Abstract:

There are often problems in geothermal wellbore that its production capacity of geothermal fluids decreases or even cannot be produced due to the scaling of geothermal fluids in it. Therefore, it is particularly important to study the scaling behavior, such as to predict the scaling location of geothermal fluid in the wellbore. Artificial neural networks (ANNs) can be used to develop a new model for predicting the scaling location in geothermal wellbore. Because it has no nature of mechanism modeling, it can only be used as a new proxy model. The three-layer ANNs structure was successfully trained using the temperature, pressure and well depth data of the geothermal fluids at the wellhead and bottomhole as input variables, and the appropriate accuracy of the ANNs proxy model was achieved with a relative error of less than 10%. The scaling locations predicted by the ANNs proxy model were analyzed and compared with the actual measured wellbore scaling locations, and the reasons for the errors were analyzed. The results showed that the new ANNs proxy model could serve as a practical tool to reliably predict scaling locations in geothermal wellbores.

Key words: geothermal fluids, scaling, neural networks, prediction, wellbore, precipitation

摘要：

地热井筒中常存在因地热流体结垢而导致的生产能力下降甚至无法生产的问题，因此研究地热流体在井筒中的结垢位置等行为具有重要的应用价值。人工神经网络（ANNs）可用于开发预测地热井筒中结垢位置新模型。由于其没有机理建模的性质，故只可作为一种新的代理模型。本文以地热流体在井口和井底的温度、压力以及井深等参数作为输入变量，成功训练了三层ANNs结构，以小于10%的相对误差实现了ANNs代理模型的合适精度。对ANNs代理模型预测的结垢位置进行了分析，并与现场测量的井筒结垢位置进行了比较，分析了产生误差的原因。结果表明，新建的ANNs代理模型可作为一种实用工具，能够可靠地预测地热流体在井筒中的结垢位置。

关键词: 地热流体, 结垢, 神经网络, 预测, 井筒, 沉淀