Credibility analysis method and device for deformation early warning of side wall of underground powerhouse of pumped storage power station
Through the information vector machine early warning method and wireless sensor system, the problem of insufficient accuracy and credibility of the prediction model in the deformation warning of side wall of the underground factory of the power pumping and storage station is solved, and the high-precision and fast warning effect is achieved, which enhances the underground Evaluation of the stability of the cave chamber.
Patent Information
- Application Number
- CN202411184896.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-27
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-08-27
AI Technical Summary
In the existing technology, in the deformation warning of the side wall of the underground plant of the pumping and storage power station, the prediction model has problems such as difficult to determine parameters, inaccurate prediction results, unrepresentative credibility analysis indicators, and inconvenient and fast equipment.
Using an early warning method based on information vector machine, a nonlinear mapping relationship is established through Gaussian process algorithm and Bayesian statistical learning theory, a high-precision three-way displacement sensor and wireless bridge are used for data transmission, and the average prediction variance is updated in real time to improve early warning Trustworthiness.
It significantly improves the accuracy and credibility of deformation warning of the edge wall of the underground factory of the pumping and storage power station, reduces the computational complexity and memory usage, provides a fast and convenient early warning device, and enhances the evaluation and support of the stability of the underground caverns. Reference for care treatment measures.
Smart Images

Figure CN119066520A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of engineering geological disaster early warning engineering, and relates to a credibility analysis method and device for early warning of deformation of side walls of underground powerhouses in pumped-storage power stations. Background Art
[0002] The underground powerhouse of a pumped-storage power station is the "heart" of the pumped-storage power station. The main and auxiliary powerhouses are like huge underground palaces. They are buildings for installing reversible turbines, generator sets and various auxiliary equipment. They are not only a complex of hydraulic structures, mechanical and electrical equipment, but also a place for operators to carry out production activities. The underground powerhouse cavern engineering of a pumped-storage power station has the characteristics of large span, high side walls, many cavern intersections, and complex underground cavern group structure. The excavation of the underground powerhouse cavern of a pumped-storage power station is very likely to cause deformation of the surrounding rock, and it is at risk of collapse at any time during construction and operation. Therefore, strengthening the control of the deformation of the side walls of the underground powerhouse of a pumped-storage power station is important for the safety of the underground powerhouse of a pumped-storage power station.
[0003] Displacement is an important observation index in the deformation process of the side wall of the underground powerhouse of a pumped storage power station. The prediction of side wall deformation displacement is of great significance in engineering. By studying the laws contained in the monitored data of the side wall deformation displacement of the underground powerhouse of a pumped storage power station, the displacement time series analysis method is used to predict the future deformation displacement of the side wall of the underground powerhouse of a pumped storage power station. This is one of the important ways to warn the deformation of the side wall of the underground powerhouse of a pumped storage power station. At present, the methods for using the displacement time series analysis method to warn the deformation of the side wall of the underground powerhouse of a pumped storage power station include multivariate regression, ARMA, grey system theory, artificial neural network (ANN), support vector machine (SVM), etc. However, there are still some deficiencies in the reliability analysis of the prediction results of the deformation of the side wall of the underground powerhouse of a pumped storage power station in the existing studies. The main reasons are (1) the defects of the established prediction model, including the difficulty in determining the parameters and the inaccurate prediction results. (2) The indicators selected for the reliability analysis are not representative. (3) The prediction of the reliability needs to be determined based on the measured value, and the measured value is a lagged "future" value, so the reliability analysis is uncertain. (4) The device used is not convenient and fast enough. Summary of the invention
[0004] The information vector machine-based credibility analysis method and device for early warning of deformation of underground powerhouse side walls adopted by the present invention can effectively solve the above-mentioned problems.
[0005] Information Vector Machine (IVM) is an approximate algorithm based on Gaussian process algorithm. Due to the application of Bayesian statistical learning theory and kernel method, IVM has many advantages such as adaptive acquisition of hyperparameters, strong adaptability to high-dimensional and complex nonlinear problems, and probabilistic meaning of prediction output. At the same time, IVM maintains the process variance implied by the use of kernel function, and these variances can be tracked, and by using the selection of active subsets, a sparse representation is provided for the model, thereby significantly reducing the time and space complexity of learning. Compared with Gaussian process, the information vector machine algorithm can reduce computational complexity and memory usage without losing accuracy. Through the information vector machine learning algorithm, the nonlinear mapping relationship of the deformation and displacement time series of the side wall of the underground powerhouse of the pumped storage power station can be well established, thereby realizing the prediction of the deformation of the side wall of the underground powerhouse.
