초록 ▼

A method and apparatus for training and optimizing a neural network for use in controlling multivariable nonlinear processes. The neural network can be used as a controller generating manipulated variables for directly controlling the process or as part of a controller structure generating predicted

A method and apparatus for training and optimizing a neural network for use in controlling multivariable nonlinear processes. The neural network can be used as a controller generating manipulated variables for directly controlling the process or as part of a controller structure generating predicted process outputs. The neural network is trained and optimized off-line with historical values of the process inputs, outputs, and their rates of change. The determination of the manipulated variables or the predicted process outputs are based on an optimum prediction time which represents the effective response time of the process output to the setpoint such that the greatest change to the process output occurs as a result of a small change made to its paired manipulated variable.

대표청구항 ▼

A system for providing an apparatus for use in controlling a process at a desired setpoint level, the process having process inputs and outputs, said system comprising: (a) a data base of training patterns representing historical values of the process inputs and process outputs; (b) a set of future

A system for providing an apparatus for use in controlling a process at a desired setpoint level, the process having process inputs and outputs, said system comprising: (a) a data base of training patterns representing historical values of the process inputs and process outputs; (b) a set of future time steps defining a future time horizon; (c) a prediction model constructed with the training patterns contained in said data base, for predicting the process outputs over said future time horizon; (d) a sensitivity processor utilizing said prediction model for determining the effect in the process outputs as a result of changes made to the historical values of the process inputs, said sensitivity processor producing predicted process outputs; (e) a first processing element for computing a prediction time where a greatest value in the predicted process outputs occurs, the prediction time being an optimum prediction time; (f) a second processing element for computing the predicted process output as advanced by the optimum prediction time; (g) said apparatus comprising: (i) an input means for receiving input variables for operating said apparatus, the input variables comprising the historical values of the process inputs and outputs from said data base, optimum prediction times, and the predicted process outputs as advanced by the optimum prediction times; (ii) an output means for generating output variables for use in controlling the process in a preferred manner; and (iii) a processing means for mapping said input means to said output means, said processing means comprising a function for performing said mapping, said function determining an optimum prediction time, and a predicted process output for performing said mapping, the optimum prediction time representing an effective response time of the process to a change in a desired setpoint level and the predicted process output representing a process output as advanced by the optimum prediction time; (h) a training system for training said apparatus in accordance with a training algorithm; and (i) a third processing element for configuring said apparatus to receive all of said input variables and for operating said training system to train said apparatus with all of said input variables thereby producing output variables for use in controlling the process.