Prediction for Dissolved Oxygen in Water of Fish Farm by Using General Regression Neural Network

摘要

This paper uses general regression neural network (GRNN) to predict dissolved oxygen in water of fish farm for aquaculture. In Taiwan, aquaculture is one of main economic activities. In here, an important issue is how to keep dissolved oxygen in water of fish farm in normal range. Most aquaculture operators use air pump to keep dissolved oxygen in water in normal range, hut turn on or off the air pump by their experience so that the cost can’t control effectively. Therefore, this paper collects the related data by sensors and then use GRNN to predict dissolved oxygen in water. In GRNN, there are two main process, including learning and recalling. By the data, GRNN can find the relationship between input data and output data. From the experimental results, the R-value and MAPE in training are 0.99099 and 1.8744%, respectively; in test, the R-value and MAPE are 0.95036 and 5.2137%, respectively. It can show that GRNN can huild an effective model for dissolved oxygen in water of fish farm.