Dynamic prediction method and system for settlement deformation of bimodal transformer substation foundation

Abstract

The invention provides a bimodal substation foundation settlement deformation dynamic prediction method and system, and relates to the technical field of foundation settlement, and the method comprises the steps: deploying a bimodal sensing monitoring network on a target substation to collect foundation mechanical deformation data and foundation deformation positioning data; building a foundation deformation data processing channel to carry out preprocessing and weighted fusion on the foundation mechanical deformation data and the foundation deformation positioning data to obtain foundation settlement deformation fusion data; and training a foundation settlement deformation prediction network to dynamically predict the foundation settlement deformation fusion data, determining a foundation settlement deformation prediction quantity, and triggering a settlement abnormity response mechanism based on the foundation settlement deformation prediction quantity to carry out graded response early warning. The technical problems that in the prior art, transformer substation foundation settlement risk identification lags behind, the disposal efficiency is low, and safe and stable operation of a transformer substation is difficult to guarantee can be solved, and the technical effect of guaranteeing long-term safe and stable operation of the transformer substation is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of foundation settlement, and particularly relates to a dual-mode substation foundation settlement deformation dynamic prediction method and system. BACKGROUND

[0002] As a core hub of the power system, the substation is prone to uneven settlement due to the influence of factors such as compression of the rock-soil body, fluctuation of the underground water level, continuous action of the equipment load, and disturbance of the surrounding engineering during long-term service. If timely monitoring and prediction are not performed, it may cause equipment tilting, cable trench cracking, and even large-area power outage accidents. However, the existing foundation settlement monitoring technology mostly adopts a single-mode sensing scheme, or can only collect foundation mechanical deformation data but lacks accurate positioning information, or can only realize position monitoring but is difficult to meet the deformation measurement accuracy requirement, and does not consider the dynamic change of the equipment reliability caused by factors such as strong electromagnetic interference of the substation, fluctuation of temperature and humidity, and sensor aging, and the early warning mechanism does not combine the importance of the settlement area, and is prone to over-treatment or insufficient treatment.

[0003] In summary, the existing technology has the technical problems of late substation foundation settlement risk identification, low treatment efficiency, and difficulty in guaranteeing the safe and stable operation of the substation. SUMMARY

[0004] The purpose of the present application is to provide a dual-mode substation foundation settlement deformation dynamic prediction method and system to solve the technical problems of late substation foundation settlement risk identification, low treatment efficiency, and difficulty in guaranteeing the safe and stable operation of the substation in the prior art.

[0005] In view of the above problems, the present application provides a dual-mode substation foundation settlement deformation dynamic prediction method and system.

[0006] In a first aspect, the present application provides a dual-mode substation foundation settlement deformation dynamic prediction method, which is realized by a dual-mode substation foundation settlement deformation dynamic prediction system, wherein the dual-mode substation foundation settlement deformation dynamic prediction method comprises: Deploy a bimodal sensing monitoring network on the target substation, which specifically includes a magnetostrictive sensing array and a WAPI anchor network, collect ground mechanical deformation data and ground deformation positioning data through the bimodal sensing monitoring network; build a ground deformation data processing channel, which is composed of a bimodal data preprocessing channel and a bimodal data fusion channel in series; the ground deformation data processing channel is used to preprocess and weightedly fuse the ground mechanical deformation data and the ground deformation positioning data to obtain ground settlement deformation fusion data; a ground settlement deformation prediction network is trained, the ground settlement deformation fusion data is dynamically predicted by introducing the ground settlement deformation prediction network, ground settlement deformation prediction is determined, and a settlement abnormal response mechanism is triggered based on the ground settlement deformation prediction to perform hierarchical response early warning.

[0007] In a second aspect, the application also provides a bimodal substation ground settlement deformation dynamic prediction system for executing the bimodal substation ground settlement deformation dynamic prediction method of the first aspect, wherein the bimodal substation ground settlement deformation dynamic prediction system comprises: A data acquisition module is arranged to deploy a bimodal sensing monitoring network on the target substation, which specifically includes a magnetostrictive sensing array and a WAPI anchor network, and collect ground mechanical deformation data and ground deformation positioning data through the bimodal sensing monitoring network; a channel building module is arranged to build a ground deformation data processing channel, which is composed of a bimodal data preprocessing channel and a bimodal data fusion channel in series; a data processing module is arranged to preprocess and weightedly fuse the ground mechanical deformation data and the ground deformation positioning data by using the ground deformation data processing channel to obtain ground settlement deformation fusion data; a dynamic prediction module is arranged to train a ground settlement deformation prediction network, introduce the ground settlement deformation prediction network to dynamically predict the ground settlement deformation fusion data, determine ground settlement deformation prediction, and trigger a settlement abnormal response mechanism based on the ground settlement deformation prediction to perform hierarchical response early warning.

[0008] The one or more technical solutions provided in the application have at least the following technical effects or advantages: By deploying a dual-mode sensing monitoring network composed of a magnetostrictive sensor array and a WAPI anchor network at the target substation, the deformation and positioning dual-dimensional data are completed; a ground deformation data processing channel with dynamic weight adjustment is built to improve data reliability; a hybrid prediction network composed of a TCN time convolution network and a GRU gate recurrent unit is designed to reduce prediction error and predict the settlement trend in advance; a hierarchical response warning mechanism based on settlement prediction and regional importance is established, which changes passive monitoring to active prediction and disposal, avoids equipment tilt and power interruption, and realizes the technical effect of ensuring the long-term safe and stable operation of the substation.

[0009] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the following specific embodiments of the present application can be implemented according to the content of the specification, and in order to make the above and other purposes, characteristics and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described. It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the present application or 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 are only exemplary, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the provided drawings.

[0011] Figure 1 Flowchart of a dual-mode substation foundation settlement deformation dynamic prediction method of the present application; Figure 2 Structure diagram of a dual-mode substation foundation settlement deformation dynamic prediction system of the present application.

[0012] Explanation of reference signs: data acquisition module 11, channel building module 12, data processing module 13, dynamic prediction module 14. DETAILED DESCRIPTION

[0013] The present application provides a dual-mode substation foundation settlement deformation dynamic prediction method and system, which solves the technical problems of late risk identification, low disposal efficiency and difficulty in ensuring the safe and stable operation of the substation in the prior art.

[0014] Below, the technical solutions in the present application will be described clearly and completely with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application. In addition, it should be noted that, for the convenience of description, only parts related to the present application are shown in the drawings, not all.

[0015] Embodiment one, please refer to the attached Figure 1 The present application provides a kind of double mode substation ground subsidence deformation dynamic prediction method, wherein the double mode substation ground subsidence deformation dynamic prediction method is applied to a kind of double mode substation ground subsidence deformation dynamic prediction system, the double mode substation ground subsidence deformation dynamic prediction method specifically includes the following steps: S100: deploy double mode sensing monitoring network on target substation, the double mode sensing monitoring network specifically includes magnetostrictive sensing array and WAPI anchor point network, ground mechanical deformation data and ground deformation positioning data are collected by the double mode sensing monitoring network.

