In order to better adapt to the variation in solar irradiation and to improve the efficiency of the photovoltaic generator, i.e. to maximize the power delivered to the grid. Several criteria's for optimizing the efficiency of the photovoltaic system have been applied. Among them, the algorithms for tracking the optimal operating point of the photovoltaic panels that called Maximum Power Point Tracking (MPPT). In this article, a PV generator (GPV) has been connected to the power grid, as a result, direct consequence is in the deterioration of the voltage wave and thus the quality level of the energy supplied to the consumers. To overcome these problems of harmonic pollution, active power filtering is proposed as an efficient solution to improve grid power quality. This paper therefore proposes to examine the characteristics of an association between a photovoltaic generator (PVG) that aims at injecting active power into the electrical grid and a parallel active filter that has the task of eliminating disturbances present in this grid. The theory of the twophase method with Adaline harmonic extraction is applied for the extraction of the reference currents according to the DQ reference frame. Finite set mode predictive current control (FS-MPCC) applied on PAF has been proposed in order to compensate undesirable harmonic, and reactive power resulting from a non-linear load. A Global Maximum Power Point Tracking (GMPPT) algorithm based Adaline method has been suggested for extracting power from PVG. A simulation under Matlab/Simulink of the global system proves the robust performance capability of the suggested Adaline Neuro-Predictive (ANP) control to simultaneously provide harmonic current compensation, power factor correction and solar power energy injection into the grid.