A Non-linear Control Model Of Ultra-supercritical Unit

Prerequisites for control optimization of ultra-supercritical unit are to analyze the characteristics of ultra-supercritical unit in depth, build the model for control and obtain accurate signals. The model for control should be simple, and possess favorable accuracy. Some state variable signals needed by control optimization cannot be obtained with traditional methods. Those signals could be gotten through constructing their relationships with other known signals. The constructed signals should be accurate and easy to realize. To achieve the goals and take modeling as paramount, the following works have been carried out in this thesis.A scheme based on hybrid modeling method for control model of ultra-supercritical unit is proposed. Models built with single modeling method always have high complexity, low accuracy or low commonality. However, the hybrid modeling method could take advantage of both methods. When using the hybrid modeling method, there is overcompensation or undercompensation of data information for model structure. In order to get the maximum use out of data information, and simplify the structure of control model to the fullest extent as well, the best model structure was defined and control model built problem was therefore transferred to the best structure built problem.The simplest model structure containing the most effective information is given which is critical for obtention of the best model structure. In order to evade determination of the transition point position in an effective way and minimize the possible number of compartment with fixed boundary when deriving a general lumped compartmental model structure that respects the conservation of energy, momentum, and mass, suitable lumped parameters are chosen for the compartment of the boiler including two-phase region. And then another lumped model structure was developed with the sub-minimal number of compartment in the same way.The control model of ultra-supercritical unit possessing the most simplified model structure is built. A data selection method for open-loop identification based on model structure was developed and close-loop identifiability of power unit was proved as well. Parameters of models were determined with close-loop operation data of Taizhou1000MW ultra-supercritical unit and open-loop experimental data of Pinghai1000MW ultra-supercritical unit. The models were verified by data comparison experiments. Models with different structures were compared in accuracy and the ratio of accuracy and structure complexity of each model was analyzed. Results indicate that the model with the most simplified structure has the highest ratio of accuracy and structure complexity and it can be used as control model. Finally, based on multi-sensors information fusion technique and mechanism analysis, reheat steam flow signal, coal calorific value signal and oxygen content of exhaust gas signal are constructed for control optimization of ultra-supercritical unit using two important state signals constructed in the control model which are main steam flow signal and boiler heat signal. The constructed signals were verified with operation data.