CN115240394A - Method and system for monitoring and early warning water level of accident oil pool of transformer substation - Google Patents
Method and system for monitoring and early warning water level of accident oil pool of transformer substation Download PDFInfo
- Publication number
- CN115240394A CN115240394A CN202211158932.XA CN202211158932A CN115240394A CN 115240394 A CN115240394 A CN 115240394A CN 202211158932 A CN202211158932 A CN 202211158932A CN 115240394 A CN115240394 A CN 115240394A
- Authority
- CN
- China
- Prior art keywords
- water level
- oil pool
- accident oil
- monitoring
- historical
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B31/00—Predictive alarm systems characterised by extrapolation or other computation using updated historic data
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01F—MEASURING VOLUME, VOLUME FLOW, MASS FLOW OR LIQUID LEVEL; METERING BY VOLUME
- G01F23/00—Indicating or measuring liquid level or level of fluent solid material, e.g. indicating in terms of volume or indicating by means of an alarm
- G01F23/04—Indicating or measuring liquid level or level of fluent solid material, e.g. indicating in terms of volume or indicating by means of an alarm by dip members, e.g. dip-sticks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/52—Scale-space analysis, e.g. wavelet analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/19—Recognition using electronic means
- G06V30/191—Design or setup of recognition systems or techniques; Extraction of features in feature space; Clustering techniques; Blind source separation
- G06V30/19147—Obtaining sets of training patterns; Bootstrap methods, e.g. bagging or boosting
Abstract
The invention provides a method and a system for monitoring and early warning the water level of an accident oil pool of a transformer substation, wherein the method comprises the following steps: continuously acquiring a water gauge image in the accident oil pool through a camera, and identifying the water gauge image in the accident oil pool to obtain a real-time water level reading in the accident oil pool; acquiring historical water level data of each monitoring point in the peripheral area of the current transformer substation and historical water level data in the accident oil pool, and establishing a water level change prediction model according to the historical water level data of each monitoring point and the historical water level data in the accident oil pool; acquiring real-time water level data of each current monitoring point, inputting a water level change prediction model, and predicting and outputting the water level change condition in the accident oil pool; and carrying out water level monitoring and early warning according to the predicted water level change condition in the accident oil pool. The invention can not only carry out real-time water level monitoring, but also carry out the early warning of the future water level change trend, thereby improving the prospect of the water level monitoring and early warning in the accident oil pool.
Description
Technical Field
The invention belongs to the technical field of monitoring of the water level of an accident oil pool of a transformer substation, and particularly relates to a monitoring and early warning method and system for the water level of the accident oil pool of the transformer substation.
Background
In the main electrical equipment of a transformer substation, an oil-immersed power transformer still has wide application because of the characteristics of economic investment, simple and convenient maintenance, low environmental requirement in operation and the like, wherein transformer oil plays roles of insulation, cooling and heat dissipation. When oil is sprayed in case of transformer accident, a large amount of transformer oil splashes out of the transformer in a short time and is discharged to the periphery. If no special protection measures are adopted, the pollution can be caused to the internal and peripheral environments of the transformer substation, and the fire disaster is easily caused after the accident oil injection. Therefore, no matter from the aspects of environmental protection or fire safety, the part of accident oil injection is required to be safely and orderly led to a special accident oil pool, so that the accident oil injection is isolated from an external fire source, cooled and stored, separated, recovered and treated in the future, and is utilized as much as possible. The accident oil pool can smoothly collect and store oil spills when an accident happens, so that the leaked oil is prevented from causing a fire or explosion, and the personal safety of substation equipment and workers is protected.
Under normal conditions, the accident oil pool only can store a small amount of rainwater. However, when the rainy season comes or severe weather such as typhoon and heavy rain occurs, rainwater can gradually seep into the accident oil pool, so that a large amount of accumulated water is stored in the accident oil pool. If oil-filled electrical equipment damages this moment, and a large amount of ponding can't discharge fast in the pond, will lead to equipment oil can't in time get into the accident oil pond and take place oil and leak the accident to cause the conflagration easily, and then bring the potential safety hazard for electrical equipment's operation. Therefore, the water level in the accident oil pool needs to be accurately monitored, and early warning is timely carried out.
The existing transformer substation water level monitoring system mostly uses various sensors to monitor the water level of an accident oil pool, and the system is mainly divided into a contact type system and a non-contact type system. Among them, the detection methods using the float type and press-set type contact sensors are easily affected by impurities in water, so that the detection error is large, and the detection methods using the non-contact sensors such as the laser method, the radar method, the ultrasonic method and the like are too high in cost and the equipment is easily damaged. For example, the patent with the application number of 201921071376.6 measures the distance of the water surface of the accumulated water in the accident oil pool through the ultrasonic sensor, transmits the measured value to the central processing unit to be compared with the preset distance value, and sends out early warning information after the measured value exceeds the preset value for a certain number of times. The water level monitoring mode based on machine vision can improve the monitoring accuracy to a certain extent, but the current monitoring mode often has hysteresis, and when typhoon, torrential rain weather come and go too late to react, dangerous water level early warning can not be carried out in time, still have the potential safety hazard.
