Solar Power Prediction in Smart Grid Based on NWP Data and an Improved Boosting Method

Abstract

Nowadays, large-scale photovoltaic (PV) systems are penetrating into the modern electric power grid infrastructure — Smart Grid (SG), at an unprecedented rate. One of the biggest challenges for integrating PV into the SG is to overcome the intermittent and uncontrollable nature of PV power output. Hence, building a timely and efficient prediction model can be extremely helpful in system planning and market operation of grid-connected PV systems. This paper provides a prediction model based on numerical weather prediction (NWP) data and an improved boosting method. NWP data is used as a new variable in simulating the photovoltaic power generation to improve the accuracy of the model. The improved boosting method uses the Taylor formula to develop a loss function and takes measures to control the complexity of the algorithm. The model described in this paper has a proper handling of a vast number of variables included in meteorological data, runs faster than existing popular models on solar power prediction and scales to billions of examples in distributed or memory-limited settings.