Hybrid LSTM-Markovian model for greenhouse power consumption prediction: a dynamical approach

Abstract

In this paper, we consider the LSTM-Markov chain model, combining deep learning with statistical methods, to forecast greenhouse power consumption. By analyzing real-time data spanning two and a half years, the model captures temporal and sequential dependencies in seasonal energy usage patterns. Comparative analysis against CNN-LSTM, LSTM, and CNN models across different seasons highlights its superior accuracy and predictive capability. Particularly during seasonal transitions, the LSTM-Markov model demonstrates exceptional precision. Its effectiveness in optimizing resource allocation and enhancing energy efficiency in greenhouse operations offers valuable insights for stakeholders, enabling informed decision-making and sustainable agricultural practices.