In this paper, a FPNN compensated NFTCSMC for renewable energy systems is described that is capable of producing low THD under non-linear rectifier loading and fast transience under transient loading. The importance of the NFTCSMC over classic FTCSMC is finite system-state convergence time and singularity-free. However as the system uncertainty bounds are overestimated or underestimated, the chatter or steady-state errors will occur in the NFTCSMC. The FPNN is thus employed to estimate the upper bound of parameter uncertainties and external disturbances, in order to diminish the over-conservatism of the NFTCSMC design. Once such over-conservative switching gain is decreased, the chatter existed in NFTCSMC can be suppressed, thus enhancing the performance of the renewable energy system. The stability analysis of the improved technique has been carried out using the Lyapunov method, which proves the finite-time reachability of the sliding surface, the asymptotic stability of the closed-loop system, and the finite-time convergence to zero of tracking errors. Therefore, we believe that the improved technique will be helpful for control design of artificial intelligence related systems in future. Simulations and experiments on a renewable energy system prototype using a DSP have been developed to verify the applicability of the improved technique.