A method for transmission tower strain time series prediction and instability warning based on ARIMA-LSTM combined model
Through the combination of ARIMA-LSTM combined model and finite element model, the problem of difficult to predict the strain changes and instability of the transmission line towers in the prior art is solved, and accurate prediction and real-time early warning of the tower strain are achieved in the short term to ensure the safety and stability of the power grid.
Patent Information
- Application Number
- CN202210612708.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-05-31
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2042-05-31
AI Technical Summary
It is difficult for the prior art to accurately predict and instability warning of the transmission line poles and towers through strain monitoring data, and it is impossible to promptly warn of the towers' instability, affecting the safety and stability of the power grid.
The ARIMA-LSTM combination model is used to predict the tower strain time series, combine the finite element model to simulate the load situation, establish a short-term prediction model and formulate a hierarchical warning criterion for instability to achieve real-time early warning.
Accurate prediction and real-time hierarchical warning of the strain change trend of the transmission line pole towers is achieved, helping power grid workers to take timely response measures to prevent the impact of the tower instability on power grid safety.
Smart Images

Figure CN114970273B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of disaster prevention and mitigation of power transmission lines, and in particular relates to a transmission line tower strain time series prediction and instability early warning method based on an ARIMA-LSTM combined model. Background Art
[0002] Transmission lines are the great arteries for the safe and reliable operation of power grids. The transmission lines under construction in my country are long and wide, and inevitably pass through many areas with harsh climate or geological environments. Ice, strong winds, ground subsidence and other working conditions will seriously endanger the safe and stable operation of transmission line towers. Therefore, it is of great engineering practical value and research significance to carry out online monitoring of the safety status of towers and timely warn of tower instability through monitoring data.
[0003] At present, domestic and foreign scholars and experts have proposed some online monitoring methods to evaluate the status of pole towers, including pole tower tilt monitoring, deformation monitoring, and strain monitoring. The first two detection methods can only indirectly reflect the overall load and load balance status of the tower, and it is difficult to detect local failure of the tower and provide early warning of the instability of the tower. Whether the tower is damaged or not mainly depends on the strain size of the steel structure and the bolt connection. Therefore, online monitoring of the tower strain has become an important means of safety assessment of the tower. Compared with mature strain monitoring methods, there are mainly resistance strain gauges and fiber Bragg grating strain measurement. Strain monitoring data can only help on-site operation and maintenance personnel judge the safety status of the tower at that time, and cannot play a warning role. Therefore, how to analyze the collected strain time series and accurately predict the change of the tower strain in the short term in the future, so as to provide real-time warning of the instability of the tower is a problem that needs to be solved urgently. Summary of the invention
[0004] The purpose of the present invention is to provide a transmission line tower strain time series prediction and instability warning method based on the ARIMA-LSTM combined model, so as to facilitate the grid staff to take timely response measures and prevent the instability of the transmission tower from causing adverse effects on the safe and stable movement of the grid to a certain extent.
[0005] The above technical problems of the present invention are mainly solved by the following technical solutions:
[0006] A method for predicting the strain time series of a transmission line tower and for early warning of instability based on an ARIMA-LSTM combined model, characterized in that it comprises the following steps:
[0007] Step S1. Determine the parameters used by the ARIMA model for the collected transmission tower strain time series, build the ARIMA model to predict the strain time series training samples, and obtain the ARIMA model prediction value;
[0008] Step S2. Use the LSTM model to fit the residual sequence after the ARIMA model prediction in step 1 to obtain the predicted value of the residual sequence;
[0009] Step S3. Construct another LSTM model to fit the relationship between the predicted value of the ARIMA model in step 1, the predicted value of the residual sequence obtained by fitting the LSTM model in step 2, and the collected transmission tower strain time series, and establish a short-term prediction model for the tower steel structure strain time series;
[0010] Step S4. By establishing a finite element model of the transmission tower and conductor system, changing the actual situation of the transmission line load simulation, and giving the tower safety status classification corresponding to the strain change range of the corresponding measuring point according to the simulation results;
[0011] Step S5. For the measured strain time series of the tower steel structure, the strain time series short-term prediction model established above is used to perform rolling prediction on the strain data, and combined with the tower instability classification warning criterion corresponding to the strain change range formulated in step S4, a real-time warning is given for the future safety status of the tower in the short term.
