The Study of Medium- and Short-term Prediction for Artux Earthquake (M_S=6.9) and Usunan Earthquake (M_S=5.8)

In this paper, the process of medium- and short-term prediction (submitted in special cards) of the Artux earthquake (MS=6.9) and the Usurian earthquake (MS=5.8) in Xinjiang area, is introduced. The imminent seismic risk regions are judged based on long- and medium-term seismic risk regions and annual seismic risk regions determined by national seismologic analysis, combined with large seismic situation analysis. We trace and analyze the seismic situation in large areas, and judge principal risk regions or belts of seismic activity in a year, by integrating the large area’s seismicity with geodetic deformation evolutional characteristics. As much as possible using information, we study synthetically observational information for long-medium- and short-term (time domain) and large-medium -small dimensions (space domain), and approach the forecast region of forthcoming earthquakes from the large to small magnitude. A better effect has been obtained. Some questions about earthquake prediction are discussed.

“Small earthquake modulation window” and its applications in medium－term prediction

Based on the observations of many years, it has been found that “small earthquake modulation windows” exist inthe situation of some special geological structures, which respond sensitively to the variations of regional stressfields and the activities of earthquake swarms greater than moderate strong magnitude, and can supply some precursory information. More than two “small earthquake modulation windows” can also provide a general orientation of the first main earthquake of a earthquake cluster. Compared with “seismic window” based on frequency itis no doubt that the “modulation-window” has an unique characteristic of applicational significance to mediumterm earthquake prediction with a time scale of two or three years.

Seismicity Quantification and Its Application to Medium-term Earthquake Prediction

Bed on the analysis of each parameter describing seismicity,we think A(b)-value can betterquantitatively describe the feature of the enhancement and quietness of seismicity in this paper. Thedata of moderate or small earthquakes during 1972～1996 in North China are used in space scanningof A(b)-value. The result shows that 2～3 years before most strong earthquakes there wereObviously anomaly zones of A(b)-value with very good prediction effect. Some problems about themedium-term prediction by using A(b)-value are also discussed.

The middle-long term prediction of the February 3,1996 Lijiang earthquake(M_S=7) by the "criterion of activity in quiescence

Earthquake activities in history are characterized by active and quiet periods. In the quiet period, the place where earthquake M_≥6 occurred means more elastic energy store and speedy energy accumulation there. When an active period of big earthquake activity appeared in wide region, in the place where earthquake (M_≥6) occurred in the past quiet period, the big earthquake with magnitude of 7 or more often occur there. We call the above-mentioned judgement for predicting big earthquake the 'criterion of activity in quiescence'. The criterion is relatively effective for predicting location of big earthquake. In general, error of predicting epicenter is no more than 100 km. According to the criterion, we made successfully a middle-term prediction on the 1996 Lijiang earthquake in Yunnan Province, the error of predicted location is about 50 km. Besides, the 1994 Taiwan strait earthquake (M_s=7.3), the 1995 Yunnan-Myanmar boundary earthquake (M_s=7.2) and the Mani earthquake (M_s=7.9) in north Tibet are accordant with the retrospective predictions by the 'criterion of activity in quiescence'. The windows of 'activity in quiescence' identified statistically by us are 1940-1945, 1958-1961 and 1979-1986. Using the 'criterion of activity in quiescence' to predict big earthquake in the mainland of China,the earthquake defined by 'activity in quiescence' has magnitude of 6 or more; For the Himalayas seismic belt, the Pacific seismic belt and the north-west boundary seismic belt of Xinjiang, the earthquake defined by 'activity in quiescence' has magnitude of 7, which is corresponding to earthquake with magnitude of much more than 7 in future. For the regions where there are not tectonically and historically a possibility of occurring big earthquake (M_s=7), the criterion of activity in quiescence is not effective.

Gravity observations along the eastern margin of the Tibetan Plateau and an application to the Lushan MS 7.0 earthquake

This paper introduces relative and absolute gravity change observations in the eastern portion of the Tibetan Plateau. We analyze and discuss a change that occurred in 2010 in the gravity along the eastern margin of the plateau and the relationship between this change and the 2013 Lushan Ms7.0 earthquake. Our results show that： （1） before the Lushan MsT.0 earthquake, gravity anomalies along the eastern margin of the Tibetan Plateau changed drastically. The Lushan earthquake occurred at the bend of the high gradient zone of gravity var- iation along the southern edge of the Longmenshan fault zone. （2） The 2013 Lushan earthquake occurred less than 100 km away from the epicenter of the 2008 Wenchuan earthquake. Lushan and Wenchuan are located at the center of a four- quadrant section with different gravity anomalies, which may suggest that restoration after the Wenchuan earthquake may have played a role in causing the Lushan earthquake. （3） A medium-term prediction based on changes in gravity anoma- lies was made before the Lushan Ms7.0 earthquake, in par- ticular, a prediction of epicenter location.

