Battery-powered electric vehicles (EVs) are gaining more popularity nowadays due to increasing concerns about global warming and the depletion of fossil fuels. The performance of EV battery is affected by various weather conditions. So, the continuous monitoring of battery states is necessary to ensure the reliable operations and safety of the battery. The battery management system (BMS) plays a crucial role in monitoring and executing optimal battery operations. For the monitoring of battery operations, the estimation of key parameter i.e., state of charge (SoC) information is very much needed. This paper proposes an adaptive convolutional neural network (CNN) prediction model to estimate the battery SoC levels for different operating conditions. The proposed CNN model is structured with five layers including input, hidden, and output layers by considering the system complexity and performance time. The proposed SoC prediction model has been tested in different scenarios such as estimation at various ambient temperatures, charging and discharging phases of battery, chemical compositions, and different C-rate values to show its performance efficiency and robustness. Furthermore, the simulation results of the proposed SoC estimation model have also been compared with the other deep learning models such as feed-forward neural network (FNN) and recurrent neural network (RNN) by using different performance metrics to show its superior prediction accuracy levels. Finally, it is clearly evident that the proposed adaptive CNN model accurately estimates the battery SoC values under different operating conditions.