Use of dynamic programming and particle swarm optimization techniques for solving security and unit commitment problems

Abstract

Power system security and Unit Commitment are two important tasks in power system operation. Power system security is the ability of the system to continue supplying power to consumers in spite of faults and occasional breakdown of some equipment. Power systems are designed with generation capacities larger than peak demands so that the total demand at any time can be met even in the event of failure of one or two largest generating units. The generating units have to run continuously to meet the load demand. Unit Commitment is the task of selecting the best generating unit(s) to be committed. The committed units must satisfy the fuel cost and CO₂ emissions constraints. In this thesis, we considered a small power system (microgrid) consisting of ten subsystems. Although, each subsystem contains Photovoltaic, battery and Micro Gas Turbines, their capacities are not identical. A Microgrid Central Energy Management System is utilised for solving the security and Unit Commitment problems. This energy management system is a MATLAB program written for the purpose of solving the above two problems. In the simulation, we first used Dynamic Programming for solving security and Unit Commitment problems. A fault was applied to one of the subsystems causing a loss of some generation. Dynamic Programming has been able to correctly identify the fault and allocate the necessary amount of generation from another subsystem to compensate for the lost power. Being cheap and environmentally friendly, Photovoltaic should be the primary generation allocated even if they belong to a different subsystem. Unfortunately, the Dynamic Programming allocated Micro Gas Turbines in some cases even though sufficient Photovoltaic generation was available. This is due to the fact that the Dynamic Programming provides only a local solution instead of a global solution.

Dynamic Programming is not, therefore, totally acceptable for solving the security and Unit Commitment problems. Next, we resorted to a second technique i.e., Particle Swarm Optimization. We applied a fault to a subsystem just as we did in the concept of Dynamic Programming. This technique also correctly identified the fault and allocated the available Photovoltaic generation from another subsystem to compensate for the lost generation. Particle Swarm Optimization used Micro Gas Turbines only when all Photovoltaic generation from all the subsystems has been used. This technique has satisfied both the security as well as Unit Commitment problems in an acceptable manner. The Particle Swarm Optimization has also been able to correctly identify the faults. When, we applied two random faults in the same or different subsystems. This technique has also recommended necessary Photovoltaic generation from another subsystem. It is obvious that the Particle Swarm Optimization was accurate and more acceptable than the Dynamic Programming for solving security and Unit Commitment problems.

Date of Award	Jul 2019
Original language	English
Supervisor	Kamal Debnath (Supervisor) & Friso De Boer (Supervisor)