[0006] The credibility analysis of the information vector machine for the deformation warning of the side wall of the underground powerhouse of the pumped storage power station is evaluated by the fit R of the predicted value. The fit R is calculated by combining the prediction results of the warning model with the average prediction variance e of the training set. The higher the fit R, the higher the credibility. The average prediction variance e will be updated in time, that is, the variance between the measured value and the predicted value is calculated every time a new measured value is obtained, and then the average prediction variance is recalculated. Using the average prediction variance of the existing data instead of the variance between the measured value and the predicted value to participate in the calculation of the fit R can improve the credibility of the warning indicator, because this indicator does not depend on the "future" measured value, and the analysis result is more certain.
[0007] The information vector machine credibility analysis device for the deformation warning of the side wall of the underground powerhouse of the pumped storage power station is the carrier of the above-mentioned warning model. The device uses a high-precision three-way displacement sensor to ensure the accuracy of the displacement data. At the same time, the use of a wireless bridge for data transmission is not only convenient and fast in the layout of equipment, but also has a long transmission distance and high speed.
[0008] The present invention provides a new method and device for improving the accuracy of early warning of deformation of the side walls of underground powerhouses of pumped storage power stations, solves the practical engineering problem of deformation prediction of the side walls of underground powerhouses to a certain extent, and provides a reference for evaluating the overall stability of underground caverns and the selection of support treatment measures.
[0009] The purpose of the present invention is to provide an information vector machine credibility analysis method and device for early warning of deformation of side walls of underground powerhouses of pumped storage power stations, aiming at deformation and damage of rock masses between caverns caused by excavation of underground cavern groups in pumped storage power stations.
[0010] In order to achieve the above object, the present invention adopts the following technical scheme:
[0011] In the first part, the present invention provides a credibility analysis method for early warning of deformation of side walls of underground powerhouses in pumped storage power stations, comprising the following steps:
[0012] Step 1: Construct an early warning model for the deformation of the underground powerhouse wall of a pumped storage power station based on an information vector machine, referred to as the early warning model; sort out the historical measured data of the deformation of the underground powerhouse wall of the pumped storage power station and divide it into two sets: a training set and a test set. The training set is used to train the early warning model. The test set is used to test the trained early warning model, and the parameters of the early warning model are modified and adjusted according to the test results.
[0013] Step 2: Establish an early warning model based on an information vector machine. The present invention adopts an information vector machine machine learning method, which is similar to a Gaussian process machine learning method. The construction of the Gaussian process model mainly includes the following four steps:
[0014] (1) Define the edge approximation function
[0015] Based on the Bayesian formula, the joint distribution of the latent variable set f and the output observation value y is expressed as:
[0016]
[0017] In the formula, p(y n |f n ) is the noise model, which shows the connection between the latent variable f and the output observation y. Integrating formula (11) can get the marginal likelihood function:
[0018]
[0019] Where B is a diagonal matrix, and its nth diagonal element is β n ,β n The value of
[0020] (2) Calculate the posterior distribution
[0021] Based on the Bayesian formula and multivariate Gaussian distribution, the posterior distribution of f can be obtained by combining equations (4) and (5):
[0022]
[0023] An important assumption in the Gaussian process model is that the noise model must obey a Gaussian distribution. The information vector machine constructs an approximate function through ADF to replace the non-Gaussian posterior distribution to maintain the applicability of the model. In the ADF algorithm, the approximation process of the true posterior distribution is to replace the training data n in J with i Added to I, the posterior distribution (f) can be updated as:
[0024]
[0025] Using moment matching to minimize KL divergence, the new approximation is q i (f) = N(f|μ i ,∑ i ), where μ i Yes i The mean vector of (f), ∑ i Yes i The covariance matrix of (f) is updated as follows:
[0026]
[0027] in,
[0028] In summary, the ADF algorithm can be used to approximate any noise model using the Gaussian noise model.
[0029] (3) Kernel parameter θ learning
[0030] The kernel parameter θ is estimated by maximizing the marginal likelihood function of formula (5). The objective function including θ is as follows:
[0031] θ ML =argmax p(y|X,θ) (9)
[0032] The training process of Gaussian process is to select kernel parameters directly through training samples. Classification prediction samples can be predicted by determining kernel parameters.