[0016] Specifically, deploy double mode sensing monitoring network and collect data in target substation, i.e. the substation that carries out ground subsidence deformation monitoring and prediction, the collaborative monitoring system composed of two different monitoring principle sensing modules, through multidimensional data acquisition, complementary advantages are realized, avoid the limitation of single mode monitoring, double mode sensing monitoring network specifically includes magnetostrictive sensing array and WAPI anchor point network, wherein magnetostrictive sensing array refers to sensing device using the mutual conversion characteristics of mechanical deformation and electromagnetic signal in magnetostrictive material, can accurately capture the mechanical deformation such as small stretching and compression of ground, it is deployed in key stress points and subsidence sensitive areas of ground, such as under equipment foundation, uneven subsidence prone section of ground, ground mechanical deformation data are collected through its mechanical-electromagnetic conversion function, the data refers to the data reflecting the physical deformation state of ground, such as vertical displacement amount, horizontal offset amount, deformation rate of ground, etc., with the characteristics of high precision and strong real-time;WAPI anchor point network refers to anchor point monitoring network constructed based on WAPI wireless local area network security standard, position positioning is realized through signal interaction between anchors, it is deployed in the whole substation, ensure that signal coverage has no blind area, through collecting signal propagation time delay, signal strength and other parameters between anchors, ground deformation positioning data are generated, the data refers to geographic coordinate data reflecting the position of ground deformation, can clearly mark the specific direction of deformation in substation, make up for the defect that single deformation data does not know position;Two sensing devices work together to form double mode sensing monitoring network, and complete the synchronous collection of ground mechanical deformation data and ground deformation positioning data.

[0017] S200: build a foundation deformation data processing channel, the foundation deformation data processing channel is composed of a bimodal data preprocessing channel and a bimodal data fusion channel in series.

[0018] Specifically, to build a foundation deformation data processing channel, the bimodal data preprocessing channel and the bimodal data fusion channel need to be constructed respectively, and then they are connected in series to form the channel. The channel is the core intermediate link between the data acquisition of the bimodal sensing monitoring network and the subsequent foundation settlement deformation prediction network, responsible for converting the collected foundation mechanical deformation data and foundation deformation positioning data into high-quality and high-credibility fusion data, providing a data basis for accurate prediction. The bimodal data preprocessing channel is a channel that integrates the adaptability preprocessing program in parallel for the different data characteristics of the magnetostrictive sensing array and the WAPI anchor network. Its role is to preliminarily purify and standardize the original data. The bimodal data fusion channel is constructed based on real-time sensors, which dynamically weights and fuses the preprocessed data to construct the bimodal data fusion channel. After the construction of the above two channels, the bimodal data preprocessing channel and the bimodal data fusion channel are connected in series, so that the original data collected by the bimodal sensing monitoring network first enters the bimodal data preprocessing channel to complete the purification and standardization, and then the processed single-source data is input into the bimodal data fusion channel for dynamic weighted fusion, and finally the foundation settlement deformation fusion data is output.

[0019] S300: using the foundation deformation data processing channel to preprocess and weight fuse the foundation mechanical deformation data and the foundation deformation positioning data to obtain the foundation settlement deformation fusion data.

[0020] Specifically, a ground deformation data processing channel composed of a dual-mode data preprocessing channel and a dual-mode data fusion channel in series is adopted to preprocess and weightedly fuse the ground mechanical deformation data collected by the magnetostrictive sensing array and the ground deformation positioning data collected by the WAPI anchor network. The preprocessing is to eliminate noise interference, correct system error, and unify data format through a customized program, so as to convert the disordered original data into standardized data meeting the subsequent fusion standard. Specifically, for the ground mechanical deformation data, a program containing data cleaning, noise filtering, and temperature compensation is run according to the mechanical-electromagnetic conversion characteristics of magnetostrictive materials; for the ground deformation positioning data, a program containing coordinate conversion, error correction, and data alignment is run in combination with the signal parameter characteristics of the WAPI anchor network, such as signal propagation delay and strength. Finally, two types of standardized single-mode data are output. After the preprocessing, the standardized data enters the dual-mode data fusion channel to perform weighted fusion. Based on the dynamic allocation of weights according to the real-time reliability of the data, the two types of single-mode data are integrated into comprehensive data, which is a comprehensive data set containing both the accurate value of the ground deformation and the specific location where the deformation occurs, and both the deformation measurement accuracy of the magnetostrictive sensing array and the positioning coverage advantage of the WAPI anchor network are retained.

[0021] S400: training a ground settlement deformation prediction network, introducing the ground settlement deformation prediction network to dynamically predict the ground settlement deformation fusion data, determining the ground settlement deformation prediction value, and triggering a hierarchical response warning based on the ground settlement deformation prediction value.

[0022] Specifically, the ground settlement deformation prediction network is trained. First, a neural network with a specific architecture is constructed, and historical settlement data is used for training and optimization. Finally, a ground settlement deformation prediction network with the ability to predict future ground settlement deformation, strong generalization ability, and high prediction accuracy is obtained.

[0023] After the network training is completed, the network is introduced for dynamic prediction. Dynamic prediction refers to inputting the real-time generated ground settlement deformation fusion data into the trained prediction network, and outputting the ground settlement situation in a specific future period in real time by the network. The ground settlement deformation fusion data output by the ground deformation data processing channel in real time contains the current ground deformation value and the corresponding position information, which are input into the trained prediction network. The network first extracts the time sequence features in the fusion data through the TCN layer, then mines the long and short term dependencies between the data through the GRU layer, and finally outputs the settlement value, deformation rate, and corresponding occurrence position in a specific future period in real time, i.e., the ground settlement deformation prediction value.

[0024] After obtaining the foundation settlement deformation prediction, an abnormal response and a hierarchical early warning are performed. First, the prediction is compared with a preset settlement threshold. If the prediction reaches or exceeds a certain threshold, an abnormal settlement response mechanism is triggered immediately. The mechanism is a pre-set automatic response rule system, which includes threshold judgment, early warning level matching, emergency instruction generation and other functional modules, and can realize seamless connection from risk identification to response initiation. After the response mechanism is started, the foundation settlement deformation early warning level is determined through an internal algorithm, taking into account the magnitude of the prediction exceeding the threshold and the importance of the corresponding area. The early warning level is usually divided into three levels: level I, mild early warning; level II, moderate early warning; and level III, severe early warning. Finally, a hierarchical response early warning is executed based on the determined early warning level. Differentiated response measures are matched according to different early warning levels to avoid over-treatment or under-treatment. For example, when the early warning level is I, the system automatically sends a short message to the operator to remind him to increase the monitoring frequency of the corresponding area. When the early warning level is II, in addition to intensified monitoring, the on-site inspection program is immediately started, and technical personnel are arranged to carry out a preliminary safety assessment report. When the early warning level is III, the system directly triggers the emergency shutdown plan, and at the same time, the emergency repair team is linked to push the precise location and prediction data of the settlement area to guide the team to carry out settlement control and structure protection work, avoiding the expansion of the accident.

[0025] Further, the deployment of the dual-mode sensing monitoring network on the target substation includes: The structure distribution data and equipment layout data of the target substation are collected, and a three-dimensional reconstruction is performed based on the structure distribution data and equipment layout data to establish a three-dimensional model of the substation. Geological exploration analysis is performed on the target substation to generate a geological structure model of the substation. The geological structure model of the substation and the three-dimensional model of the substation are analyzed by spatial fusion to obtain an integrated spatial model of the substation. Based on the integrated spatial model of the substation, a monitoring network is deployed to construct a dual-mode sensing monitoring network.

