Automatic water level prediction and optimization method based on ARIMA model driving
The automated water level prediction method driven by the ARIMA model solves the problem of response lag in water level control in foundation pit engineering, realizes accurate dynamic prediction of water level, and improves engineering safety and management level.
Patent Information
- Application Number
- CN202511641905.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-10
- Publication Date
- 2026-02-03
AI Technical Summary
Existing technologies for water level control in foundation pit engineering suffer from response lag and insufficient foresight, making it difficult to cope with emergencies and leading to a high risk of engineering accidents.
An automated water level prediction method driven by the ARIMA model is adopted. By combining data preprocessing, difference operation, autocorrelation function and partial autocorrelation function, the optimal ARIMA model is established and the water level prediction model is updated in real time to capture dynamic changes.
It enables accurate dynamic prediction of foundation pit water levels, improving the safety assurance capabilities and intelligent management level of deep foundation pit projects.
Smart Images
Figure CN121456850A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of geotechnical engineering monitoring and groundwater control, and particularly relates to an automatic water level prediction and optimization method based on an ARIMA model. BACKGROUND
[0002] Foundation pit engineering is an important link in building construction, and the accurate control of the underground water level is directly related to the stability of the foundation pit and the safety of the surrounding environment. If the underground water level is too high, it will cause the increase of the soil permeation pressure, and cause the collapse of the side wall or the uplift of the bottom of the foundation pit. If the water level is too low, it may cause the settlement of the surrounding stratum, and cause the cracking of the adjacent buildings. The traditional water level control mainly relies on manual monitoring or simple automatic equipment, and the start and stop of the water pump are triggered by setting fixed thresholds (such as the upper limit of the safety water level HS1 and the over-pumping control water level HS2). Although this method can maintain the water level within a limited range, it has defects such as response lag and lack of predictability, and it is difficult to deal with sudden heavy rain or underground pipe leakage, and it is easy to cause engineering accidents.
[0003] In recent years, automatic monitoring technology has been gradually applied to the field of foundation pit water level control. For example, the laser reflection monitoring technology developed by the Shenzhen Institute of Geological Environment uses a laser emitter and a receiver to measure the reflection time of the laser beam on the water surface to calculate the water level height, realizing non-contact measurement. The memory metal wire water level sensor designed by Shanghai Pudong Road Bridge Group uses the deformation characteristics of memory metal to improve the measurement sensitivity in complex environments. However, these technologies are still limited to the data collection level, and lack the ability to predict the trend of water level changes, resulting in a passive response state of the control strategy.
[0004] Currently, researchers use the ARIMA model (autoregressive moving average model) to analyze the steel support axial force monitoring data, and find that the original data of the axial force can be converted into a stationary sequence after first-order difference, and the autocorrelation coefficient (ACF) and the partial autocorrelation coefficient (PACF) both show a tailing feature, which meets the applicable conditions of the ARMA model. This achievement verifies the effectiveness of the ARIMA model in the processing of time series data of foundation pits, but directly migrating to water level prediction faces new challenges. In particular, the dynamic underground water is affected by rainfall infiltration, pumping interference, soil anisotropy, etc., and its fluctuation amplitude and frequency are significantly higher than the change of the axial force. Taking the Shanghai soft soil area as an example, a single heavy rainfall can cause the water level of the foundation pit to rise by 0.5 m within 2 hours, while the change amplitude of the axial force is usually less than 5%.
[0005] Therefore, to solve the problems existing in the prior art, an automatic water level prediction and optimization method based on an ARIMA model is proposed, which is a problem that needs to be solved by those skilled in the art. SUMMARY
[0006] Therefore, the application provides an ARIMA model driven automatic water level prediction and optimization method, which can capture and predict the dynamic change of the water level of a foundation pit, and significantly improves the safety guarantee capability and intelligent management level of deep foundation pit engineering.
