In view of the difficulty in measuring the internal wear of the coal mill in actual operation and the uncertainty of the maintenance time

it is proposed to use the current of the coal mill as the internal wear characteristic parameter

and establish the multi-model based on the least square support vector machine(LSSVM) by using the idea of multi-modeling based on the period of time

thus to obtain time series coal mill current model values under typical operating conditions. On this basis

a long and short-term memory(LSTM) neural network is applied to establish a coal mill current sequence prediction model to calculate the coal mill current. The time sequence change of the coal mill shows the trend of performance degradation of the coal mill. Moreover

taking the actual operating data of a medium-speed coal mill in a 660 MW coal-fired unit in a power plant as the research object

the method in this paper is used to predict the operation status of the coal mill

and the results are compared with the prediction results of the cyclic neural network(RNN) model. The final results show that the LSTM neural network model in this paper can obtain a clear and accurate time-series change trend of the coal mill current under the same operating conditions

which is consistent with the actual changes on the site

and the prediction accuracy is higher than that of the RNN. The overhaul current of the coal mill can accurately determine the time node for overhaul of the coal mill.