A robust state estimation method based on adaptive kernel density estimation theory

To focus on the drawbacks of existing robust state estimation methods and algorithms, a novel robust state estimation algorithm based on adaptive kernel density estimation was hereby proposed. The new robust state estimation model based on adaptive kernel density estimation theory was firstly presented. In order to determine the bandwidth of adaptive kernel density, the approximative expression of optimal kernel bandwidths is deduced using mean integral square error of probability density estimation, and the characteristic of the state variable components having stronger correlation with the local direct measurements than with all un-direct measurements; and the constraints of kernel bandwidths for state variable components, including initial constraint, modified constraint and standard deviation constraint, are further designed, for insuring measurement redundancy and system observability and overcoming the issue of misidentification and unidentification of bad data; based on the above expression and constraints, the composite determination method of kernel bandwidths is finally made. Successful application of practice power system of the algorithm illustrate that: the new state estimation model, the expression of approximative optimal kernel bandwidths, constraints of kernel bandwidths and its composite determination method are all effective and reasonable. © 2015 Chinese Society for Electrical Engineering.