Multi-station runoff medium-and-long-term rolling probability prediction method considering prediction uncertainty associated evolution characteristics

Abstract

The invention discloses a multi-station runoff medium-and-long-term rolling probability prediction method considering prediction uncertainty associated evolution characteristics. The method comprisesthe following steps: collecting and arranging long-series medium-and-long-term runoff data information of a basin where a station to be predicted is located, and selecting a proper prediction model tocarry out rolling simulation-prediction and analyze prediction error statistical characteristic parameters; coupling a Copula function to improve a yoke model so as to reflect the space-time correlation characteristic of the forecast error under the condition of considering the correlation evolution characteristic of the time history and the space dimension of the multi-station runoff forecast error; and randomly sampling the rolling forecast error scene sequence by adopting a Monte Carlo method, and superposing the rolling forecast error scene sequence on a single-value forecast result to obtain a rolling probability prediction scene sequence. The yoke model theory is improved to finely describe the medium-and-long-term runoff prediction error uncertainty associated evolution characteristics, so the prediction precision and reliability can be improved, and more accurate support information is provided for a water resource management decision.

Description

technical field [0001] The invention relates to medium and long-term hydrological forecasting in the field of water conservancy engineering, in particular to a medium- and long-term rolling probability forecasting method for multi-station runoff considering the associated evolution characteristics of forecast uncertainty. Background technique [0002] Improving the accuracy of multi-site medium- and long-term (monthly, seasonal, and annual) runoff prediction is crucial to improving the accuracy of water resource assessment and improving the efficiency of water resource development and utilization. Affected by the chaos of the mid-to-long-term weather system, the mid-to-long-term prediction model based on physical causes is still difficult to guarantee the prediction accuracy in theoretical methods. At present, the most commonly used mathematical statistics methods are used to construct prediction models, such as periodic mean superposition method, neural network method, and w...

AI Technical Summary

Problems solved by technology

However, in the case of multi-station runoff forecasting, since each station has the same or similar climatic conditions, the hydrological elements of each station have a certain spatial correlation, so the forecast errors of each station also have a certain spatial correlation, while the traditional martingale The model does not take this spatial correlation into account, which leads to deviations between the statistical characteristics of the model's parameter estimates and error sequences and the actual situation, and ultimately reduces the prediction accuracy

Images