IMPROVED POWER SYSTEM REALIZATION AND INTEGRATION

Abstract

The main goal of this research is to pave a new path to solve electric power system problems from a realistic perspective. The problems covered in this thesis are power transmission lines, power flow (PF) or load flow (LF) analysis, economic load dispatch (ELD), power and energy forecasting, optimal relay coordination (ORC), fault location, and state estimation. In this journey, we try to push the borders by digging everywhere to find other alternatives to precisely solve many existing real-world power system problems. This study considers many hidden facts and phenomena.

To ensure accomplishing this ambitious task with some sorts of intelligence, advanced tools are involved; including traditional and meta-heuristic optimization algorithms and machine learning (ML) computing systems. Some superior hybrid optimization algorithms and ML computing systems are developed.

The mission starts from the Telegrapher's equations where the distributed- and lumped-parameter transmission line models are built on. The realization phase is done by considering the effects of surrounding weather, system frequency, load current, and cable design/status/age. Some innovative techniques are proposed to solve the inherent weaknesses in explaining the deviation in distributed series and shunt parameters of lines under sag. This realization is applied to enhance the solutions of PF, ELD, short-circuit analysis, power system stability and ORC problems.

To avoid tedious and highly time-consuming computational methods, a new set of optimization-free/modeling-free techniques are designed to solve ELD problems.

Through the realization and integration phases, many new innovative ideas are presented. Also, because many problems heavily depend on ML tools, so a new computing system is designed to achieve the accuracy and precision criteria without losing the explainability and interpretability criteria. That is, compromising between the strengths of classical linear regression (LR) and nonlinear regression (NLR) analysis and modern artificial neural networks (ANNs) and support vector machines (SVMs).

To judge the performance of each technique, many theoretical and real-world test systems and datasets are used with considering different scenarios and conditions.