Umeå University's logo

Accurate prediction of battery energy output is essential forenergy management and battery system optimization. However, the complex nonlinear relationships among battery parameters make reliable prediction challenging for traditional modeling approaches. This paper proposes a hybrid quantum-classical neural network (HQNN) framework forbattery energy prediction. The proposed model integrates classical neural network layers for feature extraction and dimension reduction with avariational quantum circuit for nonlinear feature transformation. Computational results on a battery dataset demonstrate that the HQNNachieves competitive performance compared with several classical machine learning models, including Random Forest, XGBoost, and multilayer perceptron. In particular, the HQNN provides lower predictionerrors and improved R2scores, indicating its ability to capture complex nonlinear relationships in battery data. Furthermore, SHAP-basedfeature importance analysis is conducted to interpret the influence ofbattery variables on the prediction results. The findings suggest thathybrid quantum machine learning models offer promising potential forenergy-related prediction tasks.