[0033] (4) Data prediction
[0034] Calculate the optimal kernel parameter θ ML After that, according to p(f|X,θ)=N(f|0,K) and the properties of multivariate Gaussian distribution, the joint distribution of [ff(x)] can be obtained. The integration of the joint distribution and the posterior distribution of f is obtained as follows:
[0035]
[0036] in, is the posterior covariance function;
[0037] μ(x)=k T K -1 ∑By represents the posterior mean function.
[0038] Once the distribution of f(x) is determined, the function value of f(x) at x can be predicted, which is the prediction process.
[0039] For the binary classification problem of Gaussian process, the classification result is a probability value. Therefore, the cumulative Gaussian function in the S-type function can be used for transformation to obtain the probability noise model of Gaussian process classification:
[0040]
[0041] Normalization constant Z i :
[0042]
[0043] Therefore, the approximation to q i (f) The parameter g in the mean vector and covariance matrix update formula in With γ in Can be expressed as:
[0044]
[0045] Among them, g in Indicates g i The nth element of in It represents Γ i The nth diagonal element of The nth diagonal element υ in for:
[0046]
[0047] Then the approximate Gaussian distribution of the noise model is Medium n With β n The calculation formula is as follows:
[0048]
[0049] In summary, the approximate formula of the marginal likelihood function for the classification problem is:
[0050] p(y)≈N(m|0,K+B -1 )(18)
[0051] When the number of training samples is N, the computational complexity of solving the gradient of the marginal likelihood function to θ during the θ learning process is O(N 3 ), and takes O(N 2 ) memory. When the value of N is large, the computational efficiency will be greatly affected, which also limits the wide applicability of the model. The information vector machine machine learning algorithm approximates the Gaussian process machine learning process by introducing active subset selection and parameter optimization:
[0052] (1) Activity subset selection:
[0053] The posterior differential entropy of each data point in J is calculated in the information vector machine, and the data point with the maximum posterior differential entropy is selected and added to I. When the i-th information vector is selected for I, the posterior differential entropy of the n-th data point in J is:
[0054]
[0055] For Σ in formula (10), i-1 A sparse table is used to reduce memory usage, so the continuous vector product of the original variance matrix Σ0=K can be used to obtain Σ i :
[0056]
[0057] Where M i is an i×n-order matrix, with the kth row being n k represents the kth information vector included in I. Since ζ i-1,n is the covariance matrix Σ i-1 The nth diagonal element in . Then the posterior covariance matrix ζ i The update formula for , expressed in diagonal form:
[0058]
[0059] The update formula of the posterior mean output vector is:
[0060]
[0061] In summary, the activity subset can be determined according to the information vector machine activity subset selection algorithm:
[0062] 1. Initialization. Set the number of information vectors in the active subset to d; m = 0; set ζ0 = diag(K); μ = 0; the inactive subset is all training sample data J; the active subset I is an empty set, and s0 is an empty matrix.
[0063] 2. When i = 1, traverse all training samples n∈J and calculate g according to equations (13), (14), and (15): in , γ in and in ; Calculate ΔH according to formula (19) in .
[0064] 3. Calculate the data point n corresponding to the maximum a posteriori differential entropy i = argmax n∈J ΔH in .
[0065] 4. Update m according to equations (16) and (17) n and βn , and use equations (21) and (22) to calculate ζ i and μ i .
[0066] 5. Extensions To M i-1 , and then update M i .
[0067] 6. The nth i The data is added to I and removed from J.
[0068] 7. Repeat 1 to 6 until d data are selected and entered into I.
[0069] Through the above steps, the activity subset I used for training can be obtained.
[0070] (2) Parameter Optimization
[0071] The information vector machine algorithm uses the training samples in the active subset to replace all training samples for Gaussian approximation, and the marginal likelihood function is:
[0072]
[0073] The kernel function parameter θ is included in K I In the above example, the optimal θ can be obtained by maximizing the marginal likelihood function of equation (23) using the scaled conjugate gradient method.
[0074] Step 3: Train the early warning model based on information vector machine. The training set in the historical measured data is divided into multiple groups of displacement time series, which are input as input vectors into the preliminarily established early warning model. The k-fold cross validation (K-CV) method is used to randomly divide the samples into 10 parts (k = 10), and one of them is selected as the verification sample and the other 9 as the training sample. This is repeated 10 times. Finally, the algorithm is estimated with the average accuracy of the 10 verification results. A reliable classification model based on information vector machine with optimal denoising parameters can be obtained when the accuracy reaches 95%. The relevant parameters of the early warning model are continuously modified according to the fit between the predicted value and the measured value, and the optimal displacement time series length L and the average prediction variance e are determined.
