Research of machine learning models in measuring the level of petroleum products with capacitive level meters

The paper investigates methods for measuring the level of petroleum products using machine learning algorithms. Three approaches are considered: discrete, continuous and combined (discrete-continuous) measurement method. The main technological parameters of the operation of capacitive level meters for measuring the level of light petroleum products are highlighted. A combined method is considered that combines the advantages of discrete and continuous measurements, allowing minimizing systematic errors. A comparative analysis of the accuracy of various methods based on random forest modeling is carried out. Key statistical indicators such as MSE, MAPE, and R2 have been calculated, confirming that the combined method is significantly superior to traditional approaches in terms of prediction accuracy. Technical conclusions have been drawn confirming the effectiveness of using machine learning in this task.