The invention provides a collaborative
model prediction control method for a regional cooling
system penetrating through a source network load full chain, and belongs to the technical field of
intelligent control of cooling systems. Collecting and preprocessing meteorological and operation characteristic data of the
system; inputting the data into an Attention-LSTM cold load prediction model, determining an optimal step length in combination with
thermal inertia analysis, and outputting a cold load prediction sequence; based on the unified energy path theory, a
pipe network quasi-steady-state hydraulic and
thermal coupling mechanism model is constructed; by taking load prediction as input and the
coupling model as constraint, establishing an optimization model containing three targets of pump consumption power and the like; rolling optimization is carried out by adopting a multi-target
particle swarm optimization algorithm, and optimal operation parameters meeting thermal unbalance degree constraints are obtained; and in combination with the peak-valley
electricity price, a cold source unit and
cold storage device collaborative scheduling strategy is formulated, and a final control instruction is generated. The phenomena of insufficient cold supply at the
tail end and local
supercooling caused by hydraulic imbalance of a
pipe network and large
hysteresis characteristics are effectively solved, and
optimal control over the operation cost is achieved.