CN-121982831-A - Landslide body deformation abnormality identification and early warning method based on seasonal decomposition residual errors

CN121982831ACN 121982831 ACN121982831 ACN 121982831ACN-121982831-A

Abstract

A landslide body deformation abnormality identification and early warning method based on seasonal decomposition residual errors belongs to the technical field of slope deformation monitoring. The method comprises the steps of data acquisition and preprocessing, seasonal decomposition, residual error anomaly identification, comprehensive early warning and attribution judgment. According to the method, a complete intelligent analysis and early warning technology system for landslide deformation is established through technical means such as seasonal decomposition of time sequence data, residual error component extraction, machine learning anomaly identification, multisource information fusion and judgment, and the like, so that anomaly accurate identification and grading early warning under the premise of effectively stripping seasonal interference are realized, and reliable technical guarantee is provided for early prevention and control of landslide disasters.

Inventors

ZHOU NIANHUA
ZOU XINGHONG
WANG LIJUN
QIAN HONGJIAN
WU SHUANGLI
ZHANG YULONG
WANG WENTAO
JIANG HONGJUN
YANG SHIWEI
HE LIDONG
Jing Weichen

Assignees

中国电建集团昆明勘测设计研究院有限公司
国家能源集团金沙江旭龙水电有限公司

Dates

Publication Date: 20260505
Application Date: 20260109

Claims (6)

1. A landslide body deformation abnormality identification and early warning method based on seasonal decomposition residual errors is characterized by comprising the following steps: S1, data acquisition and preprocessing, namely acquiring time sequence deformation data of monitoring points of a landslide body, and performing data cleaning and initial standardization processing to form time-ordered deformation sequence data; S2, seasonal decomposition, namely decomposing the standardized deformation sequence data by adopting a seasonal decomposition algorithm to obtain a trend item reflecting long-term variation, a seasonal item reflecting periodic fluctuation and a residual item remained after stripping the trend item and the seasonal item; s3, residual error anomaly identification, namely analyzing the characteristics of a residual error sequence, carrying out anomaly identification on the residual error sequence formed by the residual error items, and identifying an abnormal residual error point which exceeds a normal fluctuation range and taking the abnormal residual error point as a criterion of landslide body deformation anomaly; s4, comprehensively early warning and attribution judgment, namely when abnormality is detected, comprehensively integrating the long-term trend of the trend item and the periodic rule of the seasonal item, evaluating the risk level of the landslide body and issuing early warning information, and simultaneously, carrying out auxiliary judgment on the cause of the abnormality by combining contemporaneous geological and meteorological data.
2. The landslide body deformation abnormality identification and early warning method based on seasonal decomposition residual error as set forth in claim 1, wherein the data cleaning and initial standardization processing adopts a Z-Score standardization method to conduct linear interpolation filling on the missing values, and the calculation formula is as follows: ; Wherein, the As the raw data is to be processed, As the mean value of the sequence, The standard deviation of the sequences is given as the standard deviation of the sequences, Is normalized data.
3. The landslide body deformation abnormality identification and early warning method based on seasonal decomposition residual error as set forth in claim 1, wherein the seasonal decomposition algorithm decomposes a standardized deformation sequence into trend, seasonal and residual error terms, and the calculation formula is as follows: ; Wherein, the The deformation sequence data after the normalization at the time t, For the trend term value at the time t, For the seasonal term value at time t, And the residual term value at the time t.
4. The landslide body deformation abnormality recognition and early warning method based on seasonal decomposition residual error as recited in claim 1, wherein the residual error abnormality recognition adopts isolated forests for automatic recognition, and the method comprises the steps of: s3-1, taking residual sequences of all monitoring points as input features; s3-2, carrying out recursive random segmentation on the sample by constructing 100-200 isolated trees; S3-3, calculating the average path length of each residual data point in all the isolated trees by recursively and randomly dividing a sample space; S3-4, calculating the anomaly score of each residual data point, judging and outputting residual difference constant points according to a preset anomaly score threshold; the anomaly score calculation formula is as follows: ; Wherein, the For the anomaly score to be the one in question, For the sample The number of samples of the data set, For the path length of a single residual sample in an orphan tree, Is based on The calculated average path length is used to determine, Is the path length mean of each sample point of the sample dataset.
5. The landslide body deformation abnormality identification and early warning method based on seasonal decomposition residual error as set forth in claim 1, wherein the comprehensive early warning and attribution judgment comprises the following specific steps: S4-1 risk level multidimensional evaluation, namely dividing the landslide body risk level into three levels of attention, warning and alarm based on the abnormal score size and the spatial distribution condition of the residual difference constant points and whether acceleration phenomenon occurs in the point trend item at the same period; s4-2 abnormal conditions are attributed to two types of abnormal conditions, namely: when the residual error abnormal point is identified, automatically acquiring synchronous rainfall data, judging whether the rainfall and continuous rainfall are performed, monitoring earthquake activities with the magnitude greater than 3 Richner level, and comprehensively judging the construction conditions near the landslide body, and then attributing the abnormality to deformation triggered by external environment disturbance; a spontaneous destabilization judgment step of, when the obvious external disturbance is not present, attributing the abnormality to spontaneous destabilization in which the internal structure of the landslide body is deteriorated if the change rate of the trend term exceeds 2 standard deviations of the historical average value thereof; And S4-3, grading early warning release and feedback, namely releasing corresponding early warning information according to the determined risk level and attribution result, automatically raising the risk level of the abnormal judged as spontaneous destabilization by one level, and generating a comprehensive research report containing abnormal point positions, risk levels, cause inference and treatment suggestions.
6. The system formed by the landslide body deformation abnormality identification and early warning method based on seasonal decomposition residual error as set forth in claim 1, wherein the system comprises the following modules: the data acquisition and preprocessing module is used for acquiring, cleaning and standardizing time sequence deformation data; The seasonal decomposition module is used for decomposing the standardized time sequence data by a seasonal decomposition algorithm and outputting a trend item, a seasonal item and a residual item; the residual error anomaly identification module is used for automatically identifying abnormal points of the residual error sequence by adopting an isolated forest algorithm, and identifying and marking residual error constant points; And the comprehensive early warning and studying and judging module is used for evaluating the risk level, judging the cause of the abnormality and issuing early warning information by combining residual error characteristics, trend item states, rainfall, earthquake and construction data when the abnormality is detected.