[0026] Specifically, first, basic data collection is carried out, focusing on collecting structure distribution data of the target substation, which reflects the spatial characteristics of the above-ground building structure of the substation, including the position coordinates, size parameters, structure materials and mutual connection relationships of the main control building, equipment foundation, cable trench, fence and other structures, as well as equipment layout data, which records the spatial configuration of the core power equipment of the substation, including the installation position, weight distribution, rated load and relative distance to the surrounding structure of devices such as transformers, circuit breakers and disconnectors. Then, based on these two types of data, three-dimensional reconstruction is performed using BIM building information model or three-dimensional laser scanning technology, i.e. two-dimensional data is converted into a three-dimensional visual model through digital technology to construct a three-dimensional model of the substation that can intuitively present the above-ground structure and equipment layout. The model can clearly show the spatial position of the settlement sensitive area, such as under the heavy equipment foundation and the structure connection.

[0027] After the completion of the above-ground model construction, the underground geological data acquisition is carried out, and the geological exploration analysis of the area where the target substation is located is carried out. The professional analysis process of detecting the underground soil layer distribution, the physical and mechanical properties of rock-soil body, the underground water level depth, and the information of adverse geological bodies such as soft soil layer and karst cave distribution is carried out through drilling, geophysical prospecting, and soil test. According to the exploration results, the underground geological stratification characteristics, the compressibility and stability parameters of each soil layer are sorted out, and the substation geological structure model which can accurately reflect the underground geological conditions is generated. The model can clearly mark the weak areas of underground settlement such as high compressibility soft soil layer distribution area and underground water level change sensitive area, and supplement the underground dimension basis of monitoring planning.

[0028] Then, the spatial fusion analysis is carried out, which means that the above-ground three-dimensional model and the underground geological model are superimposed, calibrated and correlated in the same spatial coordinate system, and the accurate spatial matching of the above-ground structure, equipment and underground geological conditions is realized. Through the unified coordinate reference, the substation three-dimensional model and the substation geological structure model are spatially aligned, so that the position of a certain equipment foundation on the ground can be directly corresponded to a specific soil layer underground, forming a substation integrated spatial model containing the global information of above-ground structure-equipment-underground geology. The model can intuitively identify the high-risk areas of the superposition of important equipment on the ground and weak geology underground, and provide global decision basis for monitoring network deployment.

[0029] Finally, the monitoring network deployment is carried out based on the substation integrated spatial model. According to the risk level marked by the model, the sensor installation position, quantity and density are determined. In the high-risk area marked by the model, such as the transformer foundation area above the underground soft soil layer, the sensor is densely deployed; in the medium-risk area, such as the auxiliary building area with stable underground soil layer, the sensor is reasonably arranged; in the low-risk area, such as the open space, the deployment density is reduced; and the magnetostrictive sensor array is installed synchronously to collect the foundation mechanical deformation data, and the WAPI anchor point network is installed to collect the foundation deformation positioning data. Finally, a dual-mode sensor monitoring network covering the key risk areas of the substation and considering the above-ground and underground influencing factors is constructed, which can ensure the comprehensive and accurate collection of double-dimensional data related to foundation settlement.

[0030] Further, the construction of the dual-mode sensor monitoring network comprises: identifying a set of foundation settlement influencing factors, establishing a risk quantification evaluation index system according to the set of foundation settlement influencing factors, performing settlement risk area evaluation on the substation integrated spatial model according to the risk quantification evaluation index system, obtaining a set of substation graded risk areas, determining the deployment position of the substation monitoring sensor and the regional deployment density of the substation monitoring sensor according to the set of substation graded risk areas, and performing monitoring network deployment on the target substation based on the deployment position of the substation monitoring sensor and the regional deployment density of the substation monitoring sensor to construct the dual-mode sensor monitoring network.

[0031] Specifically, first, the foundation settlement influencing factors are identified, the target substation foundation design drawings are consulted, combined with regional geological exploration reports and similar substation settlement failure cases, the foundation settlement influencing factor set is systematically sorted out and identified, the set refers to the sum of various key factors that may induce or exacerbate the target substation foundation settlement, covering the monthly change rate of underground water level, the compression modulus of rock-soil mass, the load distribution of substation structure, such as the weight of heavy equipment, the disturbance intensity of surrounding engineering construction, seasonal freezing and thawing cycle, etc., to ensure that no core influencing factor is missed. Based on the factor set, the analytic hierarchy process is adopted combined with the scoring method of power industry experts to establish a risk quantitative evaluation index system, which converts qualitative influencing factors into a calculable standardized evaluation framework, including index layers such as underground water level monthly change > 50 cm, 30-50 cm, < 30 cm, weight layers, the weights of each factor are determined by AHP, such as the compression modulus of rock-soil mass weight 0.3, the structure load weight 0.25, the scoring layer, the quantitative standards of 5 points, 3 points and 1 point are set for the corresponding indexes, to realize the measurable and comparable of the settlement risk.

[0032] Subsequently, the index system is associated with the integrated space model of the substation, and the comprehensive risk score of each 10m x 10m space unit divided in the model is calculated one by one, for example, the underground of a certain unit is high compression soft soil, and the above is the main transformer, and the comprehensive score is obtained after superimposing the scores of other factors. Through this settlement risk area evaluation, the settlement risk degree of each space unit of the substation is calculated, sorted and classified based on quantitative indexes, and finally all units are divided into high-risk areas with scores ≥ 8, such as equipment foundation under soft soil layer area, medium-risk areas with scores 5-7, such as auxiliary building area, low-risk areas with scores < 5, such as open space, to form a set of substation graded risk area, which marks the spatial distribution of risk levels in each area.

[0033] Based on the graded risk area set, the deployment position of the substation monitoring sensor is determined in the high-risk area first, that is, the specific point that needs to be monitored is selected, the high-risk area needs to cover the key nodes such as the four corners of the equipment foundation and the cable trench corner, and the representative point is selected in the low-risk area to avoid blind area; at the same time, the regional deployment density of the monitoring sensor is determined according to the risk level difference. After the planning is completed, the monitoring network deployment is carried out in the target substation site, that is, the process of actually installing equipment according to the planning, including cable fixation and mechanical connection debugging of magnetostrictive sensing array, installation and wireless signal calibration of WAPI anchor point, synchronous construction of data transmission link to ensure real-time communication, integration of magnetostrictive sensing array responsible for collecting foundation mechanical deformation data and WAPI anchor point network responsible for collecting foundation deformation positioning data, and finally construction of a dual-mode sensing monitoring network.

[0034] Further, the data processing channel of the foundation deformation is built, including: The magnetostrictive sensor array and the WAPI anchor network are preprocessed and analyzed to generate a dual-mode data preprocessing channel; the reliability influence of the magnetostrictive sensor array and the WAPI anchor network is analyzed to obtain a set of sensor reliability influence factors; the output data of the dual-mode data preprocessing channel is fused and analyzed based on the set of sensor reliability influence factors to construct a dual-mode data fusion channel; and the dual-mode data preprocessing channel and the dual-mode data fusion channel are combined in series to build the ground deformation data processing channel.

