This paper aims at developing an artificial intelligence model to forecast the water yield of a modified solar distiller integrated with evacuated tubes and an external condenser. The model consists of a hybrid long short-term memory (LSTM) model optimized by a moth-flame optimizer (MFO) used as a subroutine to obtain the optimal internal parameters of the LSTM model that maximize the forecasting accuracy. The model performance was compared with that of the standalone LSTM model. Both developed models were trained and tested using experimental data of the modified distiller and a conventional distiller. The thermal performance of both distillers is also compared in this article. The maximum daily distillate output achieved for the modified distiller was 3920 l/m². The forecasted data of both models were compared using several statistical measures. For all measurements, LSTM-MFO outperformed standalone LSTM. The determination coefficient of the forecasted data using LSTM-MFO reached a high value of 0.999 for both solar distillers.

本文旨在开发一种人工智能模型来预测集成有真空管和外部冷凝器的改进型太阳能蒸馏器的产水量。该模型由一个由蛾焰优化器 (MFO) 优化的混合长短期记忆 (LSTM) 模型组成，用作子程序以获得 LSTM 模型的最佳内部参数，从而最大限度地提高预测精度。将模型性能与独立的 LSTM 模型进行了比较。两种开发的模型都使用改进的蒸馏器和传统蒸馏器的实验数据进行了训练和测试。本文还比较了两种蒸馏器的热性能。改造后的蒸馏器的最大日馏出物产量为 3920 l/m ². 使用几种统计方法比较了两种模型的预测数据。对于所有测量，LSTM-MFO 的表现优于独立的 LSTM。对于两种太阳能蒸馏器，使用 LSTM-MFO 预测数据的决定系数都达到了 0.999 的高值。