Hydrological regime of Lake Dongting is the overlap of discharge from Yangtze River and four other inlets around the lake (i. e. Xiang, Zi Yuan and Li Rivers). Hie operation of the Three Gorges Reservoir (TGR) inevitably disturbs this hydrological regime. In this paper, the water level processes of Chenglingji Station (outlet of Lake Dongting) were simulated by using BP neural networks, to evaluate the independent proportion/weight of water level change caused by the operation of the TGR. In the training phase, the average daily outflow discharge of TGR and the synthetic discharge of other four inlets were chosen as inputs, and the water level of Chenglingji Station as output. In the implementation phase, the outflow discharge of TGR was replaced with the inflow to simulate the water level process without regulation of TGR. The results indicate that significant impacts have been introduced due to the runoff regulation of the reservoir. The water level dropped 2.03m in 2006 and 2.11m in 2009 during the reservoir storage period, and the extreme drop was up to 3.30m and 3.12m, respectively. In view of its potential impacts on the structure and function of lake wetlands, proper attention should be paid to this occurrence.