The setting of the size compensation value during the grinding process plays a vital role in the machining accuracy of batch workpieces. Through the in-depth analysis of the cylindrical grinding process, the compensation value prediction algorithm in the grinding process was studied, and the PSO-LSSVM compensation value accurate prediction model based on fuzzy matter element analysis was proposed. Through fuzzy matter-element analysis, the optimal process parameters were quickly and accurately obtained, and the corresponding parameters were input to train the PSO-LSSVM prediction model. The parameters of the LSSVM model were optimized by PSO to improve the prediction accuracy of the model. When the predicted value was greater than the theoretical boundary, the size error was compensated, and the processing accuracy was improved by adjusting the process parameters in time. The average absolute error of the model is 0.052 57 μm and the root mean square error is 0.065 33 μm verified by online measurement and grinding processing experiments.Using the obtained model to predict the compensation value during the grinding of the workpiece outside the specimen, the average absolute error of the model is 0.096 25 μm and the root mean square error is 0.134 12 μm, which meets the accuracy requirements of the compensation value prediction. Through the processing test of batch workpieces, it is concluded that the processing accuracy of batch workpieces is significantly improved compared with that before the compensation value prediction compensation control is not use. The proposed method of compensation value prediction was applied to the grinding processing active measurement control instrument. The control instrument can realize automatic compensation adjustment and form feedback control with the machine tool, which improves the accuracy of grinding workpieces and the intelligence of the grinding processing system.