Abstract:
"Predicting the sola plantation output based on it’s maintenance output will generated more
profitability to both industrial and residential consumers The importance of this has increased
due to it’s direct production to the customers profitability. Managing the correct maintenance
patterns would reduce the maintenance cost while under performance brings low production
due to dust, voltage fluctuations, shadows etc. understanding the correct maintenance patterns
would save unwanted maintenance schedules as well as high labor cost. In this study, several
machine learning techniques were used to predict the sola plantation maintenance using the
factors affecting sola power production. The techniques namely k-nearest neighbor regression,
Artificial neural network (ANN), Decision tree, Random Forest, Linear Regression and
Bayesian Applied Regression, The algorithms were applied on the preprocessed sola production
data captured from the device magic data capturing tool, SolarEdge production monitoring tool
and predicted whether data. Once applied the algorithms, the best model was selected
comparing RMSE and MAPE. KNN was identified to be the best fitting models used for all the
maintenance patterns. The analysis was done using azure machine learning and then the
predicted values are then visualized through Power BI dashboards to present to the engineering
teams. These predictions visualized through dashboards allows the maintainers team to take
correct decisions at the right time while monitoring the progress of the actions. Moreover, the
finalized data model allows the maintainers team to gain the cutting edge over the competitors
while increasing the productions."
