Proposed article deals with implementation of quaternion neural network in application of maximum power point tracking controller (MPPT) for photovoltaic panel. The research compared ordinary 2-layer feedforward neural network with quaternion neural network. The quaternionic model confirmed promising computational potential in terms of final mean square error, 84% reduction of trainable parameters and 57% reduction of multiplication operations. Reduced computational complexity of quaternion neural network-based MPPT controller makes it possible to increase the switching frequency of DC-DC converters, and therefore to use smaller and cheaper electrical components.