International Review of Electrical Engineering (I.R.E.E.), Vol. 13, N. 3 ISSN 1827- 6660 May – June 2018 Copyright © 2018 Praise Worthy Prize S.r.l. - All rights reserved https://doi.org/10.15866/iree.v13i3.13534 204 Evaluation of Generation System Reliability Using Adaptive Neuro-Fuzzy Inference System (ANFIS) and Artificial Neural Networks (ANNs) Khanittha Wannakam, Somchat Jiriwibhakorn Abstract – This paper presents an evaluation of the reliability index of power generation systems using the Adaptive Neuro-Fuzzy Inference System (ANFIS) and Artificial Neural Networks (ANNs) to compare the results obtained from the basic method of probability. The reliability index used in this study is the Expected Energy Not Supplied (EENS) index, which is used in planning to increase the installed capacity for the adequate demand for electricity. The ANFIS and ANNs techniques will learn the relationship between the priority level, the installed capacity and the force outage rate (FOR) of the generator, which significantly affect the EENS index. The results indicated that the ANNs techniques have the best predictive performance. The best accuracy of the training data was 1.2488% and the testing data was 2.3963%, calculated using a Mean Absolute Percentage Error (MAPE). Furthermore, the ANNs took more time to learn faster than the ANFIS. Copyright © 2018 Praise Worthy Prize S.r.l. - All rights reserved. Keywords: Adaptive Neuro-Fuzzy Inference System (ANFIS), Artificial Neural Networks (ANNs), Expected Energy Not Supplied (EENS), Generation System, Reliability I. Introduction Currently, the development of industrial equipment or facilities requires electrical energy power sources. An electricity supply is therefore a basic part of an infrastructure and it is important to national development. Continued population growth and ongoing economic growth have increased the annually demands for electricity. Thus, electricity is very important in daily living and is fundamental in driving the economy. If there were insufficient power to supply the demand, this would have a severe effect on the Thai economy. To ensure the capacity of power plants used in production to be large enough to generate a sufficient amount of electricity, it is necessary to plan for reliable electrical systems in accordance with the specified criteria, then they should meet the increasing demand for electricity each year [1]-[27]. The reliability index used for evaluation is the Expected Energy Not Supplied (EENS), which is an index used for decision-making to increase the installed capacity of generators in response to the demand for electricity. In calculation by the probabilistic method, to find the Reliability Index, the Reliability Test System is used to compute the database [1]. The calculation is based on the probability principle of the reliability index calculation. For the ANFIS and ANNs applications [7], [13] the program calculates the correlation of the input and output variables by using a trial and an error method, which allows accurate and fast evaluation of the reliability index. This was used to make decisions about increasing the installed capacity to respond to the demand for electricity. Accurate and reliable evaluation are critical to the adequacy of electrical power, as they can determine the timing of future investments and operations in electricity generation. If the evaluation fails, this could affect the reliability and adequacy of the system. This may result in insufficient power to meet the needs of the consumer or in insufficient investment in building power plants. II. Generation Systems Most power generation planning is considered only for a single power system, which is used to study the adequacy of the power demand of the system and to determine its reliability. This is due to the failure status of the generator, which may result in the total system capacity being insufficient for the load requirements. In modeling, the specifics of the generator and the model of the demand for electricity [1] were considered. Generally, the operation of the electrical equipment, such as, generators, is characterized by a period of time between the available and unavailable states. Fig. 1. Conventional system model