Data Driven Operating Performance Assessment Method For Dense Medium Coal Preparation Process

Coal resources are an important pillar of our country’s economy and social development,providing power and support for the sustainable and rapid development of the economy.Our country has abundant coal reserves and its output ranks first in the world,but the proportion of raw coal selected is relatively low compared to other coalproducing countries in the world.Therefore,improving the utilization efficiency of coal resources is of great significance to ensure the sustainable development of our country.Dense medium coal preparation technology is widely used in coal preparation plants due to its high sorting accuracy,wide selection range,and easy implementation of automatic control.The main purpose of dense medium coal preparation is to remove gangue and impurities in raw coal,thereby improving the quality of raw coal.However,in the actual coal preparation process,due to a large number of disturbances and uncertain factors,the coal preparation process deviates from the set optimal working point as the washing process proceeds,resulting in a decrease in the stability of coal product quality.In order to ensure the stability of coal product quality and improve the overall economic benefits of coal preparation plants,it is of great significance to study the meticulous and robust operation performance assessment and non-optimal factor identification method of the dense medium coal preparation process.Based on the background of the dense medium coal preparation process,this paper establishes a dynamic simulation platform for the whole dense medium coal preparation process,and conducts the research of the operating performance assessment method for the densemedium coal preparation process on this basis,laying a good foundation for the realization of operation optimized control and production adjustment of the densemedium coal preparation process,and improvement of the quality of coal products.The main research content includes the following three aspects:(1)This subject is based on the dense medium raw coal preparation process.According to the technology and characteristics of the dense medium coal preparation process,the mechanism of each typical dense medium coal preparation process and the main factors affecting the separation effect of the dense medium coal preparation process are analyzed.On the basis of the existing research work,the whole process mechanism model of dense medium coal preparation including raw coal classification screen,coal medium mixing,cyclone seperation,dense medium recovery,dense medium deployment and other processes is established,and a simulation platform based on Simulink simulation environment is constructed for the whole process of densemedium coal preparation.Based on the whole process simulation platform,the changes in the main performance indicators of the whole process of the dense medium coal preparation process are analyzed through simulation and analysis of the changes in main operating variables.The simulation platform of the whole process of densemedium coal preparation provides a verification platform for the simulation of densemedium coal preparation process,the operating performance assessment method and the research of operation control algorithm.(2)Aiming at the strong nonlinearity and dynamic time-varying characteristics of the dense medium coal preparation process,a dense medium coal preparation operation performance assessment method based on the comprehensive economic index driven long short term memory(ILSTM)is proposed.On the basis of LSTM unsupervised feature learning,comprehensive economic indicator constraints are introduced to force the long and short-term memory network to learn the original data feature expression related to comprehensive economic indexes.When the operating performace is nonoptimal,this paper combines the structural characteristics of the long and short-term memory network and proposes a non-optimal factor identification method based on the ILSTM autoencoder contribution graph algorithm,which uses the contribution rate of variables to the non-optimal state to identify the process non-optimal cause.The data set generated by the simulation platform for the whole process of the dense-medium coal preparation process verifies its effectiveness.(3)Aiming at the problem of the limited sample of comprehensive economic index in the dense medium coal preparation process,this paper stack the LSTM network to a deep structure model,and proposes a dense medium coal preparation operating performance assessment method based on semi-supervised learning mechanism deep comprehensive economic index driven long-term short-term memory network(Deep ILSTM),so that the operating performance assessment model can use the limited comprehensive economic index data information to extract the complex process dynamic characteristics.LSTM can make full use of unlabeled data to mine the process dynamic characteristics in the process data,which is used to solve the problem of the operating performance assessmenf of the dense medium coal preparation process under the condition of limited comprehensive economic index data.The stacked DLSTM network is applied to the feature extraction of unlabeled data,and the labeled data is combined to constraints feature extraction process to construct DILSTM network.After the feature extraction is completed,the labeled data is used to train the performance assessment model to obtain the operation performance assessment model.Finally,the effectiveness and practicability of the proposed method are verified on the coal preparation process data set generated by the dense medium coal preparation process simulation platform.