Seismic characteristics near the epicenter of the 1303 Hongtong M=8 earthquake, Shanxi Province and its implication

In this paper, we calculated the seismic pattern of instrumental recorded small and moderate earthquakes near the epicenter of the 1303 Hongtong M=8 earthquake, Shanxi Province. According to the spatial distribution of small and moderate earthquakes, 6 seismic dense zones are delineated. Temporal distribution of ML2 earthquakes since 1970 in each seismic dense zone has been analyzed. Based on temporal distribution characteristics and historical earthquake activity, three types of seismicities are proposed. The relationship between seismic types and crustal medium is analyzed. The mechanism of three types is discussed. Finity of strong earthquake recurrence is pro-posed. Seismic hazard in mid-long term and diversity of earthquake disaster in Shanxi seismic belt are discussed.

Seismic characteristics near the epicenter of the 1303 Hongtong M=8 earthquake, Shanxi Province and its implication

In this paper, we calculated the seismic pattern of instrumental recorded small and moderate earthquakes near the epicenter of the 1303 Hongtong M=8 earthquake, Shanxi Province. According to the spatial distribution of small and moderate earthquakes, 6 seismic dense zones are delineated. Temporal distribution of ML2 earthquakes since 1970 in each seismic dense zone has been analyzed. Based on temporal distribution characteristics and historical earthquake activity, three types of seismicities are proposed. The relationship between seismic types and crustal medium is analyzed. The mechanism of three types is discussed. Finity of strong earthquake recurrence is pro-posed. Seismic hazard in mid-long term and diversity of earthquake disaster in Shanxi seismic belt are discussed.

Earthquake prediction from China's mobile gravity data

The relation between plate tectonics and earthquake evolution is analyzed systematically on the basis of 1998-2010 absolute and relative gravity data from the Crustal Movement Observation Network of China. Most earthquakes originated in the plate boundary or within the fault zone. Tectonic deformation was most intense and exhibited discontinuity within the tectonically active fault zone because of the differential movement; the stress accumulation produced an abrupt gravity change, which was further enhanced by the earthquake. The gravity data from China's Mainland since 2000 obviously reflected five major earthquakes （Ms 〉 7）, all of which were better reflected than before 2000. Regional gravity anomalies and a gravity gradient change were observed in the area around the epicenter about 2 or 3 years before the earthquake occurred, suggesting that gravity change may be a seismic precursor. Furthermore, in this study, the medium-term predictions of the Ms7.3 Yutian, Ms8.0 Wenchuan, and Ms7.0 Lushan earthquakes are analytically pre- sented and evaluated, especially to estimate location of earthquake.

Analysis on Depth Distribution and Precursor Mechanism of Small and Moderate Earthquakes

In this paper, the focus depth distribution of earthquakes with each magnitude has been analyzed. Statistic data show that the lower magnitude is, the wider focus depth distributes. With larger magnitude, the focus tends to be concentrated in upper or middle crustal layers. We analyzed the cause of focus depth distribution and explained the precursor mechanism of small and moderate earthquakes with occurring condition and characteristics of strong earthquakes. The results of this paper may be applied to determine risk sites of strong earthquakes.

New Seismic Zoning Map of Chin a and Its Application in Earthquake Disaster Mitigation

The New Seismic Zoning Map of China was prepared from 1987 to 1990 and officially promulgated in 1991.In comparison with the previous two seismic zoning maps prepared in 1957 and 1977,some new methods were applied to upgrade the method currently used for seismic hazard analysis.First,a probabilistic method was used instead of the deterministic analysis was used for previous mapping.Second,by taking advantages of the long history of historical seismic data in China,the nonhomogeneity of seismicity both in space and in time has been fully considered and hence the over-and/or underestimation of seismic hazard could be avoided.Third,the results of middle-and long-term earthquake prediction based on tectonic evidence have been incorporated into seismic hazard analysis.In addition,the attenuation laws for both intensity and peak acceleration of strong motion as the mapping parameters are also presented.Finally,an evaluation of the New Seismic Map and its effect on engineering application,such as aseismic

