Gridding method and black box method were used to acquire and analyze real-time monitoring of moisture content in heat-drying sludge. The properties and calculation methods of hot and humid air in heat-drying system were discussed to study the relationship of moisture discharge rate, drying time and sludge moisture content in the return air inlet. The latest honey badger algorithm (HBA) was tried and introduced to optimize the support vector machine (SVM). HBA-SVM regression model was established, and then compared with the regression model of SVM optimized by particle swarm optimization (PSO) and genetic algorithm (GA). Results showed that the return air inlet moisture discharge rate and sludge moisture content were non-linearly reduced with drying time, and the reduction rate of moisture discharge was slightly higher than that of moisture content. The coefficient of determination (R2) and root mean square error (RMSE) of HBA-SVM were 0.996 5 and 0.979 2, respectively, and lower dispersion and higher accuracy were achieved. By transplanting the model into the embedded system and verified by site testing, the comprehensive prediction accuracy reached more than 90%. It is concluded that the low dispersion of the prediction value and high prediction accuracy are obtained by applying the HBA-SVM regression model, which is an effective method that can be used to monitor the actual sludge moisture content.