[0012] Furthermore, the main steps of determining the ARIMA model parameters in step S1 include:
[0013] S11. Use the ADF unit root test method to test the stability of the strain time series. If it is determined to be a non-stationary time series, it will be processed into a stationary time series;
[0014] S12. Determine the model type and preliminarily determine the values of the ARIMA model parameters by observing the truncation and tailing of the autocorrelation diagram and partial autocorrelation diagram of the sequence.
[0015] Furthermore, in step S4, based on establishing a finite element model of the transmission tower and conductor system, the load of the transmission line is changed to simulate the actual situation, the component stress ratio is calculated, and the strain variation range of the steel structure at the corresponding measuring point with the load is extracted according to the failure criterion of the tower structure system, and the tower instability classification warning criterion is given.
[0016] Furthermore, the tower instability classification warning criteria are divided into three levels according to the failure conditions: slight damage, moderate damage and severe damage.
[0017] Furthermore, in step S5, by rolling prediction of the tower strain time series, the safety status of the tower can be estimated in the short term, thereby providing a real-time graded warning of tower instability.
[0018] Therefore, compared with the prior art, the present invention has the following advantages:
[0019] 1. The method of the present invention proposes a transmission tower strain time series prediction model based on the ARIMA-LSTM combined model, which can predict the change trend of the transmission tower strain time series in the short term.
[0020] 2. The method of the present invention provides a theoretical basis for the classification of the safety status of the tower corresponding to the strain change range of the corresponding measuring point through finite element simulation calculation of the transmission tower-line system. Combined with the strain prediction model, it can provide real-time graded warning for the instability of the transmission tower under extreme weather conditions, thereby facilitating the power grid staff to take timely response measures.
[0021] 3. The method of the present invention is applicable to the prediction of strain monitoring data and instability warning of transmission line towers of various voltage levels. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] Figure 1 It is a specific flow chart of the method of the present invention.
[0023] Figure 2 This is a schematic diagram of the installation position of the strain sensor on the tower.
[0024] Figure 3 It is the original data of the actual measurement of the strain of a main steel structure within a certain period of time.
[0025] Figure 4 The results of 1h rolling prediction and 2h rolling prediction in the embodiment of the present invention are compared with the measured results.
[0026] Figure 5 The embodiment of the present invention establishes a finite element model of a tower-line system. DETAILED DESCRIPTION
[0027] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be further specifically described below through embodiments and in combination with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0028] Example:
[0029] This embodiment takes the measured strain data of a 500kV transmission line tower as an example to illustrate the strain time series prediction and instability real-time warning method. Figure 1As shown in the figure, σ(k), σ(k-1)…σ(km) are the original time series of the steel structure measured by the strain sensor, with a total of m+1 data points; σ'(k), σ'(k-1)…σ'(km) are the fitting results of the ARIMA prediction model for the original time series; e(k), e(k-1)…e(km) are the residual series after the ARIMA model prediction in step one; e'(k), e'(k-1)…e'(km) are the fitting results of the LSTM model for the above residual series in step two; σ(k+1)…σ(k+h) are the prediction results obtained by the above combined prediction model, with a total of h data points; σ'(k+1)…σ'(k+h) are the prediction results of the linear part of the ARIMA prediction model; e'(k+1)…e'(k+h) are the prediction results of the LSTM model for the residual part of the original time series. Specifically, the following steps are included:
[0030] Step S1. For the collected transmission tower strain time series, determine the parameters used by the ARIMA model (autoregressive integrated moving average model), build the ARIMA model to predict the strain time series training samples, and obtain the ARIMA model prediction value;
[0031] Specifically, the selected transmission tower is a wine glass tower. By installing a fiber Bragg grating-based strain monitoring device on the tower, the original strain time series data of the main steel structure of 16 towers are obtained. The actual installation position of the strain monitoring point on the tower is as follows: Figure 2 The measured original data of the strain of a main steel structure during the period from 14:14 on May 6 to 8:14 on May 10 are shown in Figure 1. Figure 3 shown.
[0032] In the establishment of the strain prediction model, the 24-hour strain data from 8:14 on May 9 to 8:14 on May 10 was used as the test sample, and the strain data from 14:14 on May 6 to 8:14 on May 9 was used as the training sample. First, a short-term prediction model was established for this section of measured data.