[0007] In order to achieve the above-mentioned purpose, the application adopts the following technical solutions: An ARIMA model driven automatic water level prediction and optimization method comprises the following steps: S1, acquiring real-time water level time series data, preprocessing the data, and constructing water level sequence sample data set; S2, performing difference operation on the water level sequence sample data, determining the minimum difference order d through ADF unit root test, and obtaining a stationary sequence; S3, calculating the autocorrelation function and the partial autocorrelation function of the stationary sequence, and searching for the combination parameters of the autoregressive order p and the moving average order q in the preset range by using grid search; S4, establishing an ARIMA (p, d, q) model according to the minimum difference order d and each (p, q) combination; S5, traversing all (p, q) combinations, and selecting the optimal ARIMA (p, d, q) model as the final prediction model; S6, inputting the previous N water level data at the current time into the final prediction model, and predicting the future k-step water level prediction value; S7, updating the water level time series data in real time, training the ARIMA (p, d, q) model in real time according to the updated data, and updating the final prediction model.
[0008] The above method can optionally comprise the following steps in S1. Abnormal value detection is performed on the real-time water level time series data, and when there is an abnormal value, the abnormal value is removed based on the 3σ criterion; When there is a missing value, the linear interpolation method is used to fill in the missing value after removing the abnormal value; The filled data is standardized.
[0009] The above method can optionally comprise the following steps in S2. It is judged whether the number of water level sequence samples reaches a preset threshold N, and if not, the real-time water level time series data is continuously acquired; After the number of samples reaches the threshold N, the difference order d is initialized to 0, and the d-order difference operation is performed and then the ADF unit root test is performed; After determining that the sequence is stationary, the stationary sequence is output; If it is judged that the sequence is not stationary, then the difference order d is increased and the above difference test step is executed cyclically until the sequence is judged to be stationary and a stationary sequence is output; If the difference order d is higher than the preset maximum difference order d_max, then the data processing is terminated and it is output that the data cannot be stabilized.
[0010] The above method, optionally, in S5, the optimal ARIMA (p, d, q) model is selected as the final prediction model, specifically: The optimal (p, q) parameter combination is selected with the minimum Akaike information criterion or Bayesian information criterion as the target, and the optimal ARIMA (p, d, q) model is obtained.
[0011] The above method, optionally, in S7, the water level time series data is updated in real time, the ARIMA (p, d, q) model is trained in real time according to the updated data, and the final prediction model is updated, specifically: The water level time series data is updated in real time, a new data window is constructed every time a set value m of measured data is added, the ARIMA model is automatically retrained, and the optimal ARIMA (p, d, q) model corresponding to the current real-time data is obtained; A new prediction result is generated by the optimal ARIMA (p, d, q) model corresponding to the current real-time data, and a prediction error is calculated; When the prediction error continuously exceeds a set threshold, the optimal ARIMA (p, d, q) model corresponding to the current real-time data is updated as the final prediction model; When the prediction error does not exceed the set threshold, the existing final prediction model is maintained.
[0012] According to the above technical solution, compared with the prior art, the application provides an automatic water level prediction and optimization method based on an ARIMA model, which has the following beneficial effects: the application uses the powerful time series modeling capability of the ARIMA model to effectively capture and predict the dynamic changes of the foundation pit water level with strong nonlinear characteristics, on the basis of automatic time series water level data preprocessing to obtain a stationary sequence, the optimal ARIMA model is determined based on ACF / PACF combined with grid search, and after the measured data is updated, a new sliding window is constructed, the model is optimized and updated, the dynamic changes of the foundation pit water level can be captured and predicted, and efficient monitoring of the foundation pit groundwater level can be realized; the precise dynamic estimation of the foundation pit groundwater level can be realized, and the safety protection capability and intelligent management level of the deep foundation pit project can be significantly improved. BRIEF DESCRIPTION OF DRAWINGS
[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only constitute a part of the embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of the provided drawings.
[0014] Figure 1 A flow chart of an automatic water level prediction and optimization method based on an ARIMA model provided by the present application is shown in the figure. Figure 2 The time sequence of automatic underground water level collection in a specific embodiment of an automatic water level prediction and optimization method based on an ARIMA model provided by the present application is shown in the figure. Figure 3 The ARIMA model driven underground water level prediction sequence in a specific embodiment of an automatic water level prediction and optimization method based on an ARIMA model provided by the present application is shown in the figure. DETAILED DESCRIPTION
[0015] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments only constitute a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.
