Research And Application On Air-conditioning System Detection Technology Based On Soft-sensing Method

With the development of science and technology, testing technology and methods have been rapid progress, changing from traditional analog instruments, digital instruments, intelligent instruments to today’s era of virtual instruments. However, there are still many important process parameters can not be measured directly or difficult to use sensors or process instrumentation in the actual testing process, the Soft-sensing technology is an effective way to solve the problem of such tests.This paper study from chiller condenser fouling resistance coefficient and IPLV, using two ways of the Soft-sensing technology to achieve chillers cleaning the condenser fouling resistance line monitoring and alarm and IPLV detection, which are based on the mechanism of traditional modeling and data-driven modeling.First, establish the theoretical model from the basic principles of heat transfer, obtain identify fouling resistance and analysis of its impact. On this basis, then improve the theoretical model, establishe different chillers logarithmic mean temperature difference defined fouling resistance test model. Discuss the scope of the two models and factors affect the relationship between the various, meanwhile analysis the model error. Focused on the relationship between the condenser heat transfer coefficient, load rate, condenser end difference, and fouling resistance, provide theoretical support for the condenser fouling resistance line monitoring.Based on the above discussion, establish a magnitude of the attenuation-based refrigeration condenser fouling resistance cleaning alarm model, which set different alarm limits to chillers depending on the actual operating conditions, achieving fouling classification alarm, and avoid the effects due to changes in operating conditions and performance of the unit itself lead to changes in the condenser. This model can be used to line monitoring in different load factor, different units.Write "Fouling chiller condenser cleaning alarm system" in Lab VIEW2012Platform, achieving the transformation of research results software applications. Carry out uncertainty analysis performed to alarm system established for cleaning. In the case of relative uncertainty of10%or less, each measuring instrument allows uncertainty could be certain, so the associated sensor accuracy limited.Finally, process the actual project monitoring data, comparing four BP model training period to finalize the BP modeling adaptive learning rate adjustment, and then establish a structure for the5×11×1of IPLV detection model. By contrast checking and comparing the actual unit performance curves of four models, the model established by proof has higher prediction accuracy, so provides a convenient way for obtaining IPLV.