Integrate deep learning and physically-based models for multi-step-ahead microclimate forecasting

作者：

Highlights：

• The proposed novel ConvLSTM*CNN-BP fuses deep learning and physically-based models.

• ConvLSTM*CNN-BP gives accurate multi-factor/multi-horizon microclimate forecasts.

• ConvLSTM*CNN-BP overcomes the curse of dimensionality of ensemble big-data.

• ConvLSTM*CNN-BP deeply extracts ensemble data features to facilitate forecasting.

• Results show IoT data can be replaced by open data in greenhouse control.

摘要

•The proposed novel ConvLSTM*CNN-BP fuses deep learning and physically-based models.•ConvLSTM*CNN-BP gives accurate multi-factor/multi-horizon microclimate forecasts.•ConvLSTM*CNN-BP overcomes the curse of dimensionality of ensemble big-data.•ConvLSTM*CNN-BP deeply extracts ensemble data features to facilitate forecasting.•Results show IoT data can be replaced by open data in greenhouse control.

论文关键词：Microclimate forecast,Big data,Deep learning,Convolutional neural network (CNN),Long short term memory neural network (LSTM)

论文评审过程：Received 13 May 2022, Revised 19 July 2022, Accepted 6 August 2022, Available online 10 August 2022, Version of Record 24 August 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.118481