[0075] The historical measured data constitutes the training model learning sample (x i ,y i ), where i = 1, 2, ..., n, the input sample vector, where x i is the first i items of historical measured data, y i is the corresponding measured displacement data at the next moment.
[0076] Step 4: Test the early warning model based on the information vector machine. The test set in the historical measured data is composed of multiple groups of input vectors according to the optimal time series length L, and input into the trained early warning model. According to the fit between the predicted results and the measured results, the feasibility of the early warning model is judged. If the prediction result is not ideal, return to step 3 to retrain the early warning model.
[0077] Step 5: Combine the existing measured data of the deformation of the side walls of the underground powerhouse of the pumped storage power station to form a set of displacement time series samples N with a length of L as the input vector.
[0078] Step 6: Input the input vector into the trained early warning model to obtain the predicted displacement data corresponding to time A.
[0079] Step 7: Perform goodness of fit analysis on the predicted displacement data at time A and the average predicted variance of the training set to obtain the fit R and determine the credibility of the warning.
[0080] Step 8: Update the average prediction variance e. Use the high-precision three-dimensional displacement sensor deployed on site to collect the deformation displacement x of the side wall of the underground powerhouse of the pumped storage power station in real time, calculate the variance e1 between the measured value and the predicted value, and update the average prediction variance e * .
[0081] Furthermore, the degree of fit R is evolved from the goodness of fit concept in statistics, which is used to analyze the results of regression analysis in statistics. In step 3 of the present invention, the degree of fit is calculated by combining the predicted value and the average prediction variance e of the sample set. Similarly, the degree of fit R is a value between 0 and 1. The closer the value of R is to 1, the better the degree of fit, that is, the higher the credibility of the early warning model prediction.
[0082]
[0083] Where: y i represents the predicted value of deformation displacement, and e represents the average prediction variance of the sample set.
[0084] The calculation formula of the variance e1 between the measured value and the predicted value is as follows:
[0085]
[0086] Where: Represents the measured value of deformation displacement.
[0087] Update the average prediction variance e * The method is to set the number of sample displacements to n. Then the updated sample average prediction variance is:
[0088]
[0089] Similarly, as the measured deformation displacement value is updated, the sample average variance is updated in real time according to the variance between the measured value and the predicted value.
[0090] Table 1 Credibility judgment standard table
[0091]
[0092]
[0093] In the table, a, b, and c are threshold standards determined based on the actual project conditions.
[0094] The steps for constructing the displacement time series in step 3 are as follows:
[0095] (1): Input the monitored deformation displacement data of the side wall of the underground powerhouse of the pumped storage power station and construct the displacement time series sample data N according to the optimal length L.
[0096] (2): Let the historical point be l, and take the first 1 to l measured displacement data as the input vector x.
[0097] (3): Input the input vector into the early warning model to obtain the corresponding l+1th predicted displacement data. Then replace the oldest value of the original input vector with the l+1th measured displacement data to form a new input sequence with the same time length, and obtain the l+2th predicted displacement data. Then replace the oldest value of the original input vector with the l+2th measured displacement data to form a new input sequence with the same time length; and so on, rollingly construct multiple slope displacement time series samples.
[0098] Furthermore, the steps for constructing the early warning model in step 1 are as follows:
[0099] S1: Select samples. The historical measured displacement data of the side wall deformation of the underground powerhouse of the pumped storage power station are divided into a training set and a test set. The training set is used to train the early warning model based on the information vector machine. The test set is used to test the trained early warning model and modify and adjust the parameters of the early warning model according to the test results.
[0100] S2: Construct samples. Construct a displacement time series based on the measured data of the deformation displacement of the side wall of the underground powerhouse. Assume that the number of historical points is l, take the first 1 to l measured values as the input vector, and take the l+1th predicted value as the output value, thus forming the first sample; then use the l+1th measured value as the new measured value to replace the oldest value of the original input vector, forming a new input sequence with the same time length, and take the l+2th predicted value as the output value, thus forming the second sample; and so on, construct a series of displacement time series samples.
[0101] S3: Establish a sidewall deformation warning model. Train the warning model, set the initial parameters of the warning model, and use the classic k-fold cross-validation method to complete an information vector machine sidewall deformation warning model when the accuracy rate reaches 95% or above. In this way, a nonlinear mapping relationship between the displacement value at any time and the displacement value at the previous l times is established, and the optimal displacement time series length L is determined.
