Research Of Abnormal Energy Consumption Diagnosis Of Horseshoe-flame Glass Furnace Based On Improved Clustering

Glass production is a high-energy-consuming manufacturer,in which glass furnace is the indispensable facility in production,and its energy consumption is massive.The pass rate of glass products will decrease when the glass furnace is abnormal,resulting its energy lost directly.In this thesis,taking the glass furnace with horseshoe-shaped flame(horseshoe kiln)as the research object,its complex structure and scattered energy consumption of subsystem,so that abnormal causes of working condition has uncertainty.It is very challenging to implement anomaly detection of glass furnace due to we synchronously pay attention to the trend of change in the whole and local working status of it.And the glass furnace has been in full-load running state for a long time,the probability of abnormal occurrence will cumulatively increase with the running period.Therefore,it is one of the effective measures to detect and locate abnormal informations in time when anomaly occurs,which can ensure the pass rate of glass products,and then reduce energy consumption.At the present,traditional method of anomaly detection is implemented by threshold alarm,manually inspecting equipment and monitoring log system.However,on the one hand,it’s susceptible to subjective factors,and its inspection process is lagging.On the other hand,with redundantly reported in the event of anomalies,it’s difficult to capture core abnormal information.Thus,new requirements are put forward for the study of abnormal diagnosis(anomaly detection and location)in glass furnace.In this thesis,firstly established an energy balance model by analyzing the characteristics of structure and process flow in horseshoe kiln,a feature space with characteristics of energy consumption and a generalized hierarchy structure of hierarchical clustering are constructed based on this model.Next,the anomaly detection of energy consumption in horseshoe kiln is implemented by the density peak clustering based on artificial bee colony algorithm(ABC-DPC).Subsequently,the re-clustering process is realized based on an abnormal energy consumption data.In other words,the hierarchical clustering based on root cause analysis(RCA-HCA)can generalize abnormal energy consumption to accurately distinguish primary abnormal info.Ultimately,in view of the existing energy management system,anomaly detection model and anomaly location model are integrated to achieve the intelligent monitoring of energy consumption in glass furnace.The particular research details are as follows:1.Analyzing the characteristics of system structure and production technology,we can establish a macroscopical understanding of the mechanism in horseshoe kiln.Next,based on the principles of heat balance and mass balance,the local energy balance model of the horseshoe kiln and its subsystems is constructed.And then the heat balance of the energy balance model is used to construct a layered energy model,which offer the theoretical basis for constructing the feature space of the anomaly detection model and the generalized hierarchical structure of the anomaly location model.2.Traditional method in the kiln of anomaly detection is susceptible to interference,with reported redundantly in the event of anomalies.To solve this problem,based on ABCDPC algorithm,the anomaly detection model of energy consumption in horseshoe kiln was proposed.Concretely,in order to implement the anomaly detection of energy consumption in horseshoe kiln,the ABC-DPC algorithm uses feature vectors what is calculated and constructed by the energy composition of layered energy model to cluster.Among them,the ABC-DPC algorithm is an improved algorithm proposed in this thesis for the DPC algorithm,which addresses the drawbacks of DPC,such as manually setting of cutoff distance and inability to automatically divide cluster’s centers and outliers.3.For the sake of precise positioning of the anomaly subsystem,a method,anomaly positioning to energy consumption based on RCA-HCA algorithm,was designed.The model is based on abnormal energy consumption data,by continuously generalizing the feature attributes of energy consumption samples,and then constantly merging the same generalized samples until the condition of the most abstract generalization representation are met,the clustering is terminated.Outputting sets of the generalized abnormal energy consumption,the subsystem’s information associated with the characteristic of energy composition is matched to achieve anomaly location.4.Based on the above research,in order to achieve the intelligent manufacturing of glass enterprise,based on existing energy management platform,the anomaly detection and location module of horseshoe kiln was developed,with verifying its effectiveness in actual production.