Development of a Machine learning assessment method for renewable energy investment decision making

Abstract

The aim of this study is to develop a hybrid methodology based on statistical approach and machine learning algorithm to structure the investment decisions in renewable energy industry. Six main criteria including GDP rate, inflation rate, electricity demand, electricity generation from Renewable Energy Technology (RET), electricity price, and CO2 emissions are used to classify and structure the decision-making algorithm. The risk related to each feature is calculated through using standard deviation of their historical data collected from each country. The Machine Learning (ML) algorithms such as Artificial Neural Network (ANN), Logistic regression (LR), Support Vector Machine (SVM), Decision Tree (DT), Naive Bayes (NB), k-Nearest Neighbors (k-NN), Random Forest (RF), and Ada Boost (AB) have been applied on the dataset related to each RET to classify the countries from the lowest risk (best country) to the highest risk (worst country) in terms of RET investment alternatives. LR, SVM, ANN, and DT showed the best performance among other ML algorithms. The highest accuracy was 0.94 and the lowest accuracy belonged to AB with the value of 0.24 for biomass technology. The minimum and maximum MSE was obtained 0.058 and 0.771 by LR for biomass technology and AB for hydropower technology, respectively. Moreover, the electricity generation risk from RET and electricity price risk in each country are the most important criteria for decision making in RET investment.