关键词: 智能灌溉平台, 云边端协同, CNN-LSTM, 容器化编排, 灌溉决策

Abstract:

The informatisation of agricultural irrigation management represents a crucial direction for future sustainable agricultural development. However, existing control platforms suffer from limitations in decision-making model capabilities and insufficient control timeliness. To address this, this study constructs an intelligent irrigation control platform based on a ‘cloud-edge-end’ collaborative architecture. Utilising Kubernetes containerised orchestration technology, it enables flexible deployment and elastic scheduling of models and algorithms. For diverse cultivation scenarios, we developed a water balance decision model for field irrigation and a cumulative photosynthetically active radiation-based irrigation decision model for greenhouses, both utilising deep learning prediction. Concurrently, we designed a closed-loop control module encompassing ‘water demand monitoring - irrigation decision-making - intelligent control - post-irrigation evaluation,’ thereby creating a dynamic management platform covering the entire irrigation lifecycle. Results from regional farmland simulations and practical applications in protected agriculture demonstrate: in field maize simulation validation, the platform's decision-generated irrigation volume reduced by approximately 18.3% compared to experience-based irrigation, while yield increased by 27.3% and irrigation water use efficiency improved by about 56%; In controlled environment lettuce cultivation trials, water savings reached 10.02%, yields increased by 9.38%, and irrigation water use efficiency improved by 21.33%. This irrigation platform provides technical and methodological support for multi-scenario irrigation management.

Key words: intelligent irrigation platform, cloud-edge-device collaboration, CNN-LSTM, containerised orchestration, irrigation decision-making