Medium-and-long-term runoff predicting method based on improved Elman neural network

Abstract

The invention discloses a medium-and-long-term runoff predicting method based on an improved Elman neural network, comprising the following steps: S1, preprocessing data; S2, selecting a predictive factor and extracting a principal component; S3: constructing an Elman neural network model; S4, subjecting the network model to 10-fold cross validation; S5, if forecast accuracy meets a requirement, saving a network and forecast result, otherwise, going to the step S2; and S6: if the number of forecast times reaches a requirement, calculating the average number of all forecast results, otherwise,going to the step S4. The method has fast learning convergence speed and can avoid falling into local optimum values.

Description

technical field [0001] The invention relates to the field of information technology, in particular to a medium and long-term runoff forecasting method based on an improved Elman neural network. Background technique [0002] Accurate forecasting of runoff is an important basis for guiding the comprehensive development and utilization of water resources, scientific management and optimal dispatching. At present, medium and long-term hydrological forecasting methods can be roughly divided into two categories: data-driven models and process-driven models. The data-driven model refers to the establishment of the optimal mathematical relationship between the forecast object (such as annual average runoff) and the predictor (such as the previous atmospheric circulation factor) directly based on historical data without considering the physical mechanism of the hydrological process, and with the help of this mathematical The relationship predicts future hydrological variables. Comm...

AI Technical Summary

Problems solved by technology

At present, most forecasting models based on neural networks mostly use BP and improved BP neural network, which is a static forward network, and there are certain defects in directly applying it to runoff prediction related to time series, and the network There are problems such as slow learning convergence speed and easy to fall into local optimum

Images