[0035] Specifically, first, the magnetostrictive sensor array and the WAPI anchor network are preprocessed and analyzed. Preprocessing and analysis refers to analyzing the problems existing in the original data of the two types of sensors, such as noise interference, format difference, and error deviation, based on the core characteristics of the two types of sensors, and designing a targeted preprocessing program. Specifically, for the magnetostrictive sensor array, combining its mechanical-electromagnetic conversion characteristics, deformation causes electromagnetic signal changes, which are easily affected by temperature and electromagnetic interference. The preprocessing program includes data cleaning, removing abnormal electromagnetic signal values generated by sensor faults, noise filtering, using wavelet transform to filter environmental electromagnetic interference, temperature compensation, and offsetting the influence of temperature on conversion accuracy through a temperature-electromagnetic signal correction model; for the WAPI anchor network, based on its signal parameter characteristics, through signal strength and propagation time delay positioning, it is easily affected by shielding and multipath effect. The preprocessing program includes coordinate conversion, unifying the local coordinates of the anchor points to the substation geodetic coordinate system, error correction, using Kalman filter to eliminate positioning deviation caused by multipath propagation, and data alignment, synchronizing the timestamps of the dual-mode data; the two preprocessing programs are integrated in parallel to enable the original data of the two types of sensors to be processed synchronously, and finally a dual-mode data preprocessing channel is generated.

[0036] Next, the magnetostrictive sensor array and the WAPI anchor network are analyzed for reliability influence. This analysis refers to the process of identifying various factors that may cause the reliability of the data of the two types of sensors to decrease, and forming a structured set. Through field research, device parameter analysis, and historical fault backtracking, the core factors affecting reliability are identified: for the magnetostrictive sensor array, the main factors include electromagnetic interference intensity, interference generated by substation equipment operation, sensor aging degree, running time and performance decay correlation, cable connection stability, and transmission link packet loss rate; for the WAPI anchor network, the main factors include signal shielding conditions, shielding of wireless signals by buildings and equipment, environmental temperature and humidity fluctuations, influence on signal propagation speed, anchor synchronization accuracy, and time synchronization deviation between multiple anchors; these factors are classified and organized to form a set of quantifiable data reliability influence factors that cover both types of sensors, i.e., a set of sensor reliability influence factors.

[0037] Subsequently, the output data of the dual-mode data preprocessing channel is fused and analyzed based on the set of sensor reliability influencing factors, and the output data refers to the magnetostrictive mechanical deformation data and WAPI positioning data after preprocessing by the preprocessing channel, which eliminates noise, error and formats the data. The data has basic usability but does not realize advantage complementation. Fusion analysis refers to the process of combining data reliability dynamic weight adjustment to integrate two single-mode data into fusion data with precision and position information. First, the basic weight of dual-mode data is calculated by entropy weight distribution, reflecting the stability of the data itself. Then, the dual-mode weight influence function is constructed based on the set of sensor reliability influencing factors, such as down-regulating the magnetostrictive data weight when electromagnetic interference is enhanced. The function is used to modify the basic weight to obtain the target weight, which reflects the data credibility in real time. Finally, the preprocessing output data is weighted and fused using the target weight to construct a dual-mode data fusion channel that can output high-quality integrated data.

[0038] Finally, the dual-mode data preprocessing channel and the dual-mode data fusion channel are combined in series. Series combination refers to the data flow direction of the two channels, which allows the original data to flow through the channels in the order of "preprocessing first, then fusion". The original data of the magnetostrictive sensor array and the WAPI anchor network first enters the dual-mode data preprocessing channel to complete purification and standardization. The output data after preprocessing automatically flows into the dual-mode data fusion channel to complete dynamic weighted fusion. Through this sequential connection, the coherence and logic of the data processing process are ensured, and finally a complete ground deformation data processing channel is built.

[0039] Further, the dual-mode data preprocessing channel is generated, comprising: According to the mechanical electromagnetic conversion characteristics of the magnetostrictive sensor array, a data preprocessing program is constructed, which includes data cleaning, noise filtering and temperature compensation. Based on the signal parameter characteristics of the WAPI anchor network, a positioning preprocessing program is constructed, which includes coordinate conversion, error correction and data alignment. The data preprocessing program and the positioning preprocessing program are integrated in parallel to generate the dual-mode data preprocessing channel.

[0040] Specifically, first, for magnetostrictive sensor array, according to its mechanical electromagnetic conversion characteristics, mechanical electromagnetic conversion characteristics refer to the electromagnetic signal parameter change of magnetostrictive material when bearing foundation mechanical deformation, and the electromagnetic signal change also reversely causes material mechanical deformation, a special data preprocessing program is constructed, which is a set of programs for purifying, correcting and standardizing the original electromagnetic signal data output by the magnetostrictive sensor array, the purpose is to eliminate data interference and restore the electromagnetic signal corresponding to the real foundation deformation. The program contains three core operations: first, data cleaning, by setting a reasonable threshold, such as removing electromagnetic signal values exceeding 10 times the normal deformation range, filtering abnormal data caused by sensor transient failure and line interference, and filling a small amount of missing data by using linear interpolation method to ensure data integrity; second, noise filtering, using wavelet transform filtering algorithm to separate and remove environmental electromagnetic radiation, such as electromagnetic interference generated by substation equipment operation and sensor circuit noise affecting original signal, and retaining effective signal components directly related to foundation mechanical deformation; third, temperature compensation, by embedding temperature-electromagnetic signal correction model, real-time correcting the influence of environmental temperature change on the conversion accuracy of magnetostrictive material, such as the electromagnetic conversion coefficient of the material may deviate by 2% when the temperature rises by 10℃, which needs to be offset by compensation formula to ensure data accuracy.

[0041] Subsequently, for WAPI anchor network, based on its signal parameter characteristics, signal parameter characteristics refer to the calculation of target position by WAPI anchor through transmitting and receiving wireless signals using signal strength, propagation delay, angle of arrival and other parameters, these signal parameters are easily affected by factors such as shielding and multipath propagation, a suitable positioning preprocessing program is constructed, which is a set of programs for coordinate unification, error correction and time synchronization of original positioning related signal parameters output by WAPI anchor network, the purpose is to improve the accuracy and availability of positioning data. The program also contains: first, coordinate conversion, the preliminary positioning coordinates calculated by WAPI anchor based on local coordinate system, such as relative coordinate system with a device in substation as origin, are converted into national geodetic coordinate system consistent with integrated space model of substation through Gauss projection transformation algorithm, realizing coordinate unification of positioning data and space model; second, error correction, using Kalman filter algorithm to eliminate signal multipath effect, time delay error caused by superposition of reflected signals and direct signals after ground and wall reflection, shielding attenuation, such as signal strength anomaly caused by tree and equipment shielding, and other factors causing positioning deviation, the positioning error is controlled within centimeter level; third, data alignment, by extracting the timestamp of magnetostrictive sensor array data, synchronously adjusting the collection time node of WAPI anchor network data, ensuring that the two modal data correspond in time dimension.

