基于自动谱聚类与多极端学习机模型的油井油液含水率软测量

油井油液的含水率是石油生产中的一个重要参数，及时、准确的测量对提高采油生产效率具有重要的意义。针对传统人工测量所存在的不足，引入软测量技术，建立基于自动谱聚类与多极端学习机（automatic spectral clustering-multiple extreme learning machines, ASC-MELM）的软测量模型。提出一种自动谱聚类（spectral clustering, SC）算法，由改进的萤火虫算法（firefly algorithm, FA）对聚类数目和尺度参数进行优化选取，所提出的改进萤火虫算法（improved firefly algorithm, IFA）采用以一定概率跳出当前解的机制，避免传统FA过早陷入局部最优解的不足；对聚类后的不同训练子集，分别由极端学习机（extreme learning machine, ELM）建立子模型，由IFA对其中的隐含层输入权值、隐含层神经元的偏置和隐含层节点个数进行优化选取；最后，将多个子模型的结果取加权平均值输出。由国内某油田作业区一口生产井进行实例验证，结果表明所提出方法具有较高的预测精度，对于实现油井油液含水率的软测量是合理有效的。

Soft sensor method for moisture content of well oil based on automatic spectral clustering and multiple extreme learning

Moisture content of the well oil is a key production variable in the oilfield, and it has great significance for improving the oil production efficiency by timely and accurate measurement of it. In order to overcome some deficiencies of the traditional manual measurement, the soft sensor technology is introduced to establish a soft sensor model based on automatic spectral clustering - multiple extreme learning machines (ASC-MELM). An automatic spectral clustering (ASC) algorithm is proposed and an improved firefly algorithm (FA) is applied to reach an optimal selection of the clustering number and the scale parameter. The proposed improved firefly algorithm (IFA) adopts a mechanism of jumping out of the current solution at a certain probability, which can avoid the deficiency of falling into the local optimal solution earlier of the traditional FA. For different training subsets after the clustering, the multiple extreme learning machines (ELM) are adopted to establish the different sub-models, in which IFA is used to reach an optimal selection of the hidden layer input weights, the hidden layer biases and the number of the hidden layer nodes. Finally, the output is obtained by calculating the weighted average of multiple sub-models. An application example of an oil well in a domestic oilfield is given. The simulation results show that the proposed method has better predicted accuracy and it is reasonable and effective to realize the soft sensor for moisture content of the well oil.