Fuzzy logic inherited machine learning based maximum power point tracker for cost-optimized grid connected hybrid renewable systems

In this article, multitrial vector-based differential evolution algorithm (MTDE) is proposed as energy and cost management controller under Time of Use (TOU) tariff in grid associated domestic PV-wind power system tied with battery storages. To enrich the energy efficiency of a proposed power system, two optimisation algorithms are proposed in the scheduling operation. TOU billing is a cost- reflective power pricing strategy that has been found as an effective way to reduce peak energy consumption in the residential segment everywhere the world, mainly in industrialised nations. In the optimisation maximizing the cost benefit of the household energy is taken as the objective and the dispatching ratio of electricity sold to the grid and used locally is treated as the optimisation variable. Using MATLAB, the performance of proposed MTDE in the aspect of daily cost benefit and revenue growth rate are presented with the comparative analysis of gravitational search algorithm and conventional self-made for self-consumed and rest for sale (SFC RFS) mode-based energy management controller. In comparison to the SFC RFS mode, the GSA-based cost optimization offers a 21.46% increase in revenue growth, while it is improved to 38.7% using proposed MTDE algorithm. The paper also addresses the significance of fuzzy logic based maximum power point tracking (MPPT) of solar PV in enhancing the energy management of the proposed system. The prototype of fuzzy logic MPPT for 10 W solar panel is presented and the tracked maximum power is visualized using Thingspeak IoT cloud server. The tracking speed of the MPPT can be increased by introducing machine learning algorithms at the cost of memory and complexity. In this article, the linear regression-based machine learning algorithm is implemented in the hardware prototype by utilizing the dataset aggregated from fuzzy MPPT and hence the proposed MPPT inherits the characteristics of the fuzzy MPPT with increased tracking speed. Around 75 % data are used for training, 25% of data are used to test the model and it is observed that the root mean square error (rmse) is 5.1334 and mean square error is 26.3521 and the model is utilized as MPPT for the real-time inputs.