[0102] S4: Use the test set to verify the prediction performance of the early warning model. If the fit between the predicted value and the measured value is high, it is considered that the performance of the established side wall deformation early warning model meets the requirements and is feasible for predicting the side wall deformation of the underground powerhouse of the pumped storage power station; otherwise, adjust the sample and early warning model training parameters and retrain the early warning model.
[0103] In the second part, the present invention also provides a credibility analysis device for early warning of deformation of side walls of underground powerhouses of pumped storage power stations, including a field device and a terminal device. The field device includes a collection unit and a transmission unit, and the terminal device includes a transmission unit, a processing unit, and an evaluation unit.
[0104] The acquisition unit, including a high-precision three-way displacement sensor, is arranged at the displacement monitoring position of the side wall of the underground powerhouse of the pumped-storage power station and connected to the wireless device to collect the displacement change data of the side wall of the underground powerhouse of the pumped-storage power station in three directions in real time.
[0105] The transmission unit includes a wireless data transmission module. The wireless data transmission module adopts wireless bridge technology. It is a storage and forwarding device that realizes LAN interconnection at the link layer. It uses air as a medium to transmit signals, which can effectively solve the problem of difficult wired deployment. The wireless device is connected to the displacement sensor and processing unit of the acquisition unit respectively to realize data transmission between the field device and the terminal device.
[0106] The processing unit includes a microprocessor and a read-write memory, which is used to use the displacement input vector to predict the deformation displacement of the side wall of the underground powerhouse of the pumped storage power station at a certain moment using the trained early warning model, and to perform goodness of fit analysis on the predicted data combined with the average prediction variance e. The microprocessor is a central processing unit composed of a large-scale integrated circuit, which is responsible for extracting script instructions from the read-write memory, executing script instructions, and controlling other components. The read-write memory is used to store various program scripts required in the calculation process, various data generated, classification models based on information vector machines, prediction results, etc.
[0107] The evaluation unit includes a credibility calculation algorithm, which is programmed into the read-write memory and is used to calculate the goodness of fit of the predicted value, obtain the fit R, and evaluate the credibility of the warning according to the size of the fit R.
[0108] Compared with the prior art, the advantages of the present invention are:
[0109] (1) The credibility analysis method of the deformation warning of the underground powerhouse side wall of the pumped storage power station uses the existing measured values of the side wall deformation displacement to input the predicted value obtained by the warning model, and combines the average prediction variance e of the training sample set to calculate the fitting degree R, and then conducts credibility analysis to weaken the correlation between the credibility and the "future" measured value, thereby reducing uncertainty and significantly improving the accuracy of the warning.
[0110] (2) The sample data of the deformation of the side wall of the underground powerhouse of the pumped-storage power station is generally limited, which belongs to the small sample problem. The existing machine learning model is prone to the problem of "over-learning (or under-learning)" for the small sample regression problem, and there is a problem that the parameters are difficult to determine. The information vector machine based on the principle of structural risk minimization has good adaptability to the small sample regression problem.
[0111] (3) The credibility analysis device for early warning of deformation of the side wall of the underground powerhouse of the pumped-storage power station proposed by the present invention transmits the high-precision three-dimensional displacement sensor signal for monitoring the deformation of the side wall of the underground powerhouse of the pumped-storage power station to the processing unit through the transmission unit using a wireless bridge. Then, the predicted displacement of the side wall deformation at a certain moment is output through the trained early warning model based on the information vector machine, and the credibility analysis is performed based on the fit analysis result. The device can input and output signals in real time, and is quick and convenient. At the same time, the high-precision three-dimensional displacement sensor can be widely installed in different sections of the underground powerhouse side wall monitoring, and has strong applicability.
[0112] (4) The information vector machine credibility analysis device for early warning of deformation of the side wall of the underground powerhouse of the pumped storage power station provided by the present invention adopts wireless bridge technology. Compared with traditional wired technology, wireless bridge technology has the advantages of convenient deployment, long transmission distance and fast transmission speed. As long as it is in the wireless network coverage area, the on-site equipment can easily access the network. At the same time, the terminal equipment can also achieve zero configuration access, which is very convenient and fast. BRIEF DESCRIPTION OF THE DRAWINGS
[0113] Figure 1 It is a schematic diagram of the underground plant and displacement monitoring equipment layout of the embodiment;
[0114] Figure 2 is a schematic diagram of the device structure of an embodiment;
[0115] Figure 3 is a flow chart of the analysis method of an embodiment;
[0116] Figure 4 A schematic diagram of a process of constructing an early warning model based on an information vector machine according to an embodiment;
[0117] Figure 5 A schematic diagram of the displacement time series construction of an embodiment;
[0118] Figure 6It is a comparison chart of the measured value and the predicted value of the embodiment. DETAILED DESCRIPTION
[0119] The specific embodiments of the present invention are further described below in conjunction with the accompanying drawings and examples. It should be noted that the accompanying drawings only show the parts related to the present invention, not all results. And the specific examples are only for explaining the present invention, not for limiting the scope of the invention.
