Reliable renewable-energy conversion and storage are required for sustainable electric-vehicle charging in regions with unstable electricity supply. This study presents a photovoltaic–wind hybrid charging testbed comprising a 300 W photovoltaic module, a 500 W permanent-magnet wind generator, and a 12 V, 40 Ah battery. A single-switch dual-parallel-inductor buck converter regulates the photovoltaic output, while the rectified wind-generator output is stepped down to a 15 V DC charging bus. The parallel inductors share the converter output current under balanced branch conditions. An artificial-neural-network-assisted perturb-and-observe controller estimates and corrects the photovoltaic maximum-power operating point, while a proportional–integral controller adjusts the MOSFET duty ratio. An Arduino Uno implements the control functions, and an ESP8266 records source, converter, load, and battery variables. Three months of time-series measurements were used to train an artificial neural network to predict battery voltage, charging current, and state of charge. Under matched simulation conditions, the improved converter produced 15 V and 16 A, compared with 14.3 V and 15 A for the conventional converter, corresponding to an 11.9% increase in delivered output power. Among networks containing 10-200 hidden neurons, the 100-neuron model achieved the lowest reported validation MSE of 0.00145 and the regression co-efficient of 0.98 This research system can be applied to EV-charging applications and other battery-charging devices.