[0033] First, an ARIMA prediction model is established for the training samples in the measured data of the steel structure. The data is tested for stationarity, and the ADF unit root test method is used to test the sequence. The p value is 0.000191, which is less than 0.05. At the same time, the t test value is also less than the critical value of 1%. It is considered that the time series is stationary and there is no need to perform differential transformations on it. If it is determined to be a non-stationary time series, it can be processed into a stationary time series by using differential, decomposition, logarithmic and other methods. Observe the autocorrelation diagram (ACF) and partial autocorrelation diagram (PACF) of the original sequence {σ}, and determine that the best fitting model of the original sequence is ARIMA (2, 0, 0).
[0034] Step S2. Use the LSTM model to fit the residual sequence after the ARIMA model prediction in step 1 to obtain the predicted value of the residual sequence;
[0035] Specifically, the ARIMA model is a linear model and it is difficult to handle the nonlinear components of the sequence. Therefore, a long short-term memory neural network (LSTM) model is needed to fit the residual sequence {e} after the training sample prediction to obtain the predicted value {e'} of the residual sequence.
[0036] Step S3. Construct another LSTM model to fit the relationship between the predicted value of the ARIMA model in step 1, the predicted value of the residual sequence obtained by fitting the LSTM model in step 2, and the collected transmission tower strain time series, and establish a short-term prediction model for the tower steel structure strain time series;
[0037] Specifically, another LSTM model is used to fit the relationship between the sequence {σ'}, {e} and the measured value {σ}. Then, the training sample data is used as the input, and the constructed combined prediction model is used to predict it to obtain the transmission tower strain time series prediction result. The prediction process is a rolling prediction, that is, every time new strain data is obtained, it is substituted into the model to obtain the predicted data.
[0038] Step S4. By establishing a finite element model of the transmission tower and conductor system, changing the actual situation of the transmission line load simulation, and giving the tower safety status classification corresponding to the strain change range of the corresponding measuring point according to the simulation results;
[0039] Specifically, according to the strength theory in material mechanics, when the load borne by the rod reaches a certain size, the material will generally yield or break first at the most dangerous point of the stress state and enter a dangerous state. Based on the failure criteria currently formulated for the reliability analysis of transmission towers, the main basis includes: 1. The structure becomes a mechanism; 2. The displacement exceeds the allowable value; 3. The main component fails. Therefore, based on the location of the measured steel structure strain measurement point and the size of the strain change, the graded failure criteria are formulated as shown in the following table. According to the failure situation, it is divided into three levels: slight damage, moderate damage and severe damage, where ξ represents the component stress ratio, where the stress ratio is the ratio of the actual stress borne by the steel structure to its yield strength, l represents the displacement of the tower top, and h represents the tower height.
[0040] Table 1 Failure criteria of tower structure system
[0041] category Minor damage Moderate damage Severe damage Tower diagonal material ξ>1.0 ξ>1.15 —— Tower main material 0.8<ξ<1.0 1.0<ξ<1.15 ξ>1.15
[0042] According to the structural drawings of the transmission tower and the relevant parameters of the conductor, a finite element model of the tower-line system is established. Figure 5According to the failure criterion of the tower structure system formulated in the above table, by continuously changing the size of the external load on the tower to perform mechanical simulation calculations on the tower-line system, the strain changes of the angle steel at each measuring point under different working conditions can be obtained. The strain calculation results of each measuring point under different working conditions are shown in Table 2.
[0043] Table 2 Strain calculation results of angle steel under different working conditions
[0044]
[0045] It can be seen from the above table that with the increase of unbalanced tension and external load, the strain value of the steel structure unit at each measuring point also changes significantly. When the strain of the angle steel at each measuring point exceeds about 600με, the auxiliary materials of the tower begin to yield. At this time, the stress ratios of the steel structure units at measuring points 9 and 16 are 1.34 and 1.24, and the stress ratios at the other measuring points are in the range of 0.8 to 1; when the strain change of the angle steel at measuring points 15 and 16 exceeds about 1300με, and the strain change at the other measuring points exceeds about 1000με, the tower suffers moderate damage. At this time, the stress of some main materials exceeds their allowable stress and enters the yield state. The stress ratios at measuring points 13 and 14 reach 1.13 and 1.09; when the strain change of the main material measuring point exceeds 1400με and the strain change of the auxiliary material measuring point exceeds 2000με, the tower is severely damaged, will lose its bearing capacity and face a greater risk of tower collapse.
[0046] Step S5. For the measured strain time series of the tower steel structure, the strain time series short-term prediction model established above is used to perform rolling prediction on the strain data, and combined with the tower instability classification warning criterion corresponding to the strain change range formulated in step S4, a real-time warning is given for the future safety status of the tower in the short term.
