The invention discloses a multi-
source data fusion-based circulating water culture
water quality accurate regulation and control method, and belongs to the technical field of
aquaculture. The method comprises the following steps: synchronously acquiring
water quality parameters, fish school images and
water body particulate matter multi-dimensional data through an acquisition
system consisting of a
water quality sensor, an
underwater high-definition camera and a
water body particulate matter sensor; a dynamic
association model is constructed to fuse multi-
source data, a
deep learning image
recognition algorithm is introduced to quantify the fish school food grabbing intensity and
swimming speed, early warning of fish school stress is achieved, meanwhile, a
time sequence prediction
algorithm is adopted to construct a water quality prediction model, and real-time and historical data and a feeding plan are fused to predict the key water quality parameter change trend in several hours in the future; on the basis of the prediction result and the early warning
signal, operating parameters of the aerator and the pH adjusting pump are regulated and controlled in real time, and a feeding strategy is dynamically optimized. Cooperative monitoring and advanced regulation and control of water quality and fish
shoal behaviors are achieved, the culture water quality is effectively stabilized, fish
shoal stress damage is reduced, and the benefits of circulating water culture are improved.