[0120] The schematic diagram of the main and auxiliary powerhouses and busbar tunnel of a hydropower station is as follows: Figure 1 As shown in FIG. 1 , there are two monitoring points on the side walls of the main and auxiliary powerhouses. The measured displacement data of monitoring point 1 are shown in the table below. The present invention provides a credibility analysis device for early warning of deformation of the side walls of underground powerhouses in pumped storage power stations, such as Figure 2 shown.
[0121] This example selects 100 displacement sample data of a monitoring point on the side wall of an underground powerhouse of a pumped storage power station in the early stage to predict the deformation displacement of the side wall in the next five periods, so as to specifically demonstrate the information vector machine credibility analysis method for early warning of deformation of the side wall of the underground powerhouse of a pumped storage power station proposed by the present invention. The specific process is as follows: Figure 3 shown.
[0122] The process of building the early warning model is as follows: Figure 4 shown.
[0123] Step 1: The first 100 periods of historical measured data of the deformation of the side wall of the underground powerhouse of a pumped storage power station are sorted out and divided into a training sample set and a test sample set. The training sample set is used to train the early warning model based on the information vector machine. The test sample set is used to test the trained model and modify the model parameters according to the test results.
[0124] The specific method of constructing the training sample set is as follows:
[0125] The following 100 periods of historical measured data are divided into several groups in chronological order, with every 10 periods of data as a group. Because the k-fold cross-validation method (k=10) is adopted, these groups of data are randomly divided into 10 parts, and one of them is selected as a verification sample and the other 9 are selected as training samples. This is repeated 10 times. Finally, the algorithm is estimated with the average accuracy of the 10 verification results. When the accuracy is 95%, a reliable classification model with optimal denoising parameters based on information vector machine can be obtained.
[0126] As shown in Tables 2 and 3, the 10 data periods from January 20, 2023 to January 29, 2023 are divided into one group, and the first nine data periods are taken as x i ,(x i ,y i ), where x i =[x i1 ,x i2 ,xi3 ,x i4 ,x i5 ,x i6 ,x i7 ,x i8 ,x i9 ], the measured data of the tenth period is used as the output value y i . Then the data on January 20, 2023 is replaced by the data on January 30, 2023 to form a new set of samples, and so on.
[0127] Table 2 Monitoring data of the first 100 periods at monitoring point 1
[0128]
[0129]
[0130] Table 3 Monitoring data of monitoring point 1 for the last 5 periods
[0131] date Actual measured value / mm 2023 / 4 / 27 18.1 2023 / 4 / 28 18 2023 / 4 / 29 18.3 2023 / 4 / 30 18.45 2023 / 5 / 1 18.84
[0132] Step 2: Establish an early warning model based on information vector machine. In terms of IVM parameter setting, first, the more commonly used and effective Gaussian radial basis kernel function RBF (RadiusBasedKernel) is selected as the kernel function of IVM; secondly, the number of information vectors is set to 160; the parameter θ of the RBF kernel function is adaptively obtained by maximizing the marginal likelihood function during the model training process.
[0133] Step 3: Train the early warning model based on information vector machine. The training set of historical measured data is divided into multiple groups of displacement time series, which are input as input vectors into the initially established early warning model. The construction method of displacement time series is as follows: Figure 5 shown.
[0134] The k-fold cross-validation method (k=10) was used to continuously modify the relevant parameters of the early warning model according to the fit between the predicted value and the measured value, and to determine the optimal displacement time series length L and the average prediction variance e.
[0135] Step 4: Test the early warning model based on the information vector machine. The test set in the historical measured data is composed of multiple groups of input vectors according to the optimal time series length L, and input into the trained early warning model. According to the fit between the predicted results and the measured results, the feasibility of the early warning model is judged. If the prediction result is not ideal, return to step 3 to retrain the early warning model.
[0136] Based on the training sample set of the above engineering example, the optimal displacement time series length L is 15 periods, and the average prediction variance e is 0.9.
