Geographical Data and Predictions of Wind-Mill Energy Production

Clustering Perspective Forecasting Model
Students: Weisi Li
Supervisor: Philippe Bonnet
Level: MSc, Semester: Spring 2020
Tags: Machine Learning, Energinet, Electricity Forecasting

Wind energy is a rapidly growing and renewable resource. Denmark has abundant wind resources, and more than 50% of the electricity every year comes from windmills. Therefore, Energinet, as the operator of the Danish national electricity and gas system, expects to forecast the wind power generation and incorporate it into the market. Here, the search and statistics of electricity prices are worthy of attention. The electricity price prediction needs a combination of the windmill energy forecast that applies them to the Danish electricity market. Currently, Energinet uses the K nearest neighbor(KNN) model for windmill energy prediction, which can provide short-term energy prediction. However, at present, the improvement of prediction models is the main problem. For example, (a) use more reliable weather data, (b) mine other relevant data, © use better prediction models, (d) extrapolate into long-term forecasting.

The goal of the project is to design and implement a clustering perspective forecasting model. Moreover, the project aims to apply geographic data (such as ground roughness, elevation, windmill location) and accurate weather data to support the construction of the forecasting model.

The intended learning outcomes for this project are: – Design and implement data collecting and cleaning methods – A small supplementary survey on the clustering-based wind-mill forecasting methods – Design and implement a forecasting model for the electricity production of Den- mark windmills based on weather and geographic data – Compare and analysis the model to the model in use at Energinet – Evaluation and reflection of the forecasting model