Technical high -parallel neural network with multi -processors on the GPU NVIDIA GK110 for the control of a waste boiler

摘要

Successful learning (neural network) of controlling the combustion of an MVA,characterized by the use of the following measuring inputs and control outputs using the network structure (sheet 2 -Figure 3 network topology of the neural network):E: Input Measurement, A: Output Control,CCS: power plant identification system,x: Selection, /: not selectedKKSKlartextEAHAD10CP001Drum pressure xHBK10FT901Fire chamber xHHC00AS001Task Q Purpose of oil%/HHC21CP001Task Pressure P1 x /HHC22CP001Task Pressure P2x /HLA10CF001Primary air quantity xHLA20CP001001Primary air pressure xHLA31CP001Air pressure Roller 1x /HLA31FP901Vapour pressure under roller 1x /HLA32CP001Vapour pressure under roller 2 xHLA33CP001Vapour pressure under roller 3 xHLA34CP001 Vapour pressure 4 xHLA35CP001SVapour pressure under roller 5 -xHLA50CF001Secondary air pressure xHLA61CF001Secondary air pressure xHLA61CF001Secondary air quantity Backwall /HLA71CF901Secondary air quantity Front wall x /HNA10CQ001O2 Train /HNA20CQ001O2 Boiler x /HNA20CQ003CO Kesselende xHNA70CQ005CO Kamin /LBA10CP001FD -Druck /LBA10FF901FD -Quantity xLBA10FT901FD -Temperature xHHC00DU001Lastregler Curve Shift /HLB10AS001Drallklappe Primary Air / HLB10AS001DrallklAppendix Primary air /xHLA31AA001Air flap 1 /xHLA32AA001Air flap 2 /xHLA33AA001Air flap 3 /xHLA34AA001Air flap 4 /xHLA35AA001Air flap 5 /xHLA61DF001Secondary air flap Backwall /xHLA71DF001Secondary flaprwand /xHNA10DQ901O2 controller /LBA10DF001Lastregler Driver [t /h]/xThe associated system diagrams are defined as -Figure 2 Primary air supply 3 and secondary air supply 4.