[0137] Step 5: Combined with the existing measured data on the deformation of the side walls of the underground powerhouse of the pumped storage power station, the last 15 periods are used to form a set of displacement time series samples N with a length of 15 periods as the input vector.
[0138] Step 6: Input the input vector into the trained early warning model to obtain the predicted displacement data for the next 5 periods. Figure 6 shown.
[0139] Table 4 Prediction results credibility analysis table
[0140]
[0141] Step 7: Perform goodness of fit analysis on the predicted displacement data at time A and the sample average predicted variance to obtain the goodness of fit R and determine the credibility of the early warning. The goodness of fit R of the deformation of the side wall of the five phases of the underground powerhouse obtained by the early warning model is greater than 0.95, and the credibility level is high, as shown in Table 4.
[0142] Step 8: Update the average prediction variance e. Use the high-precision three-dimensional displacement sensor deployed on site to collect the deformation displacement x of the side wall of the underground powerhouse of the pumped storage power station in real time, calculate the variance e1 between the measured value and the predicted value, and update the average prediction variance e * .
[0143] It should be noted that the purpose of publishing the above examples is to help further understand the present invention, but those skilled in the art can understand that if various obvious changes, readjustments and substitutions are made to the present invention, they will not deviate from the protection scope of the present invention. Therefore, the present invention is not limited to the contents disclosed in the examples, and the scope of protection required by the present invention shall be based on the scope defined in the claims.
Claims
1. A credibility analysis method for early warning of deformation of side walls of underground powerhouses of pumped storage power stations, characterized in that: The following steps are involved: Step 1: Construct an early warning model for the deformation of the side wall of the underground powerhouse of a pumped storage power station based on an information vector machine, referred to as the early warning model; sort out the historical measured data on the deformation of the side wall of the underground powerhouse of the pumped storage power station and divide it into two sets: a training set and a test set; The training set is used to train the early warning model; the test set is used to test the trained early warning model and modify and adjust the early warning model parameters according to the test results; Step 2: Establish an early warning model based on information vector machine and preliminarily determine the kernel function of the early warning model according to the characteristics of sample data; Step 3: Train the early warning model based on the information vector machine; divide the training set in the historical measured data into multiple groups of displacement time series, input them into the initially established early warning model as input vectors, use the k-fold cross-validation method, continuously modify the parameters of the early warning model according to the fit between the predicted value and the measured value, and determine the optimal displacement time series length L and the average prediction variance e; Step 4: Test the early warning model based on the information vector machine; the test set in the historical measured data is composed of multiple groups of input vectors according to the optimal time series length L, and input into the trained early warning model. According to the fit between the predicted results and the measured results, whether the early warning model is feasible is judged; if the predicted results are not ideal, return to step 3 to retrain the early warning model; Step 5: Combine the existing measured data of the deformation of the side wall of the underground powerhouse of the pumped storage power station to form a set of displacement time series samples N with a length of L as the input vector; Step 6: Input the input vector into the trained early warning model to obtain the predicted displacement data corresponding to time A; Step 7: Perform goodness of fit analysis on the predicted displacement data at time A and the average predicted variance of the training set to obtain the fit R and determine the credibility of the warning; Step 8: Update the average prediction variance e; use the high-precision three-dimensional displacement sensor arranged on site to collect the deformation displacement x of the side wall of the underground powerhouse of the pumped storage power station in real time, calculate the variance e1 between the measured value and the predicted value, and update the average prediction variance e * .
2. The credibility analysis method for early warning of deformation of side walls of underground powerhouses of pumped storage power stations according to claim 1 is characterized in that: In step 3, the goodness of fit is calculated by combining the predicted value and the average prediction variance e of the sample set; the goodness of fit R is a value between 0 and 1. The closer the value of R is to 1, the better the goodness of fit is, that is, the higher the credibility of the early warning model prediction is; Where: yi represents the predicted value of deformation displacement, e represents the average prediction variance of the sample set; The variance between the observed and predicted values e1 The calculation formula is as follows: Where: Represents the measured value of deformation displacement; The method for updating the average prediction variance e* is to set the number of sample displacements to n; then the updated sample average prediction variance is: Similarly, as the measured deformation displacement value is updated, the sample average variance is updated in real time according to the variance between the measured value and the predicted value.
