<p>The primary objective of the present study is to predict the optimal performance of thermal power plants using metaheuristic algorithms. Thermal power plants are the backbone of the energy sector and very complex entities. The main aim of this study is to optimize and predict the availability of this complex system through the combining the Markov model with various metaheuristic algorithms. For this purpose, a stochastic model is developed under the concept of cold stand by redundancy and exponentially distributed random variables for failure and repair activities. The Markov birth–death process is used to develop the Chapman-Kolmogorov differential equations. The metaheuristic algorithms which include Black Hole Optimization, Clonal Selection Algorithm, Cuckoo Search algorithm, Gravitational Search Algorithm, Harmony Search Algorithm, Krill-Herd Algorithm used to predict the optimal performance measures, namely availability of thermal power plant. The numerical results of TPP availability along with estimated parameters derived through simulated closely resemble operational data of real-world thermal power plants. These numerical results revealed that the Krill Heard Optimization Algorithm yield the finest performance among all other algorithms having highest optimal predicted Availability 0.9931567. The results and findings of this study are beneficial and have practical implementation for the plant developers, system designers and maintenance engineers to design new TPP and plant maintenance strategies.</p>