Abstract: To effectively monitor the state of algae growth in seawater, a method of soft sensing for the key representative factor based on support vector machine （SVM) algorithm was investigated. First, gridding optimization was adopted to optimize the penalty factor C and parameter σ of SVM. Then, the optimal matching parameters were used to obtain the soft sensing model of concentration of seawater chlorophyll-a via the samples training. The result of soft sensing based on SVM was compared with the result of using BP neural network. The testing result shows that the soft sensing method based on SVM has more prediction accuracy and stability than the method of BP neural network. The SVM model can be used in soft sensing for the state of seawater algae growth.