[0042] After the construction of the two single-mode preprocessing procedures described above, the data preprocessing procedure is integrated with the positioning preprocessing procedure in parallel. Since the two procedures are independent for magnetostrictive data and WAPI positioning data, parallel architecture can realize the synchronous parallel preprocessing of two types of original data, avoiding the efficiency loss caused by serial processing. The integrated dual-mode data preprocessing channel is a processing channel that is specifically used to synchronize the processing of magnetostrictive sensor array original mechanical deformation related data and WAPI anchor network original positioning related data, output normalized and high-precision single-mode data. It can simultaneously receive clean and accurate time-aligned ground mechanical deformation data and ground deformation positioning data after being processed by the respective preprocessing procedures.

[0043] Further, the dual-mode data fusion channel is constructed, comprising: The entropy weight distribution calculation is performed on the magnetostrictive sensor array and the WAPI anchor network to obtain the basic dual-mode data weight factor; the dual-mode weight influence function is constructed according to the set of sensor reliability influence factors; the basic dual-mode data weight factor is dynamically influenced and corrected based on the dual-mode weight influence function to determine the target dual-mode data weight factor; the output data of the dual-mode data preprocessing channel is weighted and fused for analysis using the target dual-mode data weight factor to construct the dual-mode data fusion channel.

[0044] Specifically, firstly, the entropy weight distribution calculation is performed on the pretreated data of the magnetostrictive sensor array and the WAPI anchor network, which is specifically: collecting continuous normalized data output by the dual-mode data preprocessing channel, such as ground mechanical deformation data and ground deformation positioning data every 15 minutes within 3 days, to form two types of data sample sets; standardizing the sample sets to eliminate dimensional differences; 3. Calculate the probability of each group of data in the corresponding mode, that is, the data probability at a certain time = the data value at that time ÷ the total data of all times in that mode); based on the probability, the information entropy of each mode data is calculated, and the smaller the entropy value, the lower the data dispersion and the higher the stability; according to the formula: weight = (1-entropy value) ÷ (2-sum of entropy values of two modes), the initial weight is calculated, and the sum of the weights of the two modes is ensured to be 1, and the initial weight obtained through this process, which reflects the stability of the data itself, is the basic dual-mode data weight factor. After completing the basic weight calculation, the sensor reliability influence factor set is introduced, which is the sum of key variables affecting the real-time reliability of dual-mode sensor data, including sensor aging degree, quantified by running time, 0=new equipment, 1=service life beyond design, environmental interference intensity, such as electromagnetic interference level generated by substation equipment operation, seasonal temperature and humidity fluctuation, quantified to 0-1 interval, data transmission packet loss rate, 0=no packet loss, 1=packet loss rate>10% and the like. These factors will dynamically weaken or enhance the data reliability and need to be included in the weight adjustment system. Based on this set, a dual-mode weight influence function is constructed, and the function form is: ; wherein is the corrected single-mode weight; is the basic dual-mode data weight factor; is the influence coefficient, which is calibrated by the power industry experiment and takes a value of 0-1, such as the influence coefficient of electromagnetic interference on magnetostrictive data is 0.38, and the influence coefficient of aging degree is 0.22; is the real-time quantized value of the reliability influence factor, which is then dynamically corrected. The quantized data of the reliability influence factor set is collected in real time through the state monitoring module of the sensor, and these real-time data are substituted into the dual-mode weight influence function at each time to adjust the basic dual-mode data weight factor, so that the weight is dynamically updated with the sensor working state and environmental changes. Finally, the target dual-mode data weight factor is obtained, so as to reflect the data reliability in real time.

[0045] Next, it is necessary to clarify the processing object as the output data, that is, the processing result of the dual-mode data preprocessing channel, including the ground mechanical deformation data after data cleaning, noise filtering and temperature compensation, and the ground deformation positioning data after coordinate conversion, error correction and data alignment. Based on the target dual-mode data weight factor, the ground mechanical deformation data is multiplied by the magnetostrictive target weight to obtain the weighted contribution value of the deformation data, and the high confidence deformation data is highlighted to play a leading role in the fusion result. Multiply the ground deformation positioning data by the WAPI target weight to obtain the weighted contribution value of the positioning data, and strengthen the position correlation of the reliable positioning data. Then, through coordinate-deformation binding, the two types of contribution values are integrated to form a fusion data set with specific geographic coordinates, accurate deformation values and confidence labels, realizing the complementarity of magnetostrictive precision advantage and WAPI positioning advantage.

[0046] Finally, the above whole process is solidified into an automatic processing module, that is, the dual-mode data fusion channel is constructed. The channel is a carrier for realizing the independent specification to collaborative value-added of dual-mode data, and can stably output high-quality fusion data.

[0047] Further, the ground settlement deformation prediction network is trained, comprising: designing a hybrid neural network architecture, the hybrid neural network architecture comprising a TCN time convolution network and a GRU gated recurrent unit; collecting a ground settlement historical deformation data set of the dual-mode sensing monitoring network; based on the hybrid neural network architecture, the ground settlement historical deformation data set is trained for deformation prediction to obtain a ground settlement deformation prediction network.

[0048] Specifically, first, a hybrid neural network architecture is designed, which is a composite model architecture combining the advantages of two different neural networks and constructed for the characteristics of ground settlement time series data. The core is to improve the prediction accuracy through functional complementation. The architecture is composed of TCN time convolution network and GRU gate recurrent unit. The TCN time convolution network has the ability of time dimension convolution operation. Through causal convolution, it ensures that the prediction does not depend on future data and expands the time receptive field through dilated convolution, accurately capturing the local time series characteristics of ground settlement data, such as the slight fluctuation of ground deformation rate within a single day and the deformation mutation caused by short-term environmental changes. The GRU gate recurrent unit captures the long and short term time series dependence of the settlement data through the gating mechanism of the reset gate to filter useful information in the historical data and the update gate to retain key long-term features, such as the continuous monthly settlement increasing trend and the periodic deformation law caused by seasonal factors. In the architecture design, the TCN time convolution network receives the processed time series data first, extracts local features and converts them into feature vectors; then the feature vectors are input into the GRU gate recurrent unit to further mine the long-term time series correlation of the data; finally, a fully connected layer is added at the output end of the GRU to map the processing results to specific ground settlement prediction values, such as daily average settlement values for the next 7 days.

[0049] Then the ground settlement historical deformation data set is collected and preprocessed, which refers to the historical monitoring data set accumulated by the dual-mode sensing monitoring network during the long-term operation of the target substation. It specifically includes ground mechanical deformation historical data collected by magnetostrictive sensing array, such as hourly deformation values in the past 1-3 years, ground deformation positioning historical data collected by WAPI anchor point network, corresponding to the geographical coordinates where deformation occurs. First, these original historical data are input into the ground deformation data processing channel built earlier, and the dual-mode data preprocessing is completed in turn, i.e. cleaning abnormal values, filtering and denoising, time alignment and weighted fusion, i.e. integrating into deformation value-position-time stamp data according to dynamic weights, and then arranging the fused historical data into a structured data set according to the time sequence for performance evaluation and test set in the training process, and for verification of the generalization ability of the final model.

[0050] Then the deformation prediction training is carried out, which refers to inputting the preprocessed ground settlement historical deformation data set into the hybrid neural network architecture. After training, the ground settlement deformation prediction network is obtained, which is a model trained and optimized by historical data and has stable time series prediction ability. It can receive real-time ground settlement deformation fusion data, automatically complete local feature extraction and long-term time series dependence processing, and output ground settlement deformation prediction values for a specific period in the future.

