Optimal operation and alternative to change the production-well of geothermal power plant by using artificial neural network

Abstract

This research presents an Artificial Neural Network (ANN) of Machine Learning (ML) for the predictive maintenance of production wells that require maintenance at a suitable time at the Fang Geothermal Power Plant in Thailand. The raw data covers a period of 48 months (between 2018 to 2021). The data is gathered from log sheets and historical records, covering 1460 instances. Then, this raw data is calculated in the Thermodynamic and Ratio Power Equation. For the ANN model, the dataset has been separated into two sections the training set and the testing set, including 664 instances in total. In general, there are two types of ANN models, including Classification Algorithms and Regression Algorithms. This study applies the ANN Classification Algorithms for simulating ANN models. The manual classification technique and the K-mean clustering algorithms are applied for determining the targets of the ANN model. In the simulation of the model, the K-mean clustering algorithms produced the best result, with 99.83% accuracy. The experiment demonstrates that the predictive maintenance could predict accurately, under established criteria, and inconsistent with the previous maintenance schedules. Therefore, the ANN model will assist operators in assessing and monitoring the system to prevent loss of power generation capacity. This means the model can support the maintenance activities and optimize the operation of the Fang Geothermal Power Plant. In the case of Alternative to Change, the Production-Well is developing predictive maintenance (PM) methods to enhance operational systems that are constant, reliable, and safe. PM technology has a function that can predict potential failures and enhance the management of machine systems. However, the decision-making process on the PM should concern the cost-effectiveness analysis (economic analysis). Thus, this paper discusses a combination of economic analysis and Machine learning (ML) in optimizing the adoption of PM. The Classification Artificial Neural Network (ANN) Algorithm of ML was selected for the PM process. The results of ML and economic analysis are used to define the optimal PM application. The economic analysis is to calculate the power production rise resulting from the prediction of the Classification ANN Model. The proposed approach is to compare the optimal approach in the decision-making on maintenance strategies. This study shows that the use of the ML algorithm can increase power production in the Geothermal Power Plant by an average of 17% in 4 years, with an energy power increase of about 276,444 kWh/year.