The invention discloses a method for predicating COD load of sewage based on a vector time sequence model. The method comprises the following steps: performing definition of an excavation target in a data mining modeling process; performing data exploration and preprocessing on obtained sewage inlet related data to meet requirements of vector autoregression model modeling; selecting parameters of an maximum likelihood estimation model, selecting information criteria to determine order of the model, and performing model checking through multi-factor blending statics; and further simplifying the model through target parameters, thereby establishing a simple and effective predicating model. The acquired inlet water data determines water inlet amount and inlet water COD, and a sewage COD load attribute is constructed; test data predication is performed through the obtained predicating model, and output of the model is predicated results for a sewage COD load related variable. According to the method disclosed by the invention, the model is simple, multiple variables can be predicated at the same time, predicating precision is relatively high, and needed time is short.