3. The credibility analysis method for early warning of deformation of side walls of underground powerhouses of pumped storage power stations according to claim 1 is characterized in that: The displacement time series described in step 3 is constructed as follows: (1): Input the monitored deformation displacement data of the side wall of the underground powerhouse of the pumped storage power station and construct the displacement time series sample data N according to the optimal length L; (2): Let the historical point be l, and take the first 1 to l measured displacement data as the input vector x; (3): Input the input vector into the early warning model to obtain the corresponding l+1th predicted displacement data; then replace the oldest value of the original input vector with the l+1th measured displacement data to form a new input sequence with the same time length, and obtain the l+2th predicted displacement data; then replace the oldest value of the original input vector with the l+2th measured displacement data to form a new input sequence with the same time length; and so on, rollingly construct multiple slope displacement time series samples.
4. The credibility analysis method for early warning of deformation of side walls of underground powerhouses of pumped storage power stations according to claim 1 is characterized in that: In step 1, build an early warning model as follows: S1: Select samples; divide the historical measured displacement data of the side wall deformation of the underground powerhouse of the pumped storage power station into a training set and a test set; the training set is used to train the early warning model based on the information vector machine; the test set is used to test the trained early warning model, and modify and adjust the parameters of the early warning model according to the test results; S2: Construct samples; construct displacement time series based on measured data of deformation and displacement of side walls of underground powerhouses; assume that the number of historical points is l, take the first 1 to l measured values as input vectors, and take the l+1th predicted value as output value, thus forming the first sample; then use the l+1th measured value as the new measured value to replace the oldest value of the original input vector, forming a new input sequence with the same time length, and take the l+2th predicted value as output value, thus forming the second sample; and so on, construct a series of displacement time series samples; S3: Establish a sidewall deformation warning model; train the warning model, set the initial parameters of the warning model, use the classic k-fold cross-validation method, and complete an information vector machine sidewall deformation warning model when the accuracy rate reaches 95% or above, thereby establishing a nonlinear mapping relationship between the displacement value at any time and the displacement value at the previous time, and determining the optimal displacement time series length L; S4: Use the test set to verify the prediction performance of the early warning model; if the degree of fit between the predicted value and the measured value is high, it is considered that the performance of the established side wall deformation early warning model meets the requirements and is feasible for predicting the deformation of the side wall of the underground powerhouse of the pumped storage power station; otherwise, adjust the sample and early warning model training parameters, and retrain the early warning model.
5. A credibility analysis device for early warning of deformation of side walls of underground powerhouses of pumped storage power stations, characterized in that: It includes field devices and terminal equipment; the field devices include a collection unit and a transmission unit, and the terminal equipment includes a transmission unit, a processing unit, and an evaluation unit; The acquisition unit includes a high-precision three-way displacement sensor, which is arranged at the displacement monitoring position of the side wall of the underground powerhouse of the pumped-storage power station and connected to the wireless device to collect the displacement change data of the side wall of the underground powerhouse of the pumped-storage power station in three directions in real time; The transmission unit includes a wireless data transmission module; the wireless data transmission module adopts wireless bridge technology and is a storage and forwarding device that realizes LAN interconnection at the link layer. It uses air as a medium to transmit signals and solves the problem of difficult wired deployment; the wireless device is respectively connected to the displacement sensor and processing unit of the acquisition unit to realize data transmission between the field device and the terminal device; The processing unit includes a microprocessor and a read-write memory, which is used to use the displacement input vector to predict the deformation displacement of the side wall of the underground powerhouse of the pumped storage power station at a certain moment using the trained early warning model, and to perform goodness of fit analysis on the predicted data combined with the average predicted variance e; the microprocessor is a central processing unit composed of a large-scale integrated circuit, which is responsible for extracting script instructions from the read-write memory, executing script instructions, and controlling other components; Read-write memory, used to store various program scripts and data generated during the calculation process, classification models based on information vector machines, and prediction results; The evaluation unit includes a credibility calculation algorithm, which is programmed into the read-write memory and is used to calculate the goodness of fit of the predicted value, obtain the fit R, and evaluate the credibility of the warning according to the size of the fit R.
Citation Information
Patent Citations
Underground diaphragm wall deformation dynamic prediction method based on CV-LSTM combined model
CN113792372A
Concrete dam space deformation early warning index drawing method based on extreme value theory
CN116975512A
Foundation pit deformation automatic monitoring and early warning method and system based on BIM (Building Information Modeling) and electronic equipment
CN117589120A
Project disaster warning method and system based on collaborative fusion of multi-physics monitoring data
US20230410012A1
Integrated monitoring method for internal and external deformation of rock-fill dam
WO2022001104A1
Cited By
Rail transit key equipment fault prediction method and system
CN119397467A