Description

Landslide body deformation abnormality identification and early warning method based on seasonal decomposition residual errors Technical Field The invention relates to the technical field of slope deformation monitoring, in particular to a landslide body deformation abnormality identification and early warning method based on seasonal decomposition residual errors. Background The landslide body deformation monitoring is a core work in the fields of slope engineering and geological disaster prevention and control, and the accuracy and timeliness of the landslide body deformation monitoring are directly related to the security of people life and property in a landslide influence area. Currently, modern monitoring technology can realize high-frequency and accurate collection of landslide body displacement and deformation rate, and mass data is accumulated for stability evaluation. However, in the key link from data to decision-deformation abnormality identification and early warning, the prior art system still has a serious short plate. The current landslide body deformation early warning mainly depends on a static threshold method or a simple statistical model based on the whole deformation rate, and has three technical bottlenecks that trending and seasonal components in deformation data cannot be effectively stripped, normal fluctuation such as seasonal expansion and contraction is misjudged as an instability signal to cause frequent misinformation, creep trend and sudden acceleration deformation which are slowly accumulated are insensitive, abnormal recognition is lagged, an optimal early warning window is often missed, an early warning trigger mechanism is single, the capability of carrying out quick auxiliary judgment on the cause of an abnormal event is lacked, and the accuracy of emergency decision is affected. In the prior art, although individual researches try to introduce a complex time sequence model, the general calculation is complex, the practicability is poor, and a complete technical system from 'data seasonal decomposition' to 'residual error anomaly identification' to 'intelligent early warning and attribution' cannot be established, so that the urgent requirement of early and accurate early warning on landslide disasters cannot be met. Disclosure of Invention The invention provides a landslide body deformation abnormality identification and early warning method based on seasonal decomposition residual errors, which realizes abnormality accurate identification and grading early warning on the premise of effectively stripping seasonal interference and provides reliable technical guarantee for early prevention and control of landslide disasters. A landslide body deformation abnormality identification and early warning method based on seasonal decomposition residual errors comprises the following steps: S1, data acquisition and preprocessing, namely acquiring time sequence deformation data of monitoring points of a landslide body, and performing data cleaning and initial standardization processing to form time-ordered deformation sequence data; S2, seasonal decomposition, namely decomposing the standardized deformation sequence data by adopting a seasonal decomposition algorithm to obtain a trend item reflecting long-term variation, a seasonal item reflecting periodic fluctuation and a residual item remained after stripping the trend item and the seasonal item; s3, residual error anomaly identification, namely analyzing the characteristics of a residual error sequence, carrying out anomaly identification on the residual error sequence formed by the residual error items, and identifying an abnormal residual error point which exceeds a normal fluctuation range and taking the abnormal residual error point as a criterion of landslide body deformation anomaly; s4, comprehensively early warning and attribution judgment, namely when abnormality is detected, comprehensively integrating the long-term trend of the trend item and the periodic rule of the seasonal item, evaluating the risk level of the landslide body and issuing early warning information, and simultaneously, carrying out auxiliary judgment on the cause of the abnormality by combining contemporaneous geological and meteorological data. According to the method, a set of complete intelligent analysis and early warning technology system for landslide deformation is established through technical means such as seasonal decomposition of time sequence data, residual error component extraction, machine learning anomaly identification, multisource information fusion and research judgment, and the like, so that anomaly accurate identification and grading early warning under the premise of effectively stripping seasonal interference are realized, and reliable technical guarantee is provided for early prevention and control of landslide disasters. Drawings FIG. 1 is a flow chart of identifying and early warning landslide body abnormality. Fig. 2 is a