基于SSA-RBF的灌区流量预测

水资源短缺问题日益严重，快速准确的灌区流量测量具有重要意义。现有流量测算模型一般采用传统的测流方法或简单的神经网络模型进行处理，采用上述方法将面临测量成本、测量精度等挑战。故将麻雀搜索算法(sparrow search algorithm, SSA)与径向基神经网络(radial basis function, RBF)相结合，以渠道水深、测点流速、测点位置为输入，灌区流量为输出，设计了一种新的SSA-RBF神经网络模型用于预测灌区流量。以都江堰人民渠渠首站点在27种不同水力条件下的实测数据为基础，对SSA-RBF模型和RBF模型以及极限学习机(extreme learning machines, ELM)模型进行评估和比较，实例结果表明SSA-RBF模型能够快速准确地预测出流量，其确定系数为0.975、均方根误差为6.186、平均绝对误差为4.324、残差质量系数为0.011 9,4种评价指标以及预测结果偏差均优于ELM模型以及RBF模型，为提升灌区流量测算精度提供了思路。

Irrigation District Flow Prediction Based on SSA-RBF

The shortage of water resources is becoming more and more serious, so it is of great significance to measure the flow of irrigation areas quickly and accurately. The existing flow measurement models are generally processed by traditional flow measurement methods or simple neural network models, which will face challenges such as measurement cost and measurement accuracy. Therefore, this paper combines SSA (sparrow search algorithm) with RBF (radial basis function neural network), and designs a new SSA-RBF neural network model to predict the flow in the irrigation area, taking the channel depth, the velocity of measuring points, and the location of measuring points as the inputs and the flow in the irrigation area as the output. The SSA-RBF model, RBF model and ELM (extreme learning machine) model are evaluated and compared based on the measured data at the head station of Renmin Canal in Dujiangyan under 27 different hydraulic conditions. The example results show that the SSA-RBF model can quickly and accurately predict the flow, with the determination coefficient of 0.975, root mean square error of 6.186, average absolute error of 4.324, and residual quality coefficient of 0.011 9, The four evaluation indexes and the deviation of prediction results are better than ELM model and RBF model, providing ideas for improving the precision of flow measurement in irrigation areas.