Heat supply pipe network water loss diagnosis method based on federal transfer learning

The invention belongs to the technical field of intelligent heat supply, and discloses a heat supply pipe network water loss diagnosis method based on federal transfer learning, which comprises the following steps: acquiring local data of each participant, and generating a correlation expression graph sample set according to the local data; training in a deep residual shrinkage network by using the correlation expression graph sample set of each participant to obtain a local model of each participant; each participant uploads the parameters of the local model to the central server, and the central server aggregates the parameters and then outputs a shared model; and each participant uses a local correlation expression graph sample set to carry out transfer learning based on a local model and a shared model by utilizing a minimum entropy loss function to obtain a personalized model of each participant. According to the method, while the data island phenomenon of each heat supply party is solved, the possible c