31-03-2014, 03:49 PM
CLEAN COAL TECHNOLOGY: POWER PLANT OPTIMIZATION AND DEMONSTRATION
INTRODUCTION
Overall optimization of a coal-fired power plant is a highly complex process. One must first decide what constitutes optimal performance. Obvious answers include maximum thermal efficiency, lowest possible emissions, lowest possible cost, readily marketable by-products, and maximum system availability for power generation. In reality, these goals—and others—are interrelated. In some cases, however, these optimization goals are at odds with each other. For example, high excess air will result in better carbon burnout and less carbon monoxide but will also result in higher emissions of nitrogen oxides (NOX). These interactions must be kept in mind and addressed with any optimization program.
There are a number of relatively fixed items that affect overall plant operation. These include boiler design, cooling water conditions, burner type, design steam conditions, and environmental control systems that capture and remove particulate matter, sulfur dioxide (SO2), NOX, and mercury. Coal quality is also a major factor that affects plant performance. High moisture and/or ash content decreases efficiency and increases wear and power requirements on the pulverisers. High sulphur content results in more reagent consumption and increased by-product generation.
ARTIFICIAL INTELLIGENCE
Artificial intelligence (AI) is commonly defined as the science and engineering of making intelligent machines, especially intelligent computer programs. Relative to applications with coal-fired power plants, AI consists of aspects or considerations that deal with the following:
• Neural networks, which mimic the capacity of the human brain to handle complex nonlinear relationships and “learn” new relationships in the plant environment
• Advanced algorithms or expert systems that follow a set of pre-established rules written in codes or computer language
• Fuzzy logic, which involves evaluation of process variables in accordance with approximate relationships that have been determined to be sufficiently accurate to meet the needs of plant control systems
Neural networks (NNs) are a class of algorithms that simulate the operation of biological neurons. The NN learns the relationships between operating conditions, emissions, and performance parameters by processing the test data. In the training process, the NN develops a complex nonlinear function that maps the system inputs to the corresponding outputs. This function is passed on to a mathematical minimization algorithm that finds optimum operating conditions.
LIGNITE FUEL ENHANSMENT PROJECT
INTRODUCTION
The use of low-rank coals (lignite and sub-bituminous) has seen a significant increase in recent years. Because of the low sulphur content of such coals, many units have adopted fuel switching to meet sulphur emissions specifications, and other units have been built specifically to burn low-rank coals. However, a major disadvantage of low-rank coals is their high moisture content, typically 25 to 40 percent. When such coal is burned, considerable energy is required to vaporize the moisture it contains, thus raising the heat rate of the power plant and lowering its efficiency.
Fuel moisture has many effects on unit operation, performance, and emissions. As fuel moisture decreases, the fuel’s heating value increases so that less coal needs to be fired to produce the same electric power, thus reducing the burden on the coal-handling system. Drier coal is easier to convey as well, which reduces maintenance costs and increases availability of the coal to the handling system. When the crushed coal is gravity-fed into bunkers, the drier coal flows more readily than the wet coal, causing fewer feed hopper bridging and plugging problems. Drier coal is easier to pulverize as well so that less mill power is needed to achieve the same coal fineness. Finally, with less moisture in the fuel more complete drying of coal can be achieved in the mill, which results in an increased mill exit temperature, better conveying of coal in the coal pipes, and fewer coal pipe plugging problems.
CONCLUSION
The major conclusio from this project is that the Pegasus Technologies NN-ISB control system is a sound idea with significant potential. The Big Bend project successfully demonstrated a neural network, closed-loop operation on a full-scale boiler without causing unit upsets or violating any constraints and it also achieved operator acceptance. The NN-ISB appears to provide generating companies with an integrated solution that will assist in optimal economic and environmental real-time, online operation of a unit.
The NN-ISB is modular in design and can be readily applied to a variety of power generating units. The solution architecture and infrastructure are designed to allow full or staged deployment, depending on technology applied throughout allows unit flexibility (i.e., existing systems can be integrated within the overall solution) and is extendible (new modules/new equipment can be readily modelled and incorporated).