Calgary AI Club developed a machine learning model to predict a power plant's output using soft sensors.
In this model, a Boosted Decision Tree Regression algorithm was trained on a large labelled dataset. The model was taught using supervised machine learning, and was developed in Microsoft's Azure Machine Learning Studio. The model takes inputs from four related physical characteristics (ambient heat, ambient temperature, relative humidity and vacuum pressure) to predict the power output, with an error of 0.6%. In the oil and gas industry, there is considerable time and effort spent testing and monitoring the hydrocarbons in a pipeline or in a refinery. Using soft sensors to predict the properties of the hydrocarbons could reduce the downtime of the pipeline or refinery and could result in significant cost savings.