预测不同深度地热储层的温度是确定热储的热能储存量、热输出能力和可持续利用期限等参数的关键。针对不同约束条件下的热储温度预测问题,建立了一种基于粒子群优化算法(PSO)的长短时记忆网络(LSTM)的热储温度预测模型,对共和盆地恰卜恰地区地热井进行了预测,并通过与BP模型、LSTM模型的预测结果对比,验证该模型的有效性。结果表明,该模型预测结果的均方根误差(RMSE)、平均绝对百分误差(MAPE)、平均绝对偏差(MAD)值与BP、LSTM模型相比均最小,且RMSE最小值仅为1.192。该模型预测值与真实值的相关性系数为0.929,说明该模型的预测效果好,能实现地热系统储层温度的高效预测,为地热系统高效长久开发提供科学依据。

The temperature prediction of geothermal reservoirs at different depths is to determine the key parameters such as thermal energy storage,heat output capacity,and the sustainable utilization period of geothermal reservoirs. Taking the geothermal wells in the Qiabuqia area of Gonghe Basin as an example, this study proposes a temperature prediction model for heat reservoirs under different constraints based on particle swarm optimization (PSO) and long short-term memory network (LSTM). The prediction effect of this model is verified by comparing with those of the BP model and LSTM model. The results show that the RMSE value,MAPE value and MAD value in the prediction results of the model are the smallest compared with those in BP and LSTM models, and the minimum RMSE value is only 1. 192. The determination coefficient of the model is 0. 929,showing a good prediction effect. This OF MINING SCIENCE AND TECHNOLOGY indicates that this model could realize the prediction of reservoir temperature in geothermal system,which provides references for the efficient and long-term development of geothermal system.