Aim to solve the problem that the internal load of a wet ball mill is difficult to be accurately predicted by expert experience in the grinding process, a soft measurement method of grinding load based on ameliorated symbiotic organisms search (ASOS)-extreme learning machine (ELM) is proposed. Firstly, the soft measurement model of mill load is established by the ELM algorithm. Symbiotic organisms search (ASOS) algorithm is used to optimize the hidden layer parameters of soft sensor model. The characteristic vector of mill load is constructed with the characteristic information of cylinder vibration and vibration sound signal. The mill load parameter is the output. The grinding load test experiment and comparative analysis show that the load parameter of the mill load soft measurement model has higher prediction accuracy and higher generalization ability, which provides useful guidance for the improvement of mill grinding efficiency and control optimization.