Study on Relation Between Dynamic Pattern of Regional Vertical Strain Rate and Several Strong Earthquakes such as Lijiang(M_s7.0)and Menyuan(M_s6.4)Earthquakes

Making use of the method of obtaining regional vertical strain rate from regional preciseleveling data and gaining dynamic pattern combining with deformation data on spanningfaults, the regional vertical strain dynamic evolution characteristics of several moderatelystrong earthquakes such as Lijiang (M_s 7.0) and Menyuan (M_s 6.4) earthquakes occurredin crustal deformation monitoring areas located in the western Yunnan and Qilianshan-Hexiregion. Based on the above-mentioned facts, by studying the time-space nonhomogeneity andstrain energy accumulation status, some criteria for judging the medium. and short-termstrong seismic risk regions according to the regional vertical strain rate dynamic informationare proposed.

基于多模型融合的中长期径流集成预测方法

中长期水文预报是流域水资源规划与合理配置的重要依据。为提高中长期径流预测精度,提出了一种基于多模型融合的水库中长期径流集成预测方法。该方法将ARMA、BP、LSTM、RF和SVR等5个异质预测模型进行融合,同时采用超参数优化方法确定各模型的最优参数。将其用于青海省龙羊峡水库的中长期径流预报中,结果表明,通过Stacking融合算法建立的集成预测模型相较于单一模型,取得了更高的预测精度(R2值由0.71提升至0.82)。此方法可为提升流域中长期径流预测精度提供一定参考。

基于LSTM神经网络的现地烈度实时估算模型——以JMA烈度为例

如何快速并且准确估计目标场点烈度是地震预警中的关键问题。常用基于衰减关系的场点烈度估计和基于P波信息的现地烈度估计往往存在大震烈度低估的问题。本文提出了一种基于长短时记忆神经网络(logn short-term memery,LSTM)的现地JMA烈度持续估计模型。该模型以现地观测地震动的能量、能量增长率、地震动卓越周期和震源距作为输入,以该点的最大仪器地震烈度为预测目标。选取了日本K-NET台网记录101次地震数据作为训练集,94次地震数据作为测试集,训练了现地烈度估算LSTM神经网络模型。结果表明:在采用3 s时窗长度的序列进行预测时,高估的比例为1.51%,低估的比例为4.00%;并且,随着时窗长度的增加,高估和低估的比例也在不断降低。模型对高烈度(大于等于4.5度)样本的预测时效性随震源距的增加而增加,对大震远场高烈度区域能提供20 s以上的预警时间。

基于概率增益模型的强震概率预测研究

为提高云南地区及邻区强震中短期预测的准确度,通过强震发生前不同地震前兆指标的对比研究,发现仅用单项指标很难提高预测的准确度。挑选出能从不同侧面反映强震发生规律的6项指标,采用概率增益模型对这些指标进行综合处理,通过估算每种地震前兆的概率增益来定量判定地震发生的可能性,最终得到可用来预测云南地区及邻区M≥6.0地震未来中短期发生可能性大小的综合概率P指标。计算分析表明:①使用综合概率P预测未来半年云南地区及邻区M≥6.0强震,能通过R值检验;②取P异常阀值为20%、30%,可对强震进行不同级别预警,P值小于20%为安全,P值为20%~30%为黄色预警,为云南地区及邻区未来存在发生M≥6.0地震的可能性;P值大于30%为红色预警,表示未来半年云南地区及邻区将发生M≥6.0地震。

基于多因子多模式集成的中长期径流预测模型

提高中长期径流预测精度对于水资源调度等具有重要意义和应用价值。基于国家气候中心的130项气候因子,采用皮尔逊相关系数、最大信息系数、方差增量指标筛选主要预测因子,建立基于DS(Dempster-Shafer)证据理论的多因子综合方法;采用随机森林、BP神经网络和贝叶斯网络等建立基于水文-气象因子遥相关的中长期径流预测模型,构建基于DS证据理论的预测结果集成模型。以三峡水库为对象开展实例研究,结果表明:引入遥相关因子能有效提高预测精度;基于DS证据理论的多因子综合方法能筛选出综合性更强、稳定性更优的因子,弥补单一筛选方法的不足;基于DS证据理论的多因子多模式集成方法在径流预测精度上优于单一方法单一模型,确定性系数提高到0.823,平均相对误差降低到23.2%。