[0016] In the present application, the relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations, the term "comprising", "containing" or any other variant thereof is intended to cover non-exclusive containing, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment. Without more limitation, the element defined by the statement "including a" does not exclude the presence of another identical element in the process, method, article or equipment including the element.
[0017] Referring to Figure 1 The present application discloses an automatic water level prediction and optimization method based on an ARIMA model, which comprises: S1, real-time water level time series data is acquired, data is preprocessed, and water level sequence sample data set is constructed; S2, difference operation is performed on the water level sequence sample data, the minimum difference order d is determined through ADF unit root test, and a stationary sequence is obtained; S3, autocorrelation function (ACF) and partial autocorrelation function (PACF) of the stationary sequence are calculated, and grid search is adopted to search for combination parameters of autoregressive order p and moving average order q in a preset range; S4, an ARIMA (p, d, q) model is established according to the minimum difference order d and each (p, q) combination; S5, all (p, q) combinations are traversed, and the optimal ARIMA (p, d, q) model is selected as the final prediction model; S6, the current time N water level data are input into the final prediction model, and the future k-step water level prediction value is predicted; S7, the water level time series data are updated in real time, the ARIMA (p, d, q) model is trained in real time according to the updated data, and the final prediction model is updated.
[0018] Further, in S1, real-time water level time series data are acquired, and the data are preprocessed, specifically: The real-time water level time series data are subjected to outlier detection, and when there is an outlier, the outlier is removed based on the 3σ criterion; The data after removing the outliers are subjected to missing value detection, and when there is a missing value, the missing value is filled by using linear interpolation method; The filled data are subjected to standardization processing.
[0019] Further, in S2, difference operation is performed on the water level sequence sample data to eliminate trend and seasonality, the minimum difference order d is determined through ADF unit root test, and a stationary sequence is obtained, specifically: It is judged whether the number of water level sequence samples reaches a preset threshold N, if not, real-time water level time series data are continuously acquired; After the number of samples reaches the threshold N, the difference order d is initialized to 0, and after performing d-order difference operation, ADF unit root test is performed; After judging that the sequence is stationary, the stationary sequence is output; If it is judged that the sequence is not stationary, the difference order d is incremented and the above difference test steps are repeatedly executed until the sequence is judged to be stationary and the stationary sequence is output; If the difference order d is higher than a preset maximum difference order d_max, the data processing is terminated and it is output that the data cannot be stabilized.
[0020] Further, in S5, the optimal ARIMA (p, d, q) model is selected as the final prediction model, specifically: The optimal ARIMA (p, d, q) model is obtained by selecting the optimal (p, q) parameter combination with the aim of minimizing AIC (Akaike Information Criterion) or BIC (Bayesian Information Criterion).
[0021] Further, in S7, the water level time series data is updated in real time, the ARIMA (p, d, q) model is trained in real time according to the updated data, and the final prediction model is updated, specifically: The water level time series data is updated in real time, and a new data window is constructed every time m new measured data (such as m = 2 hours) is added, and the ARIMA model is automatically retrained to obtain the optimal ARIMA (p, d, q) model corresponding to the current real-time data; A new prediction result is generated by the optimal ARIMA (p, d, q) model corresponding to the current real-time data, and a prediction error is calculated; When the prediction error continuously exceeds the set threshold value (such as RMSE > 5 cm), the optimal ARIMA (p, d, q) model corresponding to the current real-time data is updated to the final prediction model; When the prediction error does not exceed the set threshold value, the existing final prediction model is maintained.
[0022] In one specific embodiment, the automatic water level monitoring data of the Shaoxing Metro Line 2 Chenshan South Road subway station foundation pit project is taken as an example, and the method of the present application is used for groundwater level prediction: The automatic data collection time is from 2025 / 3 / 10 to 2025 / 3 / 16, and the data is collected basically once an hour, about 24 times a day, such as Figure 2 It can be seen that compared with manual collection once a day, automatic water depth can better monitor the change of water level, especially on March 13, the medium rain day, the water level change reaches half a meter, and high-frequency monitoring and rapid prediction are urgently needed. The optimal ARIMA (1, 1, 1) model is constructed by using the present application, and the prediction is carried out on the basis of a sliding window size of 15, and the single-step water level prediction sequence is as shown in Figure 3 It can be seen that the maximum error of water level prediction is less than 10 cm, and the RMSE is 1.6 cm, which can better estimate the future change of foundation pit water level and has obvious engineering practical value.