Disclosure of Invention
In view of this, the invention provides a method and a system for monitoring and early warning the water level of an accident oil pool of a transformer substation, which are used for solving the problem of hysteresis existing in the existing accident oil pool water level monitoring mode.
The invention provides a monitoring and early warning method for the water level of an accident oil pool of a transformer substation, which comprises the following steps:
continuously acquiring a water gauge image in the accident oil pool through a camera, and identifying the water gauge image in the accident oil pool to obtain a real-time water level reading in the accident oil pool;
acquiring historical water level data of each monitoring point in the peripheral area of the current transformer substation and historical water level data in the accident oil pool, and establishing a water level change prediction model according to the historical water level data of each monitoring point and the historical water level data in the accident oil pool;
acquiring real-time water level data of each current monitoring point, inputting a water level change prediction model, and predicting and outputting the water level change condition in the accident oil pool;
and carrying out water level monitoring and early warning according to the predicted water level change condition in the accident oil pool.
On the basis of above technical scheme, preferably, to the water gauge image in the accident oil sump discernment, the real-time water level reading in discernment accident oil sump specifically includes:
preprocessing the water gauge image;
judging the quality of the water gauge image, and adjusting the brightness of the water gauge image under low illumination to obtain a high-quality water gauge image;
and extracting a water gauge area of the high-quality water gauge image, separating the scale mark from the number of the water gauge area, identifying the number obtained by separation through a pre-trained first convolution neural network model, and calculating the water level reading.
On the basis of the technical scheme, preferably, the method for judging the quality of the water gauge image and adjusting the brightness of the water gauge image under low illuminance comprises the following steps:
calculating the average value on the image gray level histogram, and if the lightness of the image is smaller than the image component gray level average value, determining the image as a water gauge image under low illuminance;
preparing a water gauge image composition data set with different light and shade in advance;
building a KinD network model, inputting the water gauge images in the data set into the KinD network model for training to obtain a trained shading adjustment model;
and adjusting the brightness of the water gauge image under low illumination through the trained brightness adjusting model, and outputting a high-quality water gauge image.
On the basis of the technical scheme, preferably, in the process of identifying the water gauge image in the accident oil pool, the minimum water level reading identification error is taken as a target, the super-parameter of the first convolution neural network and the illuminance adjusting parameter alpha of the KinD network model are optimized by adopting a gradient-based optimization algorithm, the optimized adjusting parameter alpha is input into the KinD network model to be subjected to brightness adjustment, the optimized super-parameter is input into the first convolution neural network, and the real-time water level reading is identified.
On the basis of the above technical solution, preferably, the optimization of the hyper-parameter of the first convolution neural network and the illuminance adjusting parameter α of the KinD network model by using a gradient-based optimization algorithm with the goal of minimizing the water level reading recognition error specifically includes:
respectively setting the value ranges of the hyper-parameter of the first convolution neural network and the illuminance adjusting parameter alpha of the KinD network model, and forming a D-dimensional search space by the value ranges of the hyper-parameter and the adjusting parameter alpha; initializing basic parameters of a gradient-based optimization algorithm, and randomly initializing each individual position in a D-dimensional search space;
establishing a fitness function by taking the minimum water level reading recognition error as a target, calculating the fitness value of each individual through the fitness function, and determining the current optimal vector and the worst vector;
establishing a search direction and a gradient search rule based on the current optimal vector and the worst vector, introducing a vector merging operator of a vector weighted average algorithm, and updating the current individual position based on the search direction, the gradient search rule and the vector merging operator;
updating the current individual position according to the local escape operator;
calculating the fitness of the current individual and updating the optimal vector;
and repeating the processes of updating the current individual position and calculating the individual fitness until the maximum iteration times is reached, and outputting an optimal solution.
On the basis of the above technical solution, preferably, the formula for updating the current individual position based on the search direction and gradient search rule and the vector merge operator is as follows:
wherein
The position of the nth vector in the m and m +1 iterations respectively,
to be composed of
The best vector in
Replacement with the current vector
GSR is a gradient search rule of a gradient-based optimization algorithm, and DM is a search direction;
all are random numbers within 0 to 1,
randn is a random number following a standard normal distribution,
，
is a sine function, and the rand is a random number within 0-1.
