Under the background of "dual carbon goals"

the proportion of distributed photovoltaic in the distribution network is increasing

while the power quality problems such as the node voltage limit violation and harmonic distortion become more serious. At the same time

the development of the smart grid requires protecting sensitive data in the power system. To solve the above problems

this paper proposes an active power adjustability evaluation method for distributed photovoltaic based on self-adaptive privacy-preserving federated learning

which realizes sensitive data protection and accurate evaluation of the active power adjustability. Firstly

this paper adopts a backward/forward sweep-based harmonic analysis method for distribution systems

and the self-adaptive federated learning dataset is constructed based on Monte Carlo simulation of distribution network load uncertainty. Besides

a distributed photovoltaic active power adjustability evaluation model based on self-adaptive federated learning with three layers is constructed. Finally

the improved 189-node distribution network system and the improved 279-node distribution network system are used as example systems to validate the proposed method. Case studies demonstrate that compared with the traditional federated learning method

the accuracy of the proposed method for active power adjustability evaluation increases by 32.55% and 41.70% on the two example systems

respectively

which verifies the effectiveness of the proposed method.