[0023] The various embodiments described in this specification are described with reference to a particular sequence or order, but the order of the steps can be modified so that particular sequences or orders make no significant contribution to the progress of the art. Moreover, certain features and subcombinations are of utility and can be employed without reference to other features and subcombinations. For purposes of clarity, not every embodiment or feature in this specification is described or shown. Embodiments that provide real benefits can include any embodiment or combination of features described in this specification— even if the range of benefits realized is not the full range of benefits. Those of ordinary skill can understand that information and signals can be represented using any of a variety of technologies and techniques. For the purposes of this description, the terms "information" and "signals" can be regarded as synonymous. Those of ordinary skill can appreciate that the signals can be analog or digital, and the like.
[0024] The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. An ARIMA model driven automated water level prediction and optimization method, characterized in that, Comprise: S1, obtain real-time water level time series data, preprocess the data, and construct water level sequence sample data set; S2, difference operation is carried out on water level sequence sample data, the minimum difference order d is determined through ADF unit root test, and stationary sequence is obtained; S3, the autocorrelation function and the partial autocorrelation function corresponding to the stationary sequence are calculated, and the combination parameters of autoregressive order p and moving average order q are searched in the preset range by using grid search; S4, according to the minimum difference order d and each (p, q) combination, an ARIMA (p, d, q) model is established; S5, all (p, q) combinations are traversed, and the optimal ARIMA (p, d, q) model is selected as the final prediction model; S6, input the current time N water level data into the final prediction model, and predict the future k step water level prediction value; S7, real-time update water level time series data, real-time train ARIMA (p, d, q) model according to the updated data, and update the final prediction model.
2. The automatic water level prediction and optimization method based on ARIMA model driving according to claim 1, wherein in S1, real-time water level time series data is obtained, and the data is preprocessed, specifically: Abnormal value detection is carried out on the real-time water level time series data, and when there is an abnormal value, the abnormal value is removed based on the 3σ criterion; When there is a missing value, the linear interpolation method is used to fill in the missing value after removing the abnormal value; The data after filling is standardized.
3. The automatic water level prediction and optimization method based on ARIMA model driving according to claim 2, wherein in S2, difference operation is carried out on water level sequence sample data, the minimum difference order d is determined through ADF unit root test, and stationary sequence is obtained, specifically: Judge whether the number of water level sequence samples reaches the preset threshold N, if not, continue to obtain real-time water level time series data; After the sample quantity reaches the threshold N, initialize the difference order d=0, perform d-order difference operation, and then perform ADF unit root test; After judging that the sequence is stationary, the stationary sequence is output; If the sequence is not stationary, the difference order d is increased and the above difference test steps are executed in a loop until the sequence is judged to be stationary and the stationary sequence is output; If the difference order d is higher than the preset maximum difference order d_max, terminate the data processing and output that the data cannot be stationary.
4. The automatic water level prediction and optimization method based on ARIMA model driving according to claim 3, wherein in S5, the optimal ARIMA (p, d, q) model is selected as the final prediction model, specifically: The optimal (p, q) parameter combination is selected to minimize the Akaike information criterion or the Bayesian information criterion, and the optimal ARIMA (p, d, q) model is obtained.
5. The automatic water level prediction and optimization method based on ARIMA model driving according to claim 4, wherein in S7, real-time water level time series data is updated, ARIMA (p, d, q) model is trained in real time according to the updated data, and the final prediction model is updated, specifically: Real-time update of water level time series data, every new set value m measured data, build a new data window, automatically retrain ARIMA model, get the optimal ARIMA (p, d, q) model corresponding to the current real-time data; Through the optimal ARIMA (p, d, q) model corresponding to the current real-time data to generate new prediction results and calculate the prediction error; When the prediction error continuously exceeds the set threshold, update the optimal ARIMA (p, d, q) model corresponding to the current real-time data to the final prediction model; When the prediction error does not exceed the set threshold, maintain the existing final prediction model running.