An Improved Boltzmann Machine for Strategic Investment Planning in Power System Environment

Abstract

The objective of this research is to propose an effective method to determine an optimal solution for strategic investment planning in power system environment. The proposed method will be formulated by using mean-variance analysis approach in the form of mixed-integer quadratic programming problem. Its target is to minimize the risk and maximize the expected return. The proposed method consists of two phase neural networks combining Hopfield network at the first phase and Boltzmann machine in the second phase resulting the fast computational time. The originality of the proposed model is it will delete the unit of the second phase, which is not selected in first phase in its execution. Then, the second phase is restructured using the selected units. Due to this feature, the proposed model will improve times and the accuracy of obtained solution. The significance of output from this project is the improvement of computational time and the accurate solution will be obtained. This model might help the decision makers to choose the optimal solution with variety options provided from this proposed method. Therefore, the performance of strategic investment planning in power system engineering certainly enhanced.

A Modified Boltzmann Machine for Solving Distribution System Expansion Planning in Malaysia

“…Theoretically, Hopfield network can solve the mixed-integer programming problem within minimal period of time despite providing less accurate and optimum solutions. On the other hand, Hopfield network is easily trapped into the local minimum, thus a modification was made by employing simulated annealing to escape the local minimum in the form of BM [22], [23]. The advantage of BM is it can select the optimum and accurate solutions since it yields the global optimum solutions and at the same, it requires more computational time.…”