基于深度信念极限学习机与卷积优化算法的洪水预报方法

Flood Prediction Method Based on Deep Belief Extreme Learning Machine and Convolution Optimization Algorithm

针对洪水峰高量大、汇流时间短以及流域地貌复杂,导致洪水预报难度大和预报精度不理想的问题,提出一种基于深度信念极限学习机(DBN-ELM)和改进卷积优化算法(ICOA)的ICOA-DBN-ELM模型。以渭河上游北道水文站点2006~2020年的日径流数据作为输入数据,并将该模型与BP、ELM、DBN-BP、DBN-ELM、COA-DBN-ELM模型进行对比。结果表明,所建立的ICOA-DBN-ELM模型有更好的预报精度,在洪水预报领域具有良好的应用前景。

The ICOA-DBN-ELM model based on deep belief network(DBN),extreme learning machine(ELM)and improved convolution optimization algorithm(ICOA)is proposed to solve the problems of flood prediction difficulty and unsatisfactory accuracy caused by large flood peak,short convolution time and complex basin topography.The daily runoff data of the Beidao hydrological station in the upper reaches of the Wei River from 2006 to 2020 were used as input data,and the model was compared with BP,ELM,DBN-BP,DBN-ELM and COA-DBN-ELM models.The results show that the established ICOA-DBN-ELM model has better prediction accuracy,and has a good application prospect in the field of flood prediction.