[0051] Further, the ground settlement deformation prediction network is obtained, comprising: The ground settlement history deformation data set is processed and identified based on the ground deformation data processing channel to obtain a ground settlement deformation sample set; the mixed neural network architecture is used to perform deformation prediction training on the ground settlement deformation sample set to generate an initial settlement deformation prediction network; the initial settlement deformation prediction network is subjected to loss evaluation calculation and iterative optimization update to obtain the ground settlement deformation prediction network.

[0052] Specifically, first, based on the ground deformation data processing channel built earlier, the ground settlement history deformation data set is processed and identified. The processing and identification refers to a combination of standardization preprocessing and structured labeling of original historical data, which converts chaotic historical monitoring data into standardized data meeting the requirements of mixed neural network training. The ground settlement history deformation data set, including historical mechanical deformation data of magnetostrictive sensing array and historical positioning data of WAPI anchor network, is input into the ground deformation data processing channel. Data cleaning is completed through the dual-modal data preprocessing channel, including removing abnormal values caused by sensor failure, filling in a small amount of missing data, noise filtering, filtering the influence of environmental electromagnetic interference on deformation data, temperature compensation, correcting the deviation of magnetostrictive data caused by temperature change, and coordinate conversion. The coordinate system of the unified positioning data is converted, the data is aligned, and the timestamps of the dual-modal data are synchronized. Then, the standardized data after preprocessing is structured and labeled. The input-output sample pair format is identified to form a structured training unit. Finally, all training units are integrated to obtain a ground settlement deformation sample set. This sample set is training data with time sequence correlation and accurate labels. It is divided into a training set in a ratio of 8:2 for model parameter learning and a validation set for performance checking during the training process.

[0053] Then, the designed hybrid neural network architecture is used to carry out deformation prediction training on the ground settlement deformation sample set. The deformation prediction training refers to inputting sample data into the network, calculating the prediction results through forward propagation, and adjusting the parameters through back propagation, so that the network gradually learns the time series law of settlement data. Specifically: sample input: the input features of the training set are imported into the hybrid neural network in batches. The local time series law in the input features, such as short-term deformation rate fluctuation, is extracted by the TCN time convolution network through causal convolution and dilated convolution, and the local feature vector is output. Feature transmission and processing: the local feature vector output by the TCN is transmitted to the GRU gate recurrent unit. The reset gate selects effective historical information, and the update gate retains key long-term features to mine the long-term and short-term time series dependence between the input features and the output labels, such as monthly settlement trend. Prediction output: a fully connected layer is connected after the GRU unit to map the processed features to specific settlement prediction values, completing the forward propagation of a single batch of samples. Repeat the process until all samples in the training set are traversed. At this time, the network has preliminarily mastered the settlement prediction law, and an initial settlement deformation prediction network is generated. Although the initial network can output prediction results, it has problems of low prediction accuracy and weak generalization ability due to not being optimized.

[0054] Subsequently, the initial settlement deformation prediction network is subjected to loss evaluation calculation and iterative optimization update. The loss evaluation calculation quantifies the deviation of the network prediction value from the sample true label, i.e., the output label, by using a pre-set loss function. In the scheme, the mean square error is usually used as the loss function, and the formula is: ; is the loss value, is the true settlement value, is the prediction value, is the number of samples, and the validation set data is used to evaluate the generalization ability of the initial network. If the validation set loss is much higher than the training set loss, it indicates that the network has overfitting problem. The iterative optimization update is a process of adjusting the network parameters based on the loss evaluation results. The Adam optimizer uses the back propagation algorithm to adjust the convolution kernel size, dilation coefficient of TCN, and the number of hidden layer nodes, learning rate of GRU, and other parameters in each round according to the loss gradient direction. The training set and validation set loss are recalculated after each optimization round until the pre-set convergence condition is met, such as the validation set loss decreasing for 10 consecutive rounds or reaching the pre-set training rounds. After multiple rounds of loss evaluation and parameter optimization, the prediction accuracy and generalization ability of the initial settlement deformation prediction network are improved, and a stable ground settlement deformation prediction network is finally obtained. This network can receive real-time ground settlement deformation fusion data, automatically complete time series feature extraction and settlement law matching, and output accurate ground settlement deformation prediction values.

[0055] Further, the settlement abnormal response mechanism based on the ground settlement deformation prediction triggers a hierarchical response warning, comprising: When the ground settlement deformation prediction reaches the preset settlement threshold, the settlement abnormal response mechanism triggers a warning matching of the ground settlement deformation prediction, determines the ground settlement deformation warning level, and based on the ground settlement deformation warning level, the target substation is given a hierarchical response warning and settlement emergency treatment.

[0056] Specifically, first, when the ground settlement deformation prediction output by the ground settlement deformation prediction network reaches the preset settlement threshold, the ground settlement deformation prediction contains the settlement value, deformation rate and corresponding geographic coordinates in a specific future period, and the preset settlement threshold is a risk threshold value preset by structural mechanics calculation and historical failure case analysis in combination with the structural characteristics of different regions of the target substation and the safety specifications of the power industry. When the prediction reaches any level threshold, the settlement abnormal response mechanism is activated immediately, which converts the abstract prediction into explicit response instructions to avoid the risk of delay caused by manual judgment. The settlement abnormal response mechanism quantitatively compares the ground settlement deformation prediction with the preset settlement threshold through the built-in algorithm, and calculates the comprehensive risk score by combining the importance weight of the warning area, and finally determines the corresponding ground settlement deformation warning level. In the scheme, it is divided into three levels: level I mild warning: the prediction exceeds the mild threshold but does not reach the moderate threshold, and the region is of low importance, with no direct safety risk; level II moderate warning: the prediction exceeds the moderate threshold or exceeds the mild threshold but is located in a high importance region, with potential safety hazards; level III severe warning: the prediction exceeds the severe threshold or the core region exceeds the moderate threshold, directly threatening equipment operation and structural safety, ensuring that the level division can accurately reflect the risk severity.

[0057] After the determination of the early warning level is completed, the hierarchical response early warning and the settlement emergency treatment are synchronously promoted. The hierarchical response early warning is to adopt a differentiated information transmission mode according to the early warning level, to ensure that the relevant subjects obtain the risk information in time. When the I-level mild early warning occurs, the system automatically sends a short message and a background system pop-up window to the substation operation and maintenance personnel, to clearly indicate the early warning area coordinates, the predicted value and the threshold comparison data. When the II-level moderate early warning occurs, in addition to the short message reminder, the substation duty room sound and light alarm is additionally triggered, to synchronously notify the technical person in charge to start the risk assessment. When the III-level severe early warning occurs, the emergency command platform is used to push the emergency alarm to the operation and maintenance team and the superior power management department in real time, with the early warning area three-dimensional model and the settlement trend curve attached, to facilitate the rapid grasp of the risk panorama. The settlement emergency treatment is the practical control measure corresponding to the level, aiming to timely control the risk spread. When the I-level mild early warning occurs, the monitoring frequency of the early warning area is increased, and the dynamic tracking of the data is strengthened. When the II-level moderate early warning occurs, the technical personnel are organized to go to the site to investigate the settlement causes by carrying portable monitoring equipment, and a safety assessment report is issued. When the III-level severe early warning occurs, the core equipment emergency shutdown plan is immediately started, such as disconnecting the transformer power supply in the affected area, and at the same time, the foundation reinforcement team is linked, to push the accurate positioning data of the early warning area, to guide the development of grouting reinforcement and other settlement control operations, to avoid equipment damage or power interruption accidents.