[0047] Specifically, the measured strain data of each measuring point of the tower are used for short-term rolling prediction using the combined prediction model based on ARIMA and LSTM. This is compared with the established and improved instability graded warning criteria to analyze whether the predicted strain results of the steel structure at the corresponding measuring point exceed the first, second or third warning line. The overall prediction results are output, and ultimately the future safety status of the tower can be predicted on a rolling basis.
[0048] Using the above combined prediction model, the results of 1h rolling prediction and 2h rolling prediction are compared with the measured results. Figure 4 shown.
[0049] The evaluation indicators of the prediction model are shown in the following table:
[0050] Table 3 Evaluation results of 24h rolling prediction model
[0051] Evaluation indicators 1h forecast results 2h prediction results Mean absolute error (MAE) 19.7653με 35.9483με Mean Square Error RMSE 30.1934με 54.5858με
[0052] like Figure 4 As shown in the figure, the prediction results are all within the range of 100 microstrains. According to the comparison between the prediction results and Table 2, the change of the strain time series in the short term is far less than the warning range given in Table 2, so the tower is judged to be safe. In addition, from the prediction results, the combined prediction model based on ARIMA and LSTM can accurately predict the change trend of the strain time series. The prediction results are relatively ideal, and the prediction trend is completely consistent with the trend of the measured strain curve.
[0053] The specific embodiments described herein are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in similar ways, but they will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.
Claims
1. A method for predicting the strain time series of transmission line towers and instability warning based on an ARIMA-LSTM combined model, characterized in that: The following steps are included: Step S1. Determine the parameters used by the ARIMA model for the collected transmission tower strain time series, build the ARIMA model to predict the strain time series training samples, and obtain the ARIMA model prediction value; Step S2. Use the LSTM model to fit the residual sequence after the ARIMA model prediction in step 1 to obtain the predicted value of the residual sequence; Step S3. Construct another LSTM model to fit the relationship between the predicted value of the ARIMA model in step 1, the predicted value of the residual sequence obtained by fitting the LSTM model in step 2, and the collected transmission tower strain time series, and establish a short-term prediction model for the tower steel structure strain time series; Step S4. By establishing a finite element model of the transmission tower and conductor system, changing the actual situation of the transmission line load simulation, and giving the tower safety status classification corresponding to the strain change range of the corresponding measuring point according to the simulation results; Step S5. For the measured strain time series of the tower steel structure, the strain time series short-term prediction model established above is used to perform rolling prediction on the strain data, and combined with the tower instability classification warning criterion corresponding to the strain change range formulated in step S4, a real-time warning is given for the future safety status of the tower in the short term.
2. The method for predicting the strain time series of power transmission line towers and instability warning based on the ARIMA-LSTM combined model according to claim 1 is characterized in that: The main steps for determining the ARIMA model parameters in step S1 include: S11. Use the ADF unit root test method to test the stability of the strain time series. If it is determined to be a non-stationary time series, it will be processed into a stationary time series; S12. Determine the model type and preliminarily determine the values of ARIMA model parameters by observing the truncation and tailing of the autocorrelation diagram and partial autocorrelation diagram of the sequence.
3. The method for predicting the strain time series of transmission line towers and instability warning based on the ARIMA-LSTM combined model according to claim 1 is characterized in that: In step S4, based on establishing a finite element model of the transmission tower and conductor system, the transmission line load is changed to simulate the actual situation, the component stress ratio is calculated, and the strain variation range of the steel structure at the corresponding measuring point with the load is extracted according to the failure criterion of the tower structure system, and the tower instability classification warning criterion is given.
4. The method for predicting the strain time series of transmission line towers and instability warning based on the ARIMA-LSTM combined model according to claim 3 is characterized in that: The tower instability classification warning criteria are divided into three levels according to the failure conditions: slight damage, moderate damage and severe damage.
5. The method for predicting the strain time series of transmission line towers and instability warning based on the ARIMA-LSTM combined model according to claim 1 is characterized in that: In step S5, by rolling prediction of the tower strain time series, the safety status of the tower can be estimated in the short term, so as to provide real-time graded warning of tower instability.
Citation Information
Patent Citations
Method for judging icing instability of tension support of power transmission line
CN107657090A
A method for early warning of tower emergency failure under ground settlement condition
CN109388865A
Power transmission line icing thickness prediction method based on CEEMDAN-QFOA-LSTM
CN112116162A
Power transmission and transformation project slope instability risk degree evaluation method
CN113704849A
