The efficient tracking of power system inertia involves two key processes

namely

first

to identify the best-fit model at a low computational cost;second

to improve the fit probability between the estimated and the actual inertia in time. Accordingly

the effectiveness of the inertia evaluation and its online tracking are determined by whether the above two processes are achieved. In this paper

as for process 1

the recursive algorithm is applied in order to substantially reduce the temporal and space cost of the model identification; as for process 2

Bayesian-net with maximum likelihood estimation is used to improve the fit probability between the extracted and the actual value of the inertia. Besides

the multi-thread parallel computing method is developed to lower the calculation cost substantially. Simulation results in the IEEE 39-Bus System show that

by using the proposed method

the relative errors of the identified inertia are less than 5%

the time consumption is reduced by 97%

and the communication memory is 16 bytes/per update

which are much lower than those of the previous methods.