| Energy saving generation dispatch is an important part of the electricity industry. In thisthesis, researches are concentrated on two models of short-term scheduling and AGCreal-time control. A new evolutionary algorithm: group search optimization algorithm GSO isproposed to solve the short-term scheduling with minimum coal consumption objective.Currently AGC real-time control is a difficult problem. This thesis studies CPS assessment ofinterconnected power system with the AGC command optimal allocation to achieve energysaving. And some artificial intelligent algorithms based on reinforcement learning are used onthe AGC system real-time control.Firstly, group search optimization algorithm GSO is proposed to solve the short-termscheduling with minimum coal consumption objective. Tested models use3units system and10units system. Computed results show that, the coal comsumption of GSO is0.22%lowerthan GA in the3units system; and in the10units system GSO is1.1%lower than GA.Then the real-time generation control is the key point of energy saving generationdispatch. AGC command dynamic allocation can reduce coal consumption. Q(λ) algorithmbased on reinforcement learning is proposed to apply on the IEEE two area power grid. Thecoal comsumption results of one-day load curve show that, Q(λ) algorithm is0.49%lowerthan the traditional algorithm.Finally, hierarchical Q(λ) algorithm is proposed to solve the curse of dimensionality.Upper dispatch is to classify the total power plants and lower dispatch is applied in one of theclassification group. The simulation results show that, hierarchical Q(λ) algorithm reduces thedimensions and the coal consumption is0.564%lower than traditional algorithm.The thesis is supported by the National Natural Science Foundation of China (50807016,51177051), the Fundamental Research Funds for the Central Universities （2012ZZ0020），State Key Laboratory project of Tsinghua University （SKLD10KM01）. |
PDF Full Text Request