基于STE-TCN的中短期电力负荷预测

目的 针对传统电力负荷预测模型对长序列预测精度低的问题,提出一种结合跳级卷积连接与时间编码网络的新型时序卷积神经网络(TCN)模型——STE-TCN模型。方法 首先对TCN模型加入跨周期的膨胀卷积通道(Skip-convolution)提取电力数据周期信息;再进行特征融合得到Skip-TCN网络,使网络抓取周期规律,增加信息利用长度;最后设计日期编码网络(Time encoding network)捕捉生活周期和季节性特征,与Skip-TCN进行特征融合得到STE-TCN模型,实现对电力负荷数据长序列预测。结果 实验表明:在与TCN模型和传统时序网络的对比下,Skip-TCN的预测精度均有提升,在预测长度更长的测试上提升尤为明显。结论 实验结果验证了通过对更长跨度时序关系的捕捉,STE-TCN网络改进方法有效提升了对长序列电力负荷的预测精度。

基于多模型比选耦合的降水预测

变化环境下水文时间序列的模拟预测难度不断加大,以往研究大多聚焦在模型的不同组合尝试与应用探索,但缺乏针对不同组合模型适用性与稳定性的系统研究。选择经验模态分解(empirical mode decomposition,EMD)、总体平均经验模态分解(ensemble empirical mode decomposition,EEMD)、改进的总体平均经验模态分解(modified ensemble empirical mode decomposition,MEEMD)和变分模态分解(variational mode decomposition,VMD)4种常用的分解算法,与多元线性回归(multivariable linear regression,MLR)、随机森林(random forest,RF)、BP神经网络(back propagation,BP)、卷积神经网络(convolutional neural networks,CNN)和长短期记忆神经网络(long short-term memory,LSTM)5种具有代表性的模型结合,构建20种基于“分解-预测-重构”模式的组合模型,并以华北地区密云、官厅两流域年和汛期降水为例,进行模型适用性与稳定性综合对比分析。结果表明:单一模型对密云流域年降水和汛期降水的预测结果优于官厅流域,但整体预测结果均不理想;结合分解算法后的组合模型预测结果明显优于单一模型,且该预测结果存在正负误差抵消现象,因此有助于进一步提高组合模型的整体预测精度;与基于EMD系列的分解算法相比,VMD算法对模型预测精度提升效果最显著,组合模型适用性和稳定性整体上表现为VMD-MLR>VMD-LSTM>VMD-BP>VMD-CNN。

基于气候特征分析及改进XGBoost算法的中长期光伏电站发电量预测方法

光伏发电在能源结构中的重要性不断凸显,而提高光伏发电量预测的准确性成为当前研究的关键问题。针对中长期光伏发电量预测问题,提出一个综合利用气候预测数据的中长期光伏发电量预测方法。首先,在基于气候预测数据的发电量预测框架中,根据气候预测数据特点和预测周期划分多重子模型以充分利用气候预测数据信息。其次,在进行数据预处理后,通过对气候特征衍生与交叉、特征筛选和选择,充分挖掘气候特征的高价值信息。然后,采取一种两重多阶段超参数寻优策略,对极端梯度增强(extreme gradient boosting, XGBoost)超参数进行调整以优化预测模型。最后,在真实光伏发电量数据上,以MAPE为标准评估预测水平,验证所提中长期光伏发电量预测方法的有效性。相关实验结果表明该方法可以有效提高光伏发电量预测精度。

某高速公路桥梁中长期养护规划研究

合理的养护规划对于延长桥梁结构的服役年限,降低全寿命周期养护成本具有十分重要的意义。基于此,论文通过对某高速公路桥梁历年检测数据分析及桥梁技术状况评价,建立了桥梁结构技术状况衰变模型,并根据每年检测数据进行调整,提高其预测准确性,依此制订该桥中长期养护规划方案。

基于多变量灰色遗传模型的配电网柔性中长期负荷曲线化预测方法

传统配电网柔性中长期负荷曲线化预测方法,缺乏对配电网负荷预测权重的合理有效分配,导致预测准确性不高。因此,提出基于多变量灰色遗传模型的配电网柔性中长期负荷曲线化预测方法。使用随时间反向传播(BPTT)算法训练处理后的配电网负荷数据,得到配电网负荷特征。在考虑负荷数据特征基础上,使用多变量灰色遗传模型分配负荷曲线预测权重。根据预测权重,聚类处理负荷增长等参数,输出配电网柔性中长期负荷预测曲线,从而曲线化预测配电网柔性中长期负荷。实验结果表明,设计的方法对电网柔性中长期负荷的预测误差较小、准确性较高,满足配电网柔性中长期负荷曲线化预测的需求。