Abstract: To address the issues of high lag, low accuracy, and significant errors in moisture detection of filter cake during the filter press dewatering process, a soft measurement method for filter cake moisture based on Regularized Stochastic Configuration (RSC) is proposed: seven key operational parameters, including feed concentration, feed flow rate, and feed pressure, are selected as input variables, with filter cake moisture as the output variable, to construct a data-driven nonlinear regression model. Using 200 sets of industrial field data as samples, the dataset is divided into training and testing sets in a 7∶3 ratio. The model's effectiveness is validated by comparing its performance with linear models (Linear) and Stochastic Configuration Networks (SCN), incorporating evaluation metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and the coefficient of determination (R²), along with error characteristic analysis. The results show that the RSC model significantly outperforms the comparative models, with MAE and RMSE as low as 0.45 and 0.56, respectively, and an R² of 97.65%. In terms of relative error, the RSC model achieves sample proportions of 65.63% for errors less than 1.0% and 96.88% for errors less than 2.0%. The model exhibits good independence and non-systematic characteristics in prediction errors, with autocorrelation coefficients for all lag orders except lag 0 falling within the confidence interval. By providing real-time predictions of filter cake moisture, this method enables flexible adjustment of the filter press's feed pressure, pressing pressure, and holding time, further improving dewatering efficiency, reducing energy consumption, and enhancing the continuity and stability of the production process.