<p>Electricity generation via unconventional methods, such as spontaneous electricity from water evaporation, is gaining substantial interest. A family of solid oxides has demonstrated the ability to generate electricity via water evaporation. This study experimentally investigated the electricity generation performance of power generation devices based on binary oxide nanoparticles such as aluminum oxide, titanium dioxide, zinc oxide, and silicon dioxide at different mixing ratios (20%, 40%, 60%, and 80%) under different temperature and humidity conditions. The structural characteristics of the nanoparticles themselves, their charge state, and interactions with fluids, the Debye effect, and the environment significantly affect their power generation performance. A machine learning method was adopted to tackle the complicated multiparameter coupling problem with limited experimental data. This method yields highly accurate performance optimization and efficiently guides the design of power generation devices. The neural network model uses multiple hidden layers to obtain <i>R</i>² values above 0.88, indicating the accurate prediction of power generation performance. An electricity generation device based on SiO₂ and Al₂O₃ with 40% Al₂O₃ content achieved an output power of 199.5 nW under environmental conditions of 25°C and 40% relative humidity, which is the best among all binary oxide devices explored herein.</p>