Neural network prediction control method based on optimization control platform

Abstract

The invention discloses a neural network prediction control method based on an optimization control platform, and the method comprises the steps: A, taking samples of a controlled object in a certaininput and output range, and building an initial learning set of a neural network; B, selecting initial parameters of the neural network according to the control object, wherein the initial parameterscomprise the layer number and the node number; C, initializing a neural network, mainly initializing a network weight and a threshold value; D, training an LSTM neural network to obtain a neural network prediction model; E, calculating a reference track expected to be input; F, outputting by the neural network prediction model, and generating prediction output through feedback correction; G, calculating a prediction error; and H, solving a quadratic performance function to obtain an optimal control solution, and skipping to the step E so as to continuously adjust the control signal. Defects inthe prior art can be overcome, the method can be directly applied and implemented in a thermal power station, and the problems faced by an existing advanced algorithm thermal control platform are perfectly solved.

Description

technical field [0001] The invention relates to the technical field of thermal control, in particular to a neural network predictive control method based on an optimized control platform. Background technique [0002] With the rapid development of new energy power in my country and the change of energy structure, the operating environment faced by conventional coal-fired thermal power plants is becoming increasingly complex. , The optimal control of time-varying objects cannot meet the requirements of environmental protection, peak shaving, production and other indicators of thermal power plants in the new era. In this context, the implementation of advanced algorithms in the field of thermal engineering has become the general trend. [0003] At present, various advanced control algorithms emerge in an endless stream, but most of them stay in the stage of simulation experiments, and a small number of algorithms have been implemented in engineering applications, but there are...

AI Technical Summary

Problems solved by technology

[0003] At present, various advanced control algorithms emerge in endlessly, but most of them stay in the simulation experiment stage, and a small number of algorithms have been implemented in engineering applications, but there are a series of problems such as unsatisfactory control effects, long development cycle, and difficult maintenance.

Images