ABSTRACT
The aim of the SOLETE dataset is to support researchers in the meteorological, solar and wind power forecasting fields. Particularly, co-located wind and solar installations have gained relevance due to the rise of hybrid power plants and systems. The dataset has been recorded in SYSLAB, a laboratory for distributed energy resources located in Denmark. A meteorological station, an 11 kW wind turbine and a 10 kW PV array have been used to record measurements, transferred to a central server. The dataset includes 15 months of measurements from the 1st June 2018 to 1st September 2019 covering: Timestamp, air temperature, relative humidity, pressure, wind speed, wind direction, global horizontal irradiance, plane of array irradiance, and active power recorded from both the wind turbine and the PV inverter. The data was recorded at 1 Hz sampling rate and averaged over 5 min and hourly intervals. In addition, there are three Python source code files accompanying the data file. RunMe.py is a code example for importing the data. MLForecasting.py is a self-contained example on how to use the data to build physics-informed machine learning models for solar PV power forecasting. Functions.py contains utility functions used by the other two.
ABSTRACT
Machine Learning (ML)-based methods have been identified as capable of providing up to one day ahead Photovoltaic (PV) power forecasts. In this research, we introduce a generic physical model of a PV system into ML predictors to forecast from one to three days ahead. The only requirement is a basic dataset including power, wind speed and air temperature measurements. Then, these are recombined into physics informed metrics able to capture the operational point of the PV. In this way, the models learn about the physical relationships of the different features, effectively easing training. In order to generalise the results, we also present a methodology evaluating this physics informed approach. We present a study-case of a PV system in Denmark to validate our claims by extensively evaluating five different ML methods: Random Forest, Support Vector Machine, Convolutional Neural Networks (CNN), Long-Short Term Memory (LSTM) and a hybrid CNN-LSTM. The results show consistently how the best predictors use the proposed physics-informed features disregarding the particular ML-method, and forecasting horizon. However, also, how there is a threshold regarding the number of previous samples to be included that appears as a convex function.
