pv-weather-forecast-smart-algorithm
In many PV-Home-Systems with attached battery storage, the greedy algorithm is used. This algorithm tries to charge the battery with all available PV-Power minus current household loads at any time. This leads to a high self-consumption as a minimum of energy is exported to the grid. The rapid charging with all available power also leads to high charge-currents and a high mean state of charge as the battery is in its charged state over a long period of time. This leads to faster aging of the battery and could therefore increase overall system costs. In this report, an alternative, more intelligent algorithm was developed. The aim of this new method is to decrease charge currents and the mean state of charge of the battery. The algorithm is based on a day ahead forecast and past load profiles that feed a scheduler algorithm, that tries to find an optimal charge profile for the next day. In the end, the mean state of charge and the mean c-rate of both algorithms are compared. The overall performance of the system is assessed with help of self-sufficiency and autarky. The algorithm falls back to the simple greedy algorithm, if the required battery energy at sunset can not be reached by a factor of two. This ensures that the maximum energy for the insufficient day can be harvested. A PID-Controller is enabled, when the forecast differs from the actual solar irradiation. The controller tries to match the current energy with the precalculated energy target curve. This enables the system to catch up, when previous forecasts were higher than the actual values. The penalty of this design is the higher charge currents when catching up to the target curve. A side-benefit is the lower mean state of charge. To enable the model to work with real forecasting data, an API-Interface was created. The API-Service is provided by forecast.solar and is free of charge for our purposes. The provided data for the next day is then used to calculate the optimized charging curve for the future day.
