Yalong River Hydropower Deploys AI Model for Water-Solar-Wind Operations

The Kela photovoltaic station, part of the first phase of the water-solar complementary project at the Lianghekou hydropower station on the Yalong River, has been connected to a water-wind-solar integrated smart operations model. The model was developed by Yalong River Basin Hydropower Development Co. , Ltd. together with NARI Group, built on Huawei's Pangu large model, to address the impact of meteorological changes such as cloud cover on photovoltaic output.

The photovoltaic station site is located on the western Sichuan plateau. Under the traditional model, weather information had to be transmitted and coordinated across multiple departments, limiting processing efficiency. Since the large model came online, it has been able to predict cloud trajectories, thickness, and shading duration, and automatically link dispatch and trading operations, completing hydropower supplementary generation and trading quotation strategy adjustments within milliseconds.

The model is the first to integrate weather, runoff, and wind-solar power forecasting into a single system, extending the mid-to-long term runoff forecast period from 45 days to 60 days. In late June 2026, when runoff on the Yalong River increased while spot electricity prices remained low, the operations team adjusted trading strategies and generation plans based on advance forecasts of weather, runoff, and price trends, storing surplus inflow for additional generation in July when spot prices rose. Yalong River Basin Hydropower Development Co. said the future integrated base has a planned installed capacity of 78 million kW.

The model has also incorporated monitoring data from more than 10,000 devices across the entire basin, converting decades of hydropower operations and maintenance experience into a knowledge graph. A company official said the company will continue to advance energy artificial intelligence technology research and development, relying on its rock-cavern intelligent computing center and domestic large model infrastructure.