On the basis of the above technical solution, preferably, the establishing of the water level change prediction model according to the historical water level data of each monitoring point and the historical water level data in the accident oil pool specifically includes:
dividing historical water level data of each monitoring point in the peripheral region of the current transformer substation and historical water level data in the accident oil pool according to preset time intervals, and respectively calculating the historical water level variation of each monitoring station in each time interval and the historical water level variation in the accident oil pool; a time difference exists between the historical water level data of each monitoring point and the historical water level data in the accident oil pool, and the time difference is determined according to the difference between the time point when the historical water level data of each monitoring point changes and the time point when the historical water level data in the accident oil pool changes;
establishing a position relation matrix according to the position of each monitoring point in the peripheral area of the current transformer substation; the rows and the columns of the position relation matrix represent the positions of the monitoring points, and the elements of the corresponding positions in the position relation matrix are Euclidean distances between the positions of the two corresponding monitoring points;
establishing a water level relation matrix according to historical water level variation among monitoring stations in each time interval; the rows and the columns of the water level relation matrix represent historical water level data of the monitoring points, and elements of corresponding positions in the water level relation matrix are difference values of variation quantities of the historical water level data of the two corresponding monitoring points in the same time interval;
training a second convolutional neural network model by taking the product of the position relation matrix and the water level relation matrix as input and the historical water level variable quantity in the accident oil pool as output;
and taking the trained second convolution neural network model as a water level change prediction model.
The second invention discloses a monitoring and early warning system for the water level of an accident oil pool of a transformer substation, which comprises:
real-time water level identification module: the system is used for continuously acquiring water gauge images in the accident oil pool through the camera, and identifying the water gauge images in the accident oil pool to obtain a real-time water level reading in the accident oil pool;
a water level change prediction module: the system comprises a monitoring point acquisition module, a water level change prediction module and a control module, wherein the monitoring point acquisition module is used for acquiring historical water level data of each monitoring point in the peripheral area of the current transformer substation and historical water level data in an accident oil pool, and the water level change prediction module is established according to the historical water level data of each monitoring point and the historical water level data in the accident oil pool; acquiring real-time water level data of each current monitoring point, inputting a water level change prediction model, and predicting and outputting the water level change condition in the accident oil pool;
water level monitoring and early warning module: and the water level monitoring and early warning are carried out according to the predicted water level change condition in the accident oil pool.
In a third aspect of the present invention, an electronic device is disclosed, comprising: at least one processor, at least one memory, a communication interface, and a bus;
the processor, the memory and the communication interface complete mutual communication through the bus;
the memory stores program instructions executable by the processor which are invoked to implement the method according to the first aspect of the invention.
In a fourth aspect of the invention, a computer-readable storage medium is disclosed, which stores computer instructions for causing a computer to implement the method according to the first aspect of the invention.
Compared with the prior art, the invention has the following beneficial effects:
1) According to the method, the real-time water level reading in the accident oil pool is identified, and meanwhile, the water level change condition in the accident oil pool is predicted according to the historical hydrological data around the current transformer substation, so that an accident oil pool water level prediction curve in a period of time in the future is drawn, the real-time water level monitoring can be performed, the early warning of the future water level change condition can be performed, and the foresight of the water level monitoring and early warning in the accident oil pool is improved;
2) The method comprises the steps of judging the quality of a water gauge image before water level identification, constructing a KinD network model to adjust the brightness of the water gauge image under low illumination, optimizing an adjusting parameter alpha of the KinD network model and a hyper-parameter of a first convolution neural network by adopting a gradient-based optimization algorithm, improving the gradient-based optimization algorithm by introducing a vector merging operator of a vector weighted average algorithm, expanding a local search range while enhancing the local exploration capacity, accelerating algorithm convergence, minimizing a water level reading identification error while optimizing to obtain a high-quality water gauge image, and improving the accuracy of water level identification, so that the method has better adaptability to the dim and low-illumination accident oil pool environment and better meets the requirements of the actual environment;
3) According to the invention, the position relation matrix and the water level relation matrix are established according to the historical water level data of each monitoring point and are used for establishing the water level change prediction model, so that the quantitative representation between the water level data of the accident oil pool and the water level data of the monitoring points around the transformer substation can be realized, and the accuracy of the prediction of the water level condition around the transformer substation is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a flow chart of a monitoring and early warning method for the water level of an accident oil pool of a transformer substation.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.
Referring to fig. 1, the invention provides a method for monitoring and early warning the water level of an accident oil pool of a transformer substation, which comprises the following steps:
s1, continuously acquiring a water gauge image in the accident oil pool through a camera, and identifying the water gauge image in the accident oil pool to obtain a real-time water level reading in the accident oil pool. A water level scale, a light supplementing lamp and a camera are installed in the accident oil pool in advance, and the reading on the water gauge image is recognized in a machine vision mode to obtain a real-time water level reading.
The step S1 specifically comprises the following sub-steps:
and S11, preprocessing the water gauge image.
