DEVELOPMENT OF MONITORING TECHNIQUE FOR OPERATION AND MAINTENANCE OF HYDROPOWER PLANT

Abstract

Due to an increase in population and urbanization, energy demand is rapidly increasing across the world. A major part of this energy demand is fulfilled by burning fossil fuels; however, it also causes a harmful impact on the environment. Thus, renewable energy sources can be an alternative which are generated from sources or processes that are continually replenished such as solar, wind, hydro, biomass, and geothermal energy. Among all the renewable energy resources, hydropower is one of the mature and reliable energy source with high conversion efficiency. To harness hydropower, hydro turbine is considered as the most important component of a hydropower plant in terms of energy conversion efficiency. The main causes of the reduction in turbine efficiency are silt erosion and cavitation, which are mainly associated with reaction turbines (e.g., Francis and Kaplan turbines). The reaction turbines use both the kinetic and pressure energy of water and convert it into mechanical power. Flowing water carries sediment, which passes through the underwater parts of the turbine and is responsible for material erosion. An eroded turbine takes more discharge than a rated discharge for generating the same amount of power. Moreover, cavitation affects the performance of the turbine due to part load or overload operation of the machine, and it occurs when fluid pressure drops below its vapor pressure. Erosion of underwater parts creates vibration & noise and reduces the life of the machine. Therefore, it is also difficult to predict the performance of hydro machines due to their complicated architecture and complex working conditions. Many studies have been conducted by several investigators on silt erosion and cavitation, which are available in the literature. However, a very few studies are reported on the complete scenario of operation and maintenance (O&M) of hydropower plants. Therefore, there is a need to develop a real-time condition-based monitoring system for hydropower plants. In view of the above, the present study is carried out with the following objectives (i) To identify the critical parameters related to the operation and maintenance of hydropower plants, (ii) To collect field data from the operational plant and analyze the data, (iii) To develop models for operation and maintenance of hydropower plant based on analyzed data and (iv) To develop an IoT-based system for the O&M of hydropower plants.In order to achieve the aforementioned objectives, critical parameters related to the operation and maintenance of hydropower plants have been identified. Under the present study, the analysis of historical data collected from an MB-II hydropower plant situated in the hills of Uttarakhand state in India has been conducted. It is found that the performance of this plant is severely affected by silt erosion and cavitation. The collected data were analyzed and different correlation models using multiple linear regression (MLR), Curve fitting, support vector machine (SVM), artificial neural network (ANN), and adaptive neuro-fuzzy inference system (ANFIS) methods are developed. Based on the developed models, it is found that the ANN based model can predict the efficiency of the machine with better accuracy, however, there are limitations of large memory storage and processing power requirements. Further, considering the limitations of the ANN technique, an attempt has been made to develop an ANFIS model. This method does not have such limitations despite the fact that its performance is found to be very close to the ANN technique. Moreover, due to the limitations of the NodeMCU ESP8266 board IoT based monitoring system with the ANN & ANFIS based models, the curve fitting method has been used for further development of correlations for discharge utilization factor and power factor to predict the performance of hydro turbine and power evacuation system. The O&M costs data of different hydropower plants were collected and correlations are developed for the prediction of O&M costs of hydropower plants using the curve fitting technique. It is found that the developed model predicted the O&M costs for Francis turbinebased hydropower plants with an accuracy of R2-value as 0.89 having MAPE of 3.53% at 4.45% RMSPE. Further, an IoT-based system is configured for predicting the machine efficiency, discharge utilization factor and power factor using the curve fitting models. The sample data of the plant is passed to the correlation equations over the ThingSpeak cloud to monitor the performance of the plant. The performance of the proposed correlations-based monitoring system is found satisfactory, which can be utilized in power plants to monitor the behavior of the machines in real time. It can minimize breakdowns, identify fault initiation points and improves the machine efficiency. It is recommended that the proposed monitoring system should be improved further by using the ANFIS-based correlation models. Accordingly, a similar system can be designed for the Kaplan and Pelton types of hydro turbine-based power plants in future studies.