Abstract: By combining the use of support vector regression and predictive control techniques, we find that it is possible to control tank air temperature and humidity in earthworm treatment. In this study, we use Support Vector Regression (SVR) to predict temperature and humidity in the earthworm biological treatment reactor used to treat the municipal sludge, where the temperature and humidity belong to the nonlinear dynamic model system. Base on this model, we design a nonlinear model predictive controller. Furthermore, we propose an optimization algorithm to generate online control signals under the control constraints. To build and generalize an earth worm treatment model, the Teaching Learning Based Optimization (TLBO) technique is adopted to adjust the hyper-parameters of SVR. Experimental results show that the tuned SVR model by TLBO has good regression precision and is generalizable.