Research On Image Recognition Of DCS Screen Of Power Station Simulator

With the rapid development of computer technology and automation control technology,a wave of intelligence is sweeping in,which cause various industries are actively exploring new working modes based on automation and intelligence.The use of image recognition,voice control and other technologies simulating the behaviour of the central control operator,to achieve humanoid operation of the DCS(Distributed Control System)screen of the power station simulator that provides new ideas for the upgrading of power plants and the construction of smart power plants.The thesis addresses the current situation that the power station DCS system relies on the manual operation and judgment of equipment operation status by the central control operator in the working process,which leads to a large change in personnel requirements under different working conditions and is easily affected by the subjective factors of the operator.The research of key technologies such as DCS screen target feature extraction and detection and identification is carried out to achieve high-precision detection and identification of power plant simulator DCS screen titles,equipment and other targets,and complete the power plant simulator DCS screen equipment sequence control.Based on the detailed analysis of the work content and working mode of the central control operator of the power plant,the operation behaviour of the central control duty officer on the DCS screen of the power plant simulator is abstracted into the sequential control processes of screen title reading,screen equipment status observation and equipment operation.Through the above analysis the DCS screen image recognition system of the power plant simulator is designed.The YOLOv3-RP algorithm is proposed to achieve real-time detection and identification of DCS interfaces in power plant simulators.The algorithm is based on the YOLO(You Look Only Once)algorithm,which fuses relative positions to achieve accurate recognition of multiple identical devices on the same interface,regardless of image size.The SIFT-HC algorithm is proposed for similar devices,fusing the colour histogram vector,scale invariant feature conversion algorithm local feature vector and image colour moment vector,reducing the dimensionality of the fused vectors with the help of principal component analysis to generate new feature vectors for feature point matching and image similarity calculation.Human-computer interaction and humanoid operation is achieved through voice recognition,operation capture and reproduction and other technologies.The system is designed to provide logical support for the sequential control of the equipment by designing the equipment sequence control logic based on the operation protocols.Finally,the designed system is commissioned and tested.The system was tested one by one in accordance with the operation sequence of the duty officer in the DCS screen,imitating his operation behaviour.The test results show that the system is able to accept voice commands,accord to the pre-designed sequence logic for the corresponding screen titles,equipment operation status judgement and operation,etc.,record the operation history and feedback the operation results,and complete the operation tasks.