Specifically, the water gauge image is converted from an RGB color space to an HSV color space. And separating the three components H, S and V of the water gauge image converted into the HSV space, performing graying processing respectively, and enabling the water gauge image to only contain brightness information in a graying processing mode, so that the dimension of a stored image matrix is reduced, and the gradient feature of the image is convenient to calculate.
And S12, judging the quality of the water gauge image, and adjusting the brightness of the water gauge image under low illumination to obtain a high-quality water gauge image.
Because the accident oil pool is mostly designed underground and light is not good, high-quality images cannot be obtained in real time even if a light supplement lamp is additionally arranged, and low-illumination images are not only dark but also accompanied with noise and color distortion, so that the brightness is required to be adjusted and adjusted. Specifically, the brightness value V component is used to perform image brightness evaluation on the acquired water gauge image. And calculating the average value on the image gray level histogram, and if the brightness of the image is smaller than the image component gray level average value, determining that the image is a water gauge image under low illumination, and needing to adjust the brightness.
The invention uses a KinD (bundling the Darkness) network model for shading adjustment. Preparing water gauge image samples shot under different illumination intensities in advance to form a data set, constructing a KinD network model, inputting the water gauge image samples in the data set into the KinD network model for training, and obtaining a trained shading adjustment model.
The KinD network model adopted by the invention specifically comprises the following steps: the device comprises an input layer, a feature extraction layer, a layer decomposition layer, a feature fusion layer, a reflectivity recovery layer, an illumination regulation layer and an output layer.
The input layer is used for inputting water gauge images with different shades.
The feature extraction layer performs multi-scale decomposition on the low-light image from multi-scale features of the image to obtain multi-scale representation of the original low-light image so as to fully mine multi-scale feature information of the low-light image. The multi-scale characteristic information of the image can be extracted by constructing a Gaussian pyramid.
The layer decomposition layer is used for converting the feature information of the feature extraction layer into a reflection map and an illumination map.
The characteristic fusion layer is used for further extracting and fusing the characteristics of the multi-scale reflection map and the illumination map through a residual error network so as to fully utilize complementary characteristic information between the reflection map and the illumination map and remove redundant characteristic information.
And the reflectivity recovery layer is used for recovering the reflectivity of the image after the characteristic fusion.
And the illumination adjusting layer is used for adjusting illumination of the image after the characteristic fusion to obtain a final image after the brightness is adjusted.
The output layer is used for outputting the enhanced final image.
And adjusting the brightness of the water gauge image under low illumination through the trained brightness adjusting model, and outputting a high-quality water gauge image.
When the KinD network model is trained, the input of the KinD network model is a normal illumination image and a corresponding low illumination image, the KinD network model sequentially passes through the feature extraction layer, the layer decomposition layer and the feature fusion layer, wherein the layer decomposition layer has a first loss, the illumination adjusting layer has a second loss, the reflectivity recovery layer has a third loss, and the whole loss is the sum of the three.
And S13, extracting a water gauge region of the high-quality water gauge image, separating scale marks from numbers of the water gauge region, identifying the numbers obtained by separation through a pre-trained first convolution neural network, and calculating a water level reading.
Specifically, when steps S12 and S13 are executed, since the KinD network model is alternately trained through the pair of images photographed under different exposure conditions in different scenes, the final brightness adjustment ratio is determined by an adjustment parameter α, and during testing, actual adjustment can be performed by specifying a specific α value, so as to achieve a demand condition. However, water and oil in the accident oil pool are both easily-reflective substances, and may also include other impurities which affect the light and shade adjustment or water level identification effect, and in addition, the problems of water body refraction and the like may be caused if the illumination adjustment is not accurate, and the problems of wrong water gauge region extraction, too dark digital region or overexposure on the water gauge and the like may be caused. Therefore, the water level reading identification accuracy may be reduced if the dimming is performed by a conventional averaging method or by an artificially specified adjustment parameter α. When the water level reading is identified through the first convolutional neural network, the super-parameter setting of the first convolutional neural network also has great influence on the water level reading identification effect, and the water level reading identification effect can not be ensured under the comprehensive influence of the super-parameter and the adjusting parameter alpha. In order to solve the problem, in the process of executing the steps S12 and S13, the present invention optimizes the hyper-parameter and the adjustment parameter α of the first convolutional neural network by using a Gradient-based optimization algorithm (GBO) with the goal of minimizing the water level reading recognition error, so as to obtain an optimal parameter combination, inputs the optimized adjustment parameter α into the KinD network model trained in the step S12 for adjusting the darkness, inputs the optimized hyper-parameter into the first convolutional neural network, and recognizes the real-time water level reading through the step S13.
