Aiming to struggle with the problem of low precision of nutrient solution dynamic deployment in protected cultivation. Based on support vector machine regression(SVR), a model for regulating nutrient solution was established. Firstly, the pH value, EC, K+ concentration, Ca2+concentration and NO-3 concentration of nutrient solution were collected under 13 temperatures and 50 groups of Knop nutrient solution ratio (A:99%Ca(NO3)2·4H2O, B:98%KNO3, C:99%KH2PO4, D:98%MgSO4·7H2O, E:99%EDTA-NaFe), and SVR was used to construct the index value prediction model. Then, the discrete slope method was used to calculate the discrete slope of the content response curve for nutrient solution detection index value and five compounds, and artificial fish swarm algorithm was used to obtain the maximum mutation point of discrete slope. Finally, the optimal regulation model of nutrient solution was constructed based on SVR with the amount of five compounds corresponding to the largest mutation feature site as the optimal control target value. The determination coefficients of the five compounds in the nutrient solution regulation model were 0.99, 0.98, 0.99, 0.96 and 0.99;the root mean square errors were 4.29mg,7.39mg,5.02mg,2.85mg and 3.96mg. These results showed that the fitting effect was good. Compared with the control effect of stepwise regression method to obtain the target value, the average relative errors of the five compounds were reduced by 37.65%, 49.94%, 40.53%, 50.58% and 42.84%. In the validation test, compared with the stepwise regression method, the relative average errors of five compounds in the nutrient solution regulation model was reduced by 46.42%, 52.08%, 54.03%, 53.59% and 54.54%. The average reduction rates of the five compounds were 1.69%, 5.81%, 5.85%, 3.65% and 7.08%, respectively. The nutrient solution regulation model based on SVR had the characteristics of high efficiency and energy saving, which may provide a reference for the practical production and application of protected crop cultivation.