Sigmoidal and Gaussian functions based neural neuron technique for grid interactive solar energy system enabling power quality improvement
An adaptive approach based on a neural network is presented here for the control of a grid interactive solar power generating system. This control approach is based on a summation and a product neuron, which are used to process the non-linearity of load currents. A sigmoidal and Gaussian functions are used to linearise the response of summation and product neurons. In this control, the product and summation neurons, and sigmoidal and Gaussian functions are within a single layer, which reduces the complexity of control. This control algorithm is superior over artificial neural network (ANN)-based control algorithm, which has large number of unknown weights and layers. It reduces the computational burden and complexity of the control algorithm. It shows the inherent improved performance at non-linear balanced and unbalanced loads. The control of net active and reactive components of the power separately improves the efficacy of the control at unbalanced non-linear load and reduces the voltage transients in the grid voltages. The ANN-based control also plays an active role in enhancing the power quality of the grid. The reduced complexity of control and direct power feed forward of photovoltaic array (PVA) to the grid improve the dynamic response of the grid-connected PVA-based system.