利用生成对抗网络实现地热能开发中的储层温度预测优化

Utilizing Generative Adversarial Networks to Optimize Reservoir Temperature Prediction in Geothermal Energy Development

摘要: 地热是一种分布广泛、稳定、清洁的零碳排放能源，而孔隙型热储是最重要的地热资源类型之一，从对热储中提取出来的热量，可用于地热发电站发电、供暖等。数值模拟可以用于评估地热自然状态的生产场景。前人在地热数值模拟上已经展开大量的研究。然而数值模拟通常需要大量的计算资源和时间。与数值模拟相比，代理模型不仅具有与数值模拟近似的精度，而且花费的时间更少，这是显著优于数值模拟方法的。在本研究中，基于GAN开发了一个三维代理模型来预测储层的温度分布，模型以渗透率分布作为输入，以时间作为条件，最终输出储层不同时间的温度分布。最终的结果表明，代理模型可以非常好地预测储层温度分布。

Abstract: Geothermal energy is a widely distributed, stable and clean zero-carbon emission energy source, among which pore heat storage is one of the most important types of geothermal resources, and the heat extracted from thermal storage can be used for power generation, heating and cooling in geothermal power stations. Numerical simulations can be used to evaluate production scenarios in the natural state of geothermal energy. A large number of studies have been carried out on geothermal numerical simulation. However, numerical simulations often require a lot of computational resources and time. Compared with numerical simulation, the proxy model not only has the accuracy of the same as the numerical simulation, but also takes less time, which is significantly better than the numerical simulation method. Based on GAN, a three-dimensional proxy model is developed to predict the temperature distribution of the reservoir, and the model takes the permeability distribution as the input and the time as the condition, and finally outputs the temperature distribution of the reservoir at different times. The final results show that the proxy model can predict the reservoir temperature distribution very well.