[0058] In summary, the double-mode substation foundation settlement deformation dynamic prediction method provided by the application has the following technical effects: By deploying a double-mode sensing and monitoring network composed of magnetostrictive sensing arrays and WAPI anchor point networks at the target substation, the deformation and positioning two-dimensional data are completed. A foundation deformation data processing channel with dynamic weight adjustment is built to improve the data reliability. A hybrid prediction network composed of a TCN time convolution network and a GRU gate recurrent unit is designed to reduce the prediction error and predict the settlement trend in advance. A hierarchical response early warning mechanism based on the settlement prediction value and the importance of the area is established, to change passive monitoring to active prediction and disposal, to avoid equipment tilting and power interruption, and to achieve the technical effect of ensuring the long-term safe and stable operation of the substation.

[0059] Embodiment two, based on the same inventive concept as the double-mode substation foundation settlement deformation dynamic prediction method in the aforementioned embodiment one, the application also provides a double-mode substation foundation settlement deformation dynamic prediction system, please refer to the attached Figure 2 The double-mode substation foundation settlement deformation dynamic prediction system comprises: The data acquisition module 11 deploys a dual-mode sensing monitoring network on the target transformer substation, the dual-mode sensing monitoring network specifically includes a magnetostrictive sensing array and a WAPI anchor point network, ground foundation mechanical deformation data and ground foundation deformation positioning data are acquired through the dual-mode sensing monitoring network; the channel building module 12 builds a ground foundation deformation data processing channel, the ground foundation deformation data processing channel is composed of a dual-mode data preprocessing channel and a dual-mode data fusion channel in series; the data processing module 13 pre-processes and weightedly fuses the ground foundation mechanical deformation data and the ground foundation deformation positioning data by using the ground foundation deformation data processing channel to obtain ground foundation settlement deformation fusion data; the dynamic prediction module 14 trains a ground foundation settlement deformation prediction network, introduces the ground foundation settlement deformation prediction network to dynamically predict the ground foundation settlement deformation fusion data, determines a ground foundation settlement deformation prediction value, and triggers a settlement abnormal response mechanism to perform hierarchical response early warning based on the ground foundation settlement deformation prediction value.

[0060] Further, the data acquisition module 11 in the dual-mode transformer substation ground foundation settlement deformation dynamic prediction system is used to: Acquire structure distribution data and equipment layout data of the target transformer substation, perform three-dimensional reconstruction based on the structure distribution data and the equipment layout data, establish a transformer substation three-dimensional model; perform geological exploration analysis on the target transformer substation to generate a transformer substation geological structure model; perform spatial fusion analysis on the transformer substation geological structure model and the transformer substation three-dimensional model to obtain a transformer substation integrated space model; deploy a monitoring network based on the transformer substation integrated space model to construct a dual-mode sensing monitoring network.

[0061] The data acquisition module 11 is further used to: Identify a ground foundation settlement influence factor set, establish a risk quantification evaluation index system according to the ground foundation settlement influence factor set, perform settlement risk area evaluation on the transformer substation integrated space model according to the risk quantification evaluation index system to obtain a transformer substation hierarchical risk area set, determine a transformer substation monitoring sensor deployment position and a monitoring sensor area deployment density according to the transformer substation hierarchical risk area set, and deploy a monitoring network based on the transformer substation monitoring sensor deployment position and the monitoring sensor area deployment density on the target transformer substation to construct a dual-mode sensing monitoring network.

[0062] Further, the channel building module 12 in the dual-mode transformer substation ground foundation settlement deformation dynamic prediction system is used to: The magnetostrictive sensor array and the WAPI anchor network are preprocessed and analyzed to generate a dual-mode data preprocessing channel; the reliability influence of the magnetostrictive sensor array and the WAPI anchor network is analyzed to obtain a set of sensor reliability influence factors; the output data of the dual-mode data preprocessing channel is fused and analyzed based on the set of sensor reliability influence factors to construct a dual-mode data fusion channel; and the dual-mode data preprocessing channel and the dual-mode data fusion channel are combined in series to build the ground deformation data processing channel.

[0063] The channel building module 12 is further configured to: According to the mechanical electromagnetic conversion characteristics of the magnetostrictive sensor array, a data preprocessing program is constructed, the data preprocessing program including data cleaning, noise filtering and temperature compensation; according to the signal parameter characteristics of the WAPI anchor network, a positioning preprocessing program is constructed, the positioning preprocessing program including coordinate conversion, error correction and data alignment; and the data preprocessing program and the positioning preprocessing program are integrated in parallel to generate the dual-mode data preprocessing channel.

[0064] The channel building module 12 is further configured to: The magnetostrictive sensor array and the WAPI anchor network are subjected to entropy weight allocation calculation to obtain a basic dual-mode data weight factor; a dual-mode weight influence function is constructed according to the set of sensor reliability influence factors; the basic dual-mode data weight factor is dynamically influenced and corrected based on the dual-mode weight influence function to determine a target dual-mode data weight factor; and the output data of the dual-mode data preprocessing channel is weighted and fused and analyzed using the target dual-mode data weight factor to construct the dual-mode data fusion channel.

[0065] Further, the dynamic prediction module 14 in the dual-mode substation ground settlement deformation dynamic prediction system is configured to: A hybrid neural network architecture is designed, the hybrid neural network architecture including a TCN time convolution network and a GRU gated recurrent unit; a ground settlement historical deformation data set of the dual-mode sensor monitoring network is collected; and the ground settlement historical deformation data set is subjected to deformation prediction training based on the hybrid neural network architecture to obtain a ground settlement deformation prediction network.

[0066] The dynamic prediction module 14 is further configured to: The ground foundation deformation data processing channel is used to process the ground foundation settlement history deformation data set to obtain a ground foundation settlement deformation sample set; the mixed neural network architecture is used to perform deformation prediction training on the ground foundation settlement deformation sample set to generate an initial settlement deformation prediction network; and the initial settlement deformation prediction network is subjected to loss evaluation calculation and iterative optimization update to obtain the ground foundation settlement deformation prediction network.

[0067] The dynamic prediction module 14 is further configured to: When the ground foundation settlement deformation prediction value reaches a preset settlement threshold, a settlement abnormality response mechanism is triggered to perform early warning matching on the ground foundation settlement deformation prediction value to determine a ground foundation settlement deformation early warning level; and the target transformer substation is subjected to hierarchical response early warning and settlement emergency treatment based on the ground foundation settlement deformation early warning level.

[0068] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. Figure 1 The dual-mode transformer substation ground foundation settlement deformation dynamic prediction method and specific examples in Embodiment One are also applicable to the dual-mode transformer substation ground foundation settlement deformation dynamic prediction system of the present embodiment. Based on the foregoing detailed description of the dual-mode transformer substation ground foundation settlement deformation dynamic prediction method, those skilled in the art can clearly understand the dual-mode transformer substation ground foundation settlement deformation dynamic prediction system of the present embodiment. Therefore, in the interest of brevity, the dual-mode transformer substation ground foundation settlement deformation dynamic prediction system will not be described in detail here. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant parts are described in the method section.

