Forecasting cooling load and water demand of a semi-closed greenhouse using a hybrid modelling approach. Issue 1 (31st December 2022)

• Record Type: Journal Article
• Authors: Mahmood, Farhat; Govindan, Rajesh; Yang, David; Bermak, Amine; Al-Ansari, Tareq
• Is Part Of: International journal of ambient energy. Volume 43:Issue 1(2022)
• Page Start: 8046
• Page End: 8066
• Publication Date: 2022-12-31
• Subjects: Food security -- agriculture -- machine learning -- forecasting -- greenhouse; Power resources -- Periodicals; Renewable energy sources -- Periodicals; 621.04205
• DOI: 10.1080/01430750.2022.2088617
• Languages: English
• ISSNs: 0143-0750

Abstract

Abstract : Forecasting the greenhouse cooling and water demand is critical for improving the performance, reducing energy consumption, and operating costs throughout the year. This research proposes a hybrid modelling approach by analyzing machine learning models to determine the most suitable algorithm for a semi-closed greenhouse. The models were investigated by increasing the time step, excluding the actuator data history, and using data sets based on different seasons to determine the forecasting accuracy. LSTM outperformed both SVMR and MLP with an RMSE and R 2 value of 0.352°C, 0.982 for temperature, and 1.23%, 0.954 for relative humidity. The outputs from LSTM were used as input for the analytical model to forecast the cooling load and water demand. Results illustrated that the greenhouse had a cooling demand of 6.03 and 15.33 MWh for a two-day period in winter and summer. Similarly, the water demand was for winter and summer was 5.85 and 12.48 m 3 .