Study Of Automatic Generation Control Strategy And The Units Economical Deployment

With the continuous expansion of the grid scale, the interregional connections become closer, the system structures become more complex, the uncertain disturbances increase significantly and it has become a concern and key subject to ensure the safe, stable running of interconnected power grid. To keep power grid functioning smoothly, its main aspect is to maintain power grid frequency and interregional wire power. The important function of Automatic Generation Control is to keep power grid frequency around some key value, and maintain the connection wire power near the planned value. This thesis has two directions towards AGC research: 1.Strategy for Automatic Generation Control, which is design of AGC controller; 2.Research for economical AGC units deployment.1.Put forward to adopt model-free adaptive control algorithm driven by data, which only use input, output and measured data of influence quantity from closed loop-controlled AGC system to design the AGC controller, but no need to know any inner structure and parameter information from the AGC model. It can compress all possible complex features, like nonlinearity, time-varying parameter and structure, etc, into a new variable named false partial derivative. It would enable model-free adaptive control only by adjusting the false partial derivative online. The model-free adaptive control includes three control ways: tight, partial and full, this text use full model-free adaptive control to fully dig into the hiding information. By comparison and analysis of calculation examples, it verified this arithmetic has fine nonlinear adaptation, obvious robustness and favorable Control Performance Standard(CPS) index.2.Put forward to take advantage of binary coding feature of genetic algorithm when the unit chooses variable from 0 to 1, the genetic algorithm brings genetic operator in the algorithm simulating plant growth. During iterative process, the algorithm adopts and keeps the optimal combination but later intersects and groups with the worst combination. This strategy can expand the scope for the superiors while accelerating the rate of convergence, which suppress the excessive premature and local optimum. It would better expand the scope for optimum if we sort the unit by Kmeans clustering algorithm while choosing higher mutation probability in the same unit. It would realize equal power distribution in the same unit when we work out the optimal power distribution value of the current combination by linear programming method. As for multiple optimal solutions, we can take the arithmetic mean to equally deploy the unit. By comparison and analysis of calculation examples, it verified the algorithm is effective and superior.