The optimization of the hyper-parameters of the first convolution neural network and the illuminance adjusting parameter alpha of the KinD network model by adopting a gradient-based optimization algorithm specifically comprises the following steps:
1) Respectively setting the value ranges of the hyper-parameter of the first convolution neural network and the illuminance adjusting parameter alpha of the KinD network model, and forming a D-dimensional search space by the value ranges of the hyper-parameter and the adjusting parameter alpha; basic parameters of a gradient-based optimization algorithm are initialized, and individual positions are initialized randomly in a D-dimensional search space.
2) And establishing a fitness function by taking the minimum water level reading identification error as a target, calculating the fitness value of each individual through the fitness function, and determining the current optimal vector and the worst vector.
3) Generating two random numbers p within 0 to 1 1 、p 2 Establishing a search direction and a gradient search rule based on the current optimal vector and the worst vector, introducing a vector merging operator of a vector weighted average algorithm to update the current individual position, wherein a specific formula for updating the current individual position based on the search direction, the gradient search rule and the vector merging operator is as follows:
wherein
The position of the nth vector in the m and m +1 iterations respectively,
and
all are random numbers within 0 to 1,
randn is a random number following a standard normal distribution,
，
is a sine function, and the rand is a random number within 0-1. GSR and DM are respectively a gradient search rule and a search direction of a gradient-based optimization algorithm, and the expression is as follows:
4) Updating the current individual position according to a local escape operator LEO;
5) Calculating the fitness of the current individual and updating the optimal vector;
6) And (5) repeating the steps 3 to 5 until the maximum iteration number is reached, and outputting an optimal solution.
In the process of optimizing the hyperparameter of the first convolutional neural network and the illuminance adjusting parameter alpha of the KinD network model by the gradient-based optimization algorithm, in the process of updating the individual position by the conventional gradient search rule, the gradient-based optimization algorithm is improved by introducing a vector merging operator of a vector weighted average algorithm, a new and better vector can be provided by using the vector merging operator, the improved position updating mode updates the current individual position on the basis of the search direction, the gradient search rule and the vector merging operator, and the position updating is carried out by selecting the position updating carried out by using the original search direction and the gradient search rule or carrying out the vector merging and the position updating or keeping the original position still by using the vector merging operator according to different random probabilities.
The illuminance adjusting parameter alpha in the step S12 and the hyper-parameter of the first neural network in the step S13 are optimized as a whole, the obtained optimal solution can enable the illuminance adjusting effect and the water level reading identification accuracy to be optimal, and the influence of ambient illumination in the accident oil pool on water level identification is reduced.
S2, historical water level data of all monitoring points in the peripheral area of the current transformer substation and historical water level data in the accident oil pool are obtained, and a water level change prediction model is established according to the historical water level data of all the monitoring points and the historical water level data in the accident oil pool.
Monitoring points are arranged at a plurality of key water catchments in the peripheral area of the current transformer substation in advance, rainfall is brought by severe weather such as typhoon and rainstorm and is finally reflected on the water level change of each monitoring point, and the rainfall can be considered to be the same in the peripheral area of the transformer substation, so that historical water level data of each monitoring point in a continuous time period are analyzed, and a water level change prediction model is established by combining the water level change in an accident oil pool in the corresponding continuous time period.
The step S2 specifically comprises the following sub-steps:
s21, dividing historical water level data of all monitoring points in the peripheral area of the current transformer substation and historical water level data in the accident oil pool according to preset time intervals, and respectively calculating the historical water level variation of all monitoring points in each time interval and the historical water level variation in the accident oil pool.
The historical water level data of each monitoring point and the historical water level data in the accident oil pool are corresponding in time, but because a water flow gathering process exists after rainfall, a time difference exists between the historical water level data of each monitoring point and the historical water level data in the accident oil pool, and the time difference is defined as a difference value between the time point when the historical water level data of each monitoring point changes and the time point when the historical water level data in the accident oil pool changes. The time point when the historical water level data of each monitoring point changes can be the time point when the water level of any one or more monitoring points changes or the time point when the water level of one or more monitoring points at a specified position changes.
In addition, different time interval scales can be set, the step S21 is respectively executed by taking the different time interval scales as preset time intervals, such as 5min, 10min, 20min, 30min, 60min and the like, so that multiple groups of sampling data exist under the different time interval scales, training data can be enriched, and water level change analysis under the different time interval scales can be conveniently performed.
S22, establishing a position relation matrix A according to the position of each monitoring point in the peripheral region of the current transformer substation. The rows and columns of the position relation matrix represent the positions of the monitoring points, and the elements of the corresponding positions in the position relation matrix are Euclidean distances d between the positions of two corresponding monitoring points i and j ij The value of the element on the diagonal is 0, i.e. the element A [ i, j ] of the ith row and jth column of the matrix A]=d ij 。
And S23, establishing a water level relation matrix B according to historical water level variation among all monitoring stations in each time interval. The rows and columns of the water level relation matrix represent historical water level data of monitoring points, and corresponding positions in the water level relation matrixThe set element is the difference value of the variation of the historical water level data of two monitoring points i and j corresponding to the same time interval (Δ t), namely the element B [ i, j of the ith row and the jth column of the matrix B]=∆L i -∆L j Therein is Δ L i The change amount of the historical water level data of two monitoring points i within the time interval Δ L j Is the change of the historical water level data at two monitoring points j in the time interval Δ t.
