The increased proportion of grid-connected renewable energy sources with uncertainty makes the system state variation diverse

increases the difficulty of accurate control of system and prompts system energy co-dispatch. In this paper

a kind of parameter knowledge sharing algorithm is proposed firstly to improve computational speed and accuracy by sharing deep neural network parameters. Then

a two-layer linked generation control framework is constructed

and the parameter knowledge sharing algorithm is applied to improve the computational speed and accuracy of deep neural network parameter settings by migrating parameters between similar solution objectives

and the proposed interval fast contraction algorithm is applied to allocate commands to each type of energy storage. Finally

the experimental results of a power system containing a high percentage of renewable energy sources show that the proposed algorithm

compared to the comparison algorithm

reduces the average frequency deviation and cost by 22.5% and 3.77%

respectively.