| 王慧敏,宋松柏,张更喜.基于季节长记忆双自回归模型的日径流模拟[J].水利学报,2023,54(6):686-695 |
| 基于季节长记忆双自回归模型的日径流模拟 |
| Daily runoff simulation based on seasonal long-term memory DAR model |
| 投稿时间：2022-12-04 |
| DOI：10.13243/j.cnki.slxb.20220982 |
| 中文关键词: 长记忆性|时变波动性|径流模拟|双自回归模型|渭河流域 |
| 英文关键词: long-term memory|time-varying volatility|runoff simulation|double autoregressive model|Weihe River basin |
| 基金项目:国家自然科学基金项目(52079110);江苏省自然科学青年基金项目(BK20220590) |
|
| 摘要点击次数: 1397 |
| 全文下载次数: 1719 |
| 中文摘要: |
| 传统时间序列模型无法同时考虑径流序列的长记忆性和时变波动性，且模型参数限制严格，从而使日径流序列模拟受到限制，影响径流模拟预测精度。本文提出了同时考虑非平稳性、季节性、长记忆性和时变波动性的新型双自回归模型(WOA-SFIDAR)，并与经典长记忆波动率模型(SFIAR-GARCH)进行对比，选取渭河流域4个水文站日径流序列进行模拟验证。结果表明：WOA-SFIDAR模型的模拟能力优于SFIAR-GARCH模型，模拟结果很好地保持了日径流过程的统计特性。7、8月份模拟均值误差相对较大，WOA-SFIDAR模型的误差范围(5.72～32.56)低于SFIAR-GARCH模型(7.42～48.02)。WOA-SFIDAR模拟逐月变差系数(Cv)和偏态系数(Cs)与实测序列统计值间偏差范围为0～0.51和0.02～1.31，优于SFIAR-GARCH模型(0.02～0.56和0.06～1.52);模拟结果能够保持日径流序列自相关系数(ACF)的变化趋势，且随滞时的增加，实测序列与模拟序列的ACF差距减小。文中模型扩展了水文随机模拟方法，可为日径流模拟和预报提供一种新途径。 |
| 英文摘要: |
| Traditional time series models have strict parameter ranges and cannot simultaneously consider the long-term memory and time-varying volatility of runoff series，which limits the applicability of daily runoff series and affects the simulation accuracy.This study proposes a new double autoregressive model (WOA-SFIDAR) that can simultaneously describe non-stationarity，seasonality，long-memory and time-varying volatility，and compares it with the traditional long-memory volatility model (SFIAR-GARCH).Four hydrological stations in the Weihe River basin are selected to establish models for simulation verification.The results show that the simulation ability of the WOA-SFIDAR model is better than that of the traditional SFIAR-GARCH model，and the simulation results well preserve the statistical characteristics of the daily runoff process;the simulation error of average runoff in July and August are relatively large，and the error range of the WOA-SFIDAR model (5.72-32.56) is lower than that of the SFIAR-GARCH model (7.42-48.02);the deviation ranges between the monthly variation coefficient (Cv) and skewness coefficient (Cs) simulated by WOA-SFIDAR and the statistical values of the monthly measured series are 0-0.51 and 0.02-1.31，which are lower than those of the SFIAR-GARCH model (0.02-0.56 and 0.06-1.52);the simulated daily runoff well describes the variation trend of the autocorrelation coefficient (ACF) of the measured series，and the gap in ACF shrinks with increasing lags.The model proposed in this study expands the hydrological stochastic simulation method，and provides a new approach for runoff simulation and forecasting. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |