基于高斯混合模型的电厂入厂原煤煤质一致性分析

Consistency analysis of raw coal in power plant based on Gaussian mixture model

入厂原煤煤质监督是电厂重要工作之一,但煤质监督通常针对绝对煤质,而忽略了煤质稳定性。随着发电形势变化,电厂燃煤煤质波动受到关注。将先进数据挖掘技术应用于电厂煤质参数分析,建立了基于高斯混合模型的入厂原煤煤质一致性分析方法,能判别来煤煤质是否发生异常变化,推断来煤中是否可能包含其他未知煤种,并对该未知煤种参数和含量进行最大似然估计。将原煤标准煤质参数分布与来煤样本集作为已知条件;为了判别来煤煤质一致性,提出先验假设,假设来煤中包含标准原煤和未知煤种,2种煤的煤质参数看作服从不同正态分布的随机变量;利用高斯混合模型计算所有工业分析参数在来煤样本上的联合概率;根据最大似然原理,采用来煤样本集对未知煤种含量及分布参数进行估计;若未知煤种含量估计值很小,则认为来煤中只包含单一的标准原煤;若未知煤种含量估计值不可忽略,且未知煤种参数的估计值与标准原煤相应参数有显著差异,则认为来煤中不仅包含标准原煤,还包含未知煤种。采用上述原煤煤质一致性分析方法,对电厂实际来煤样本和模拟样本进行分析。结果表明,实际来煤样本煤质与标准原煤煤质一致,而模拟样本煤质与标准原煤煤质存在明显差别,模拟样本中包含40.9%的未知煤种,未知煤种参数估计结果与实际数据吻合较好。

The incoming coal supervision is one of the most important daily work of the power plant.Coal quality supervision usually focuses on absolute coal quality,ignoring the stability of coal.With the change of power generation situation,the fluctuation of coal quality in power plants is receiving more and more attention.The advanced data mining technology was applied to the coal quality parameter analysis in power plant,and the consistency analysis method of incoming raw coal based on Gaussian mixture model was established.This approaches can judge whether the incoming coal changes abnormally,infer whether the coal may contain other unknown coal,and estimate the parameters and content of the unknown coal by maximum likelihood estimation.The quality parameters of standard raw coal and sample set of the incoming coal were taken as the known conditions.In order to judge the consistency of incoming coal,a priori hypothesis was proposed.It is assumed that the incoming coal contains standard raw coal and unknown coal.The quality parameters of the two coals are regarded as random variables subject to different normal distributions.The Gaussian mixture model was used to calculate the joint probability of all industrial analysis parameters on all incoming coal samples.According to the maximum likelihood principle,the incoming coal sample set was used to estimate the unknown coal content and distribution parameters.If the estimated content of unknown coal is very small,it is considered that the incoming coal contains only a single standard raw coal.If the estimated content of unknown coal can not be ignored,and the estimated unknown coal parameters is significantly different from the corresponding parameters of standard raw coal,it is considered that the incoming coal contains not only standard raw coal,but also unknown coal.The above raw coal quality consistency analysis method was adopted to analyze the actual incoming coal samples and simulated samples in power plant.The results show that the coal in the actual incoming coal sample is consistent with that of the standard raw coal,while there is a significant difference between the simulated sample and the standard raw coal.The simulated sample set contains 40.9%of the unknown coal.The estimation results of unknown coal parameters are in good agreement with the actual data.