During long-term operation, the parameters of an asynchronous motor can deviate greatly from the passport values, therefore various methods of parameter identification are used. The rotational speed of the asynchronous shaft of an electric motor in real identification systems is always measured with errors. Additive noise allows modeling errors in speed sensors or errors in determining speed without sensors. Estimating the derivative with noise and discretization also introduces additional errors. The presence of errors in determining the velocity leads to biased estimates when using ordinary least squares (OLS) to estimate K-parameters. The contribution presents results showing that the accuracy of estimates at speeds below the nominal is higher. The results obtained in this work can be useful in designing new algorithms for identifying the parameters of asynchronous motors. In the case of active identification, it is possible to select the best speed range for each motor model. In the case of passive identification, a specially selected weight matrix can be used depending on the speed value.