Study And Application Of Predictive Control And Power Load Forecasting Base On The DWO Algorithm

Generalized Predictive Control as a new predictive control algorithm, the main results are based on linear system controlled object, in theory and application of research results are good. Firstly, on the theory of generalized predictive control carried out a systematic introduction, starting from its basic principles, their algorithm has the characteristics of the process and the discussions are given, also discussed the further improvement of the generalized predictive control algorithm. Although the generalized predictive control algorithm has less model parameters, robust control effect is good, but for strongly nonlinear systems, as difficult to obtain accurate models in this way can hardly be good control effect. An algorithm is called direct weighted (DWO) optimal is proposed in this time for predictive control for nonlinear systems. Its principle is to first that by observing the system output, then assume a linear estimator, and then optimized by a standard function and the weight estimator, and then fitting out of the system output. The new algorithm ideas and studied the importance of nature to have given a lot of inspiration, such as the identification of the necessary characteristics of a small sample can avoid the requirement of large sample constraints in other identification algorithm . The simple form of linear estimator, compared to support vector machines to obtain high-dimensional complex simple process model , its approach too easy, there is in itself of the nature of the limited bandwidth, its computational capacity can be optimized. Note these features are directly weighted optimal (DWO) identification algorithm that has good research value, their future will be in place in the field of nonlinear system identification, it can become a competitor with a neural network, support vector machines and other powerful identification algorithm. This thesis first introduces the direct weighted optimal (DWO) algorithm for the basic principle and algorithm steps, as thesis research needs also to its nature as a given the limited bandwidth of introduction, in order to verify the effectiveness of the algorithm in this thesis also selected a complex nonlinear function of its simulation study. After completion of the above work , the paper use of this algorithm identified a unknown system model and using Taylor series expansion , transformed into a linear form of intuitive and effective, while obtained using linear prediction control predictive control law and avoids the complicated nonlinear optimization, simulation results show the feasibility and effectiveness. Load forecasting is one of the important work of the electricity sector, accurate load forecasting can be economic and reasonable arrangements for the internal power generating units start and stop, to maintain security and stability of the grid, reduce unnecessary spinning reserve capacity, reasonable arrangements unit maintenance plan to ensure normal production and life of society, effectively reducing electricity costs and improve economic and social benefits. Based on the discussion of the traditional power system short term load forecasting principle, this thesis first described some of the features it has, then load forecasting general steps was discussed on the relevant issues, finally directly weighted optimal (DWO) was used in the short-term load forecasting, the first of which based on statistical method, its abnormal load data of the historical data was corrected, and then using direct weighted optimization ( DWO) algorithm and relevant information has been properly handled forecast the load, because it has good identification ability . This paper presents the calculation of concrete steps, through the load of a particular area, without considering into weather and other factors, its result for short term load forecasting is effective.