorithm for estimating battery capacity SOH from an open-source fast charging dataset with many different charge profile types. Histogram data is created from the measured time domain data and fed into a feedforward neural network (FNN). To capture the impact of different charge profiles on aging, current and state of charge (SOC) are multiplied together to create an additional synthetic input to the estimator. To reduce the number of inputs to the FNN to only those that contain valuable information, we use principal component analysis to reduce the total number of inputs by 80%. An SOH algorithm is proposed that can estimate capacity throughout the battery's life with a 1.03% root mean square percentage error (RMSPE) and 0.68% mean absolute percentage error (MAPE).