[0069] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0070] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application also intends to include these modifications and variations.

Claims

1. A method for dynamic prediction of foundation settlement and deformation in a dual-mode substation, characterized in that, The method includes: A dual-modal sensing and monitoring network is deployed on the target substation. The dual-modal sensing and monitoring network specifically includes a magnetostrictive sensor array and a WAPI anchor network. The dual-modal sensing and monitoring network is used to collect ground mechanical deformation data and ground deformation location data. A foundation deformation data processing channel is established, which consists of a dual-modal data preprocessing channel and a dual-modal data fusion channel connected in series. The foundation mechanical deformation data and foundation deformation positioning data are preprocessed and weighted and fused using the foundation deformation data processing channel to obtain foundation settlement deformation fused data. A foundation settlement deformation prediction network is trained, and the foundation settlement deformation prediction network is introduced to dynamically predict the fused foundation settlement deformation data, determine the predicted amount of foundation settlement deformation, and trigger a settlement anomaly response mechanism based on the predicted amount of foundation settlement deformation to carry out graded response and early warning.

2. The method for dynamic prediction of foundation settlement and deformation of a dual-mode substation as described in claim 1, characterized in that, The deployment of a dual-modal sensing and monitoring network at the target substation includes: Collect structural distribution data and equipment layout data of the target substation, and perform three-dimensional reconstruction based on the structural distribution data and equipment layout data to establish a three-dimensional model of the substation; Geological exploration and analysis were conducted on the target substation to generate a geological structure model of the substation; The geological structure model of the substation is spatially fused with the three-dimensional model of the substation to obtain an integrated spatial model of the substation. Based on the integrated spatial model of the substation, a monitoring network is deployed to construct a dual-modal sensing monitoring network.

3. The method for dynamic prediction of foundation settlement and deformation of a dual-mode substation as described in claim 2, characterized in that, The construction of the dual-modal sensing and monitoring network includes: Identify and obtain a set of factors affecting foundation settlement, and establish a risk quantification assessment index system based on the set of factors affecting foundation settlement; The substation integrated spatial model is assessed for settlement risk areas according to the risk quantification assessment index system to obtain a set of graded risk areas for the substation. Based on the set of risk zones for substation classification, determine the deployment locations and regional deployment density of substation monitoring sensors; Based on the deployment location and regional deployment density of the monitoring sensors in the substation, a monitoring network is deployed on the target substation to construct a dual-modal sensing monitoring network.

4. The method for dynamic prediction of foundation settlement and deformation of a dual-mode substation as described in claim 1, characterized in that, The establishment of the foundation deformation data processing channel includes: The magnetostrictive sensing array and the WAPI anchor network are preprocessed and analyzed to generate a dual-modal data preprocessing channel. A reliability impact analysis was performed on the magnetostrictive sensing array and the WAPI anchor network to obtain a set of factors affecting sensor reliability. Based on the set of factors affecting sensor reliability, the output data of the dual-modal data preprocessing channel is fused and analyzed to construct a dual-modal data fusion channel; The dual-modal data preprocessing channel and the dual-modal data fusion channel are connected in series to form the foundation deformation data processing channel.

5. The method for dynamic prediction of foundation settlement and deformation of a dual-mode substation as described in claim 4, characterized in that, The generation of dual-modal data preprocessing channels includes: Based on the mechanical-electromagnetic conversion characteristics of the magnetostrictive sensing array, a data preprocessing procedure is constructed, which includes data cleaning, noise filtering, and temperature compensation. Based on the signal parameter characteristics of the WAPI anchor network, a positioning preprocessing procedure is constructed, which includes coordinate transformation, error correction, and data alignment. The data preprocessing program and the positioning preprocessing program are integrated in parallel to generate the dual-modal data preprocessing channel.

6. The method for dynamic prediction of foundation settlement and deformation of a dual-mode substation as described in claim 4, characterized in that, The construction of the dual-modal data fusion channel includes: Entropy weight allocation calculations are performed on the magnetostrictive sensing array and the WAPI anchor network to obtain the basic bimodal data weighting factor. Based on the set of factors affecting sensor reliability, a dual-modal weighted influence function is constructed. The target bimodal data weight factor is determined by dynamically influencing and correcting the weight factor of the basic bimodal data based on the bimodal weight influence function. The target bimodal data weighting factor is used to perform weighted fusion analysis on the output data of the bimodal data preprocessing channel to construct the bimodal data fusion channel.

7. The method for dynamic prediction of foundation settlement and deformation of a dual-mode substation as described in claim 1, characterized in that, The training foundation settlement deformation prediction network includes: Design a hybrid neural network architecture, which includes a TCN temporal convolutional network and a GRU gated recurrent unit; Collect the historical deformation dataset of foundation settlement of the dual-modal sensing monitoring network; Based on the hybrid neural network architecture, deformation prediction training is performed on the historical deformation dataset of foundation settlement to obtain a foundation settlement deformation prediction network.

8. The method for dynamic prediction of foundation settlement and deformation of a dual-mode substation as described in claim 7, characterized in that, The foundation settlement deformation prediction network includes: Based on the foundation deformation data processing channel, the historical foundation settlement deformation dataset is processed and identified to obtain a foundation settlement deformation sample set. The hybrid neural network architecture is used to train the foundation settlement deformation sample set for deformation prediction, thereby generating an initial settlement deformation prediction network. The initial settlement deformation prediction network is subjected to loss assessment calculation and iterative optimization and update to obtain the foundation settlement deformation prediction network.

9. The method for dynamic prediction of foundation settlement and deformation of a dual-mode substation as described in claim 1, characterized in that, The graded response and early warning mechanism based on the predicted foundation settlement deformation triggering settlement anomaly response includes: When the predicted amount of foundation settlement deformation reaches the preset settlement threshold, the settlement anomaly response mechanism is triggered to perform early warning matching on the predicted amount of foundation settlement deformation and determine the early warning level of foundation settlement deformation. Based on the aforementioned foundation settlement and deformation early warning level, the target substation is subject to graded response early warning and settlement emergency treatment.

10. A dynamic prediction system for foundation settlement and deformation of a dual-mode substation, characterized in that, The system is used to implement the dynamic prediction method for foundation settlement and deformation of a dual-mode substation as described in any one of claims 1 to 9, and the system comprises: The data acquisition module deploys a dual-modal sensing and monitoring network on the target substation. The dual-modal sensing and monitoring network specifically includes a magnetostrictive sensor array and a WAPI anchor network. The dual-modal sensing and monitoring network is used to collect foundation mechanical deformation data and foundation deformation location data. The channel construction module constructs a foundation deformation data processing channel, which is composed of a dual-modal data preprocessing channel and a dual-modal data fusion channel connected in series. The data processing module uses the foundation deformation data processing channel to preprocess and weightedly fuse the foundation mechanical deformation data and foundation deformation positioning data to obtain foundation settlement deformation fused data. The dynamic prediction module trains a foundation settlement deformation prediction network, introduces the foundation settlement deformation prediction network to dynamically predict the fused foundation settlement deformation data, determines the predicted foundation settlement deformation amount, and triggers a settlement anomaly response mechanism based on the predicted foundation settlement deformation amount to provide a graded response and early warning.

Images