And S24, training a second convolutional neural network model by taking the product of the position relation matrix and the water level relation matrix as input and the historical water level variable quantity in the accident oil pool as output.
And S25, taking the trained second convolutional neural network model as a water level change prediction model.
Because the water level change in the accident oil pool is directly related to the positions of all the monitoring points on the periphery and the water level change of all the monitoring points, a position relation matrix and a water level relation matrix are introduced when the water level change in the accident oil pool is predicted, the position relation matrix of all the monitoring points in the peripheral area of the current transformer substation reflects the spatial relation among all the monitoring points, the water level relation matrix reflects the relation of the water level among all the monitoring points changing along with time, and a second convolutional neural network model can be trained by taking the product of the position relation matrix and the water level relation matrix as input and taking the historical water level change quantity in the accident oil pool as output and can visually reflect the time-space change relation between the water level change in the accident oil pool and the positions and monitoring data of all the monitoring points on the periphery, so that the water level change quantity in the accident oil pool can be predicted more accurately.
And S3, acquiring real-time water level data of each current monitoring point, inputting a water level change prediction model, and predicting and outputting the water level change condition in the accident oil pool.
And acquiring real-time water level data of each current monitoring point, inputting a water level change prediction model, and predicting and outputting water level change situation data in the accident oil pool in a future period of time.
And S4, carrying out water level monitoring and early warning according to the predicted water level change condition in the accident oil pool.
Due to the time difference between the water level data of the monitoring point and the water level data in the accident oil pool, the predicted water level change condition in the accident oil pool and the real-time water level reading in the accident oil pool are superposed to obtain a predicted value of the water level in the accident oil pool, and therefore a water level prediction curve of the accident oil pool is drawn in real time.
In addition, because different time interval scales are set, water level change data under different time intervals can be obtained, namely a water level prediction curve in a period of time in the future can be predicted, so that the warning water level can be set, the time for the water level in the accident oil pool to reach the warning water level is predicted according to the water level prediction curve of the accident oil pool, and water level monitoring and early warning are carried out. And based on the predicted time when the water level in the accident oil pool reaches the warning water level, a drainage system of the accident oil pool can be started in advance, and the leakage of the accident oil pool is avoided.
Corresponding to the above method embodiment, the invention also provides a system for monitoring the water level of the accident oil pool of the transformer substation, wherein the system comprises:
real-time water level identification module: the system is used for continuously acquiring water gauge images in the accident oil pool through the camera, and identifying the water gauge images in the accident oil pool to obtain a real-time water level reading in the accident oil pool;
a water level change prediction module: the system comprises a monitoring station, a water level change prediction model, a data acquisition module, a data analysis module and a data analysis module, wherein the monitoring station is used for acquiring historical water level data of each monitoring point in the peripheral area of the current transformer substation and historical water level data in an accident oil pool, and the water level change prediction model is established according to the historical water level data of each monitoring point and the historical water level data in the accident oil pool; acquiring real-time water level data of each current monitoring point, inputting a water level change prediction model, and predicting and outputting the water level change condition in the accident oil pool;
water level monitoring and early warning module: and the water level monitoring and early warning are carried out according to the predicted water level change condition in the accident oil pool.
And predicting the time for the water level in the accident oil pool to reach the warning water level by the prediction curve, and monitoring and early warning the water level.
The above system embodiments and method embodiments are in one-to-one correspondence, and please refer to the method embodiments for brief description of the system embodiments.
The invention also discloses an electronic device, comprising: at least one processor, at least one memory, a communication interface, and a bus; the processor, the memory and the communication interface complete mutual communication through the bus; the memory stores program instructions executable by the processor, which invokes the program instructions to implement the methods of the invention described above.
The invention also discloses a computer readable storage medium which stores computer instructions, and the computer instructions enable the computer to realize all or part of the steps of the method of the embodiment of the invention. The storage medium includes: various media capable of storing program codes, such as a U disk, a removable hard disk, a read-only memory ROM, a random access memory RAM, a magnetic disk, or an optical disk.
The above-described system embodiments are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, i.e. may be distributed over a plurality of network units. Without creative labor, a person skilled in the art can select some or all of the modules according to actual needs to achieve the purpose of the solution of the embodiment.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and should not be taken as limiting the scope of the present invention, which is intended to cover any modifications, equivalents, improvements, etc. within the spirit and scope of the present invention.
Claims (10)
1. A monitoring and early warning method for the water level of an accident oil pool of a transformer substation is characterized by comprising the following steps:
acquiring a water gauge image in the accident oil pool through a camera, and identifying the water gauge image in the accident oil pool to obtain a real-time water level reading in the accident oil pool;
acquiring historical water level data of each monitoring point in the peripheral area of the current transformer substation and historical water level data in an accident oil pool, and establishing a water level change prediction model according to the historical water level data of each monitoring point and the historical water level data in the accident oil pool;
acquiring real-time water level data of each current monitoring point, inputting a water level change prediction model, and predicting and outputting the water level change condition in the accident oil pool;
and carrying out water level monitoring and early warning according to the predicted water level change condition in the accident oil pool.
2. The transformer substation accident oil pool water level monitoring and early warning method according to claim 1, wherein the identifying of the water gauge image in the accident oil pool specifically comprises:
preprocessing the water gauge image;
judging the quality of the water gauge image, and adjusting the brightness of the water gauge image under low illumination to obtain a high-quality water gauge image;
and extracting a water gauge area of the high-quality water gauge image, separating scale marks from numbers of the water gauge area, identifying the separated numbers through a pre-trained first convolution neural network model, and calculating a water level reading.
3. The method for monitoring the water level of the accident oil pool of the transformer substation according to claim 2, wherein the step of judging the quality of the water gauge image and adjusting the brightness of the water gauge image under low illumination to obtain the high-quality water gauge image specifically comprises the following steps:
calculating the mean value of the gray level histogram of the water gauge image, and if the lightness of the image is smaller than the mean value of the gray level components of the image, determining the water gauge image under low illuminance;
preparing a water gauge image composition data set with different light and shade in advance;
building a KinD network model, inputting the water gauge images in the data set into the KinD network model for training to obtain a trained shading adjustment model;
and adjusting the brightness of the water gauge image under low illumination through the trained brightness adjusting model, and outputting a high-quality water gauge image.
4. The substation accident oil pool water level monitoring method according to claim 3, wherein in the process of identifying the water gauge image in the accident oil pool, aiming at the minimum water level reading identification error, the super-parameter of the first convolution neural network and the illuminance adjusting parameter α of the KinD network model are optimized by adopting a gradient-based optimization algorithm, the optimized adjusting parameter α is input into the KinD network model for shading adjustment, the optimized super-parameter is input into the first convolution neural network, and real-time water level reading is identified.
5. The substation accident oil pool water level monitoring method according to claim 4, wherein the optimizing the hyper-parameter of the first convolution neural network and the illuminance adjustment parameter α of the KinD network model by using a gradient-based optimization algorithm with the water level reading identification error as a target specifically comprises:
respectively setting the value ranges of the hyper-parameter of the first convolution neural network and the illuminance adjusting parameter alpha of the KinD network model, and forming a D-dimensional search space by the value ranges of the hyper-parameter and the adjusting parameter alpha; initializing basic parameters of a gradient-based optimization algorithm, and randomly initializing each individual position in a D-dimensional search space;
establishing a fitness function by taking the minimum water level reading recognition error as a target, calculating the fitness value of each individual through the fitness function, and determining the current optimal vector and the worst vector;
establishing a search direction and a gradient search rule based on the current optimal vector and the worst vector, introducing a vector merging operator of a vector weighted average algorithm, and updating the current individual position based on the search direction, the gradient search rule and the vector merging operator;
updating the current individual position according to the local escape operator;
calculating the fitness of the current individual and updating the optimal vector;
and repeating the processes of updating the current individual position and calculating the individual fitness until the maximum iteration times is reached, and outputting an optimal solution.
6. The substation accident oil pool water level monitoring and early warning method according to claim 5, wherein the formula for updating the current individual position based on the search direction and gradient search rule and the vector merging operator is as follows:
wherein
The position of the nth vector in the m and m +1 iterations respectively,
to be composed of
The best vector in
Replacement with the Current vector
GSR is a gradient search rule of a gradient-based optimization algorithm, and DM is a search direction;
all are random numbers within 0 to 1,
randn is compliance standardA random number that is normally distributed is selected,
，
is a sine function, and the rand is a random number within 0 to 1.
7. The transformer substation accident oil pool water level monitoring and early warning method according to claim 1, wherein the establishing of the water level change prediction model according to the historical water level data of each monitoring point and the historical water level data in the accident oil pool specifically comprises:
dividing historical water level data of each monitoring point in the peripheral area of the current transformer substation and historical water level data in an accident oil pool according to preset time intervals, and respectively calculating the historical water level variation of each monitoring station in each time interval and the historical water level variation in the accident oil pool; a time difference exists between the historical water level data of each monitoring point and the historical water level data in the accident oil pool, and the time difference is determined according to the difference between the time point when the historical water level data of each monitoring point changes and the time point when the historical water level data in the accident oil pool changes;
establishing a position relation matrix according to the position of each monitoring point in the peripheral area of the current transformer substation; the rows and the columns of the position relation matrix represent the positions of the monitoring points, and the elements of the corresponding positions in the position relation matrix are Euclidean distances between the positions of the two corresponding monitoring points;
establishing a water level relation matrix according to historical water level variation among monitoring stations in each time interval; the rows and the columns of the water level relation matrix represent historical water level data of the monitoring points, and elements of corresponding positions in the water level relation matrix are difference values of variation quantities of the historical water level data of the two corresponding monitoring points in the same time interval;
training a second convolutional neural network model by taking the product of the position relation matrix and the water level relation matrix as input and the historical water level variable quantity in the accident oil pool as output;
and taking the trained second convolution neural network model as a water level change prediction model.
8. The utility model provides a transformer substation accident oil sump water level monitoring early warning system which characterized in that, the system includes:
real-time water level identification module: the system comprises a camera, a water gauge, a water level meter and a controller, wherein the camera is used for continuously acquiring water gauge images in an accident oil pool, and identifying the water gauge images in the accident oil pool to obtain a real-time water level reading in the accident oil pool;
a water level change prediction module: the system comprises a monitoring point acquisition module, a water level change prediction module and a control module, wherein the monitoring point acquisition module is used for acquiring historical water level data of each monitoring point in the peripheral area of the current transformer substation and historical water level data in an accident oil pool, and the water level change prediction module is established according to the historical water level data of each monitoring point and the historical water level data in the accident oil pool; acquiring real-time water level data of each current monitoring point, inputting a water level change prediction model, and predicting and outputting the water level change condition in the accident oil pool;
water level monitoring and early warning module: and the water level monitoring and early warning are carried out according to the predicted water level change condition in the accident oil pool.
9. An electronic device, comprising: at least one processor, at least one memory, a communication interface, and a bus;
the processor, the memory and the communication interface complete mutual communication through the bus;
the memory stores program instructions executable by the processor, which are invoked by the processor to implement the method of any one of claims 1 to 7.
10. A computer-readable storage medium storing computer instructions for causing a computer to implement the method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211158932.XA CN115240394B (en) | 2022-09-22 | 2022-09-22 | Method and system for monitoring and early warning water level of accident oil pool of transformer substation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211158932.XA CN115240394B (en) | 2022-09-22 | 2022-09-22 | Method and system for monitoring and early warning water level of accident oil pool of transformer substation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115240394A true CN115240394A (en) | 2022-10-25 |
| CN115240394B CN115240394B (en) | 2022-12-20 |
Family
ID=83667487
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211158932.XA Active CN115240394B (en) | 2022-09-22 | 2022-09-22 | Method and system for monitoring and early warning water level of accident oil pool of transformer substation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115240394B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115865992A (en) * | 2023-03-02 | 2023-03-28 | 中国建材检验认证集团湖南有限公司 | Wisdom water conservancy on-line monitoring system |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108596221A (en) * | 2018-04-10 | 2018-09-28 | 江河瑞通（北京）技术有限公司 | The image-recognizing method and equipment of rod reading |
| CN109508630A (en) * | 2018-09-27 | 2019-03-22 | 杭州朗澈科技有限公司 | A method of water gauge water level is identified based on artificial intelligence |
| WO2019224739A1 (en) * | 2018-05-25 | 2019-11-28 | University Of Johannesburg | System and method for real time prediction of water level and hazard level of a dam |
| KR102149495B1 (en) * | 2019-08-19 | 2020-08-28 | 고려대학교 산학협력단 | Optimization apparatus for training conditions of environmental prediction model and operating thereof |
| CN112149648A (en) * | 2020-11-24 | 2020-12-29 | 武汉光谷信息技术股份有限公司 | Water line identification and estimation method and device based on gray variable threshold |
| WO2021007812A1 (en) * | 2019-07-17 | 2021-01-21 | 深圳大学 | Deep neural network hyperparameter optimization method, electronic device and storage medium |
| CN113052239A (en) * | 2021-03-25 | 2021-06-29 | 山东大学 | Image classification method and system of neural network based on gradient direction parameter optimization |
| US20210216913A1 (en) * | 2020-01-14 | 2021-07-15 | Alibaba Group Holding Limited | Prediction method, apparatus, and system for performing an image search |
| CN215524773U (en) * | 2021-09-26 | 2022-01-14 | 仇美 | Water level detection device for hydraulic engineering management |
| CN113971779A (en) * | 2021-10-29 | 2022-01-25 | 中国水利水电科学研究院 | Water gauge automatic reading method based on deep learning |
| CN114091362A (en) * | 2022-01-24 | 2022