As IoT is a more essential and fast-growing sector in the industry, the need for service composition arises. The composition of the service is based on the specifications and needs of the user. As Quality of Service is unavoidable in an IoT network, the service selection model must have the necessary QoS constraints. The energy consumptions must also be considered in the model as the network state may sometimes directly depend on the network state. Taking all this into account, the service composition model is required that gives an optimal solution subject to the constraints. We formulated this statement into a mathematical optimization problem model and the solution of this model is obtained from the teaching-learning-based optimization algorithm, which is a metaheuristic algorithm. The mathematical model is then deployed to a service selection model that can be used in any real-time application. The analyses are made with the executed results and according to the needs of the application the algorithm can be selected.