Capacity assessment error elimination and evaluation methods

摘要

By analyzing two groups of capacity assessment data, the paper obtained such data characteristics as data discreteness and frequency domain, and established an error analysis model and an error evaluation model for the above data on the basis of their characteristics. When reducing random errors, the paper compared merits and demerits of mean filters and median filters, and improved mean filters according to data discreteness. By referring to the evaluation results of the error evaluation model, it chose median filters as the method in the paper to eliminate high-frequency random errors. In this way, not only random errors are reduced, but the rules for systematic errors are maintained. Then, neural networks were used to fit the systematic error function, aiming to reduce systematic errors and to achieve data consistency as well. Finally, the fitted systematic error function was used for actual fitting of the second measurement data that had passed the median filters, thus random errors and systematic errors were reduced.