38 results about "Multi model predictive control" patented technology

The invention discloses an extended state observer-integrated supercritical thermal power unit turbine-boiler coordinated control method. According to the method, firstly, a plurality of local linear models are established at different working condition points of a power unit through the nonlinear analysis method according to the nonlinear model of a 1000 MW supercritical thermal power unit. Secondly, based on the plurality of local linear models, an extended state observer-integrated multiple model predictive controller (MMPC) is designed. Through the introduction of an extended state observer, the disturbance of the power unit in different channels and the model uncertainty are estimated and compensated. Meanwhile, the disturbance rejection performance of the predictive controller is improved. According to the technical scheme of the invention, based on the above algorithm, the large-scale tracking function for the load of the power unit is realized on the premise that the global stability of the control system is ensured. Meanwhile, the good disturbance rejection performance is realized.

The invention discloses a subcritical coordinative control method based on multiple model predictive control. The method comprises the following steps: determining a plurality of preset local prediction models according to a non-linear work condition range, as well as dynamic responses of main steam pressure and power of a subcriticla machine set; designing a controller for each local prediction model, selecting a control variable, and obtaining an output increment of each local prediction model through an interpolation formula; and during each control period, obtaining an actual control output increment according to an output increment of each controller in a weighing manner; and correcting a feed-forward channel. By adopting the method disclosed by the invention, control output during a transition period can be smoothed, as a result, load and pressure response characteristics of the machine set can be improved effectively; and the subcritical coordinative control method based on multiple model predictive control can adapt to large-scale change in work conditions.

The invention discloses a nonlinear suppression method for solid oxide fuel cells based on multi-model predictive control. The nonlinear multi-variable coupling problem of solid oxide fuel cells is solved by online multi-model weighted predictive control. A feedforward method is adopted to compensate for the disturbance of a resistive load to achieve disturbance restraint. At the same time, constraints on hydrogen flow the air flow are added to performance indexes, and the feasibility of a control system is achieved through a feedback correction link based on a Kalman filter, and the control effect is improved. It is ensured that the fuel cells maintain a constant output voltage after external resistance changes, and the fuel utilization rate is within a reasonable range of 0.7 to 0.9.

The invention discloses a multi-model predictive control design method for a coordinated control system of a supercritical thermal power unit based on a space metric algorithm, which is used for the control of the coordinated control system of a supercritical coal-fired unit and belongs to the field of thermal power engineering and automatic control. At present, the coordinated control system of machines and furnaces in most power plants is based on the conventional PID design. When operating in a wide range of variable operating conditions, the characteristics of large hysteresis, nonlinearity, and time-varying characteristics of the controlled objects in the machine-boiler coordination degrade the performance of the coordinated control system, resulting in the unit running During the process, problems such as low load regulation rate, poor precision, and large fluctuations in main steam pressure and steam temperature occurred. The present invention appropriately simplifies the traditional predictive control to realize the simplified design of linear predictive control, secondly combines the space metric algorithm to realize the selection of specific sub-models, and finally proposes a weighted soft switching strategy with simple calculation for the synthesis of multiple sub-controllers, which can significantly Improve the load regulation performance and unit operation stability of the 660MW unit to meet the needs of the project.

The invention discloses a multi-model intelligent optimizing prediction control method for boiler loads under the low-load condition. The method specifically comprises the steps that first, load points are selected to establish a controlled object model; second, a prediction controller is designed at all the load points according to the controlled object model, and thus the optimum control increment of the prediction controller is obtained; third, if the optimum control increment in the second step meets the constraint conditions, no treatment is required; and if not, a particle swarm algorithm needs to be adopted to search for the optimum control increment; and fourth, multi-model prediction control is adopted, and an improved recursive Bayesian weighting algorithm is used for weighing the output of all sub controllers according to deviation of the output of all sub models and the actual output. Through the self-adaptation intelligent optimizing control technology, stable and safe operation of a unit under the low-load coal supply economical blending combustion condition is achieved.

A multi-model generalized predictive control system based on dynamic optimization comprises a dynamic optimization layer, an MPC layer and a basic control layer. The dynamic optimization layer is located in the upper layer and calculates the optimization value of a critical control variable as the optimal set value of the MPC layer; the MPC layer is located in the lower layer and adjusts a to-be-optimized variable based on a rolling optimization prediction algorithm on the condition that the to-be-optimized variable satisfies a model dynamic behavior, thereby tracking the optimal set value obtained in the S1; the basic control layer is located in the bottom layer and transmits the final optimization value of the to-be-optimized variable to an actuating mechanism. According to the present invention, the cost consumption of the system is reduced, the economic benefit of the system is improved, the transient performance and the model parameter jump adjusting capability of the system can be improved, at the same time, the interference of disturbance on the output of the system can be eliminated effectively.