Development of Coal Mill and Aggregate Load Area Models in Power Systems using Genetic Algorithms

• Main Author: Wei, Jianlin
• Published: University of Liverpool 2007
• Subjects: 621.402
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.485997

Quality and reliability is one of the most important issues. in power generation and distribution. With the recent advances in computer and network technology, the Operational Information Systems (OIS) have been installed in almost all power plants and substations. The data stored in databases covers long periods of time, which presents a challenge as how to extract the useful, task-oriented knowledge from the data to improve power system reliability and power quality. The thesis presents the research work in development of mathematical models for power systems by analysing the data available from on-site measurement using evolutionary computation techniques. The project contributes to aspects in power generation and distribution: coal mill modelling and electrical load area modelling. Coal-fired power stations are now obliged to vary their outputs in response to changing electricity demand and are required to operate more flexibly with more varied coal specifications. The operations of a mill need to be controlled to respond effectively to changes in plant load and coal quality. Combustion optimization relies heavily on optimization of the mill output. Frequently start-ups and shut-downs of mills bring the impact on power plant to achieve both low NOx and CO2 emissions. Operational safety and efficient combustion require better understanding to the milling process. The work described in the thesis has three new contributions: 1) Development of an improved normal grinding coal mill process model which provide more accurate prediction of mill states than the previous version; 2) Development of a new multi-segment coal mill model which covers the whole milling process from start-ups to shut-downs; 3) Development of a prototype software programme to implement the multisegment mill model on-line. The software has been passed to RWEnPower PIc. for further test. Stable operation of a power system depends on the ability to continuously match the electrical output of generation units to the electrical load. So it is important to have a reliable mechanism to predict the power load on time. Model based approach is one of the options. With the sponsorship from the National Grid Transco PIc, a study of modelling electricity area load has been carried out through this project A methodology using evolutionary computation techniques based on system measurements to construct power system area load models and achieve distribution network reduction is proposed in the thesis. Three aggregate load area model (ALAM) approaches entitled Voltage-Two-Step, Current-Two-Step and Direct-OneStep have been studied in the thesis. Simulations studies are carried out for these three approaches, and it found that the Direct-One-Step offers the best performance among the three ALAM approaches. Verification studies are performed through the project and some rules for constructing a good ALAM are obtained.