摘要： 为了提高下料装置在海上摇摆、颠簸工况下的饲料称重精度，提出了基于PSO-BP神经网络的称重误差修正算法。基于传感技术，建立配置有称重传感器和姿态传感器的密闭式试验装置，在近海养殖场测取不同配重下020°倾斜范围内的称重与姿态数据，在确定修正系数后，引入BP神经网络算法获取称重预测值。结果显示：真实质量分别为8.06 kg与12.40 kg的测试样本中，相较于直接测量法，BP神经网络算法修正后的称重数据的最大相对误差分别减小4.32%与4.36%；相较于BP神经网络算法，PSO-BP神经网络算法修正后的称重数据，其最大相对误差分别降低0.39%与0.33%。研究表明，对于海上网箱养殖业的饲料称重，运用PSO-BP神经网络算法进行误差修正具有更高的精度。

关键词: 网箱养殖, 称重, BP神经网络, 粒子群, 误差修正[]?

Abstract: In order to improve the bait weighing accuracy of the discharging device under the swaying and bumping conditions at sea, a weighing error correction algorithm based on PSO-BP neural network is proposed. Based on the sensing technology, a closed experimental device with load cell and attitude sensor was established, and the weighing and attitude data in the tilt range of 0-20 degrees under different counterweights were measured in the offshore farm, and after determining the correction coefficients, the BP neural network algorithm was introduced to obtain the predicted values of the weighing. In the test samples with real masses of 8.06 kg and 12.4 kg, the maximum relative errors of the weighing data corrected by the BP neural network algorithm were reduced by 4.32% and 4.36%, respectively, compared with the direct measurement method; the maximum relative errors of the weighing data corrected by the PSO-BP neural network algorithm were reduced by 0.39% and 0.33%, respectively, compared with the BP neural network algorithm. The maximum relative errors of the PSO-BP neural network algorithm were reduced by 0.39% and 0.33%, respectively. The PSO-BP neural network algorithm has a higher accuracy in error correction for bait weighing in offshore net-pen aquaculture.

Key words: net-pen aquaculture, weighing, BP neural network, particle swarm, error correction