Open Access Open Access  Restricted Access Subscription or Fee Access

Multi-layer Perceptron Ensemble for Estimation of Available Transfer Capability in Deregulated Environment

Rajnikant H. Bhesdadiya, Rajesh M. Patel


The available transfer capability (ATC) estimation problem can be considered as a unique answered puzzle to be solved with highly nonlinear complexity of power systems and the uncertain state of the energy market. To implement the open access in deregulated markets, system operator should post ATC at a small and regular interval on openly accessible network. This forces the quick computation of ATC of transmission path. In these situations, the neural computing is the best candidate for such problem solution due to its good learning and generalization capability. The multi-layer perceptron (MLP) is the most popular topology, applied to several practical problems. Supervised MLP learning is made by ATC data patterns generated by optimal power flow methodology, MLP inputs are real power generation and demand, and output is the ATC for the identified path. ANN training is an optimization problem of selecting optimized weight connecting the ANN layers, usually fulfilled with several optimization trials with different initial conditions. This offers a set of truly trained MLP but inferior to best as a by-product. Frequently, the MLP trained for practical problems, experiences a problem of local minima entrapment. This paper proposes a multi-layer perceptron ensemble (EMLP) solution methodology to estimate the ATC by using set of truly trained MLPs, which includes best MLP and inferior set of MLPs, to trim down local minima entrapment problem. As only converged networks are used in the formation of EMLP, the question of generalization for EMLP network does not occur. To the best of our knowledge, this is the first attempt to use ensembles concept for ATC estimation. This paper also proposes three MLP selection methods for EMLP, based on MLP performance, all possible combination, and teaching–learning based optimization. Performance based MLP selection method is handy tool, whereas all possible combination based method is most accurate. However, optimization based MLP selection method performs well considering computation, when numbers of MLP are more. A case study of 6-bus and IEEE 30-bus system is presented to demonstrate the approach feasibility. An experiment shows that the EMLP results are improved and encouraging one for test inputs, including bad estimates of MLP-best.


Keywords: All possible combination based EMLP, available transfer capability (ATC), artificial neural network (ANN), ensemble multi-layered perceprtron (EMLP), MLP selection for EMLP, multi-layered perceptron (MLP), optimum power flow (OPF), performance based EMLP, teaching–learning based optimization (TLBO)

Cite this Article


Bhesdadiya and Patel. Multi-Layer Perceptron Ensemble for Estimation of Available Transfer Capability in Deregulated Environment. Trends in Electrical Engineering. 2016; 6(1):    57–71p.


All possible combination based EMLP; Available transfer capability (ATC); artificial neural network (ANN); Ensemble Multi-layered Perceprtron (EMLP); MLP selection for EMLP ; Multi-layered perceptron (MLP); Optimum power flow (OPF); Performance based EMLP

Full Text:



  • There are currently no refbacks.