For a battery system of management, accurate status predictions is essential. Precise state-of-charge (SoC) estimation is particularly important for the secure and dependable operation of cells. But the on-board system is only able to estimate the SoC of battery cells based on other quantifiable data collected during battery operation; it is not able to measure the SoC of battery cells directly. In this study, prior to the proposal of a network, in order to determine whether convolutional neural networks (CNNs) and long short-term memory (LSTM) networks be applicable to data prediction, their respective properties are examined. The advantages of particle swarm optimisation (PSO) in optimising the hyperparameters of neural networks are studied. A simulation of a SoC prediction based on a PSO-CNN-LSTM combination model was carried out in MATLAB. CNN is used for feature extraction, whilst the LSTM model optimised by PSO is utilised for SoC prediction employing K2 lithium battery test data of the CALCE Battery Research Group at the University of Maryland. To confirm the network's application and reliability in battery SoC prediction, the performance of CNN, LSTM, and PSO-LSTM is compared with that of the proposed network, namely PSO-CNN- LSTM. The prediction performance of the network has significantly improved, according to simulation results. The suggested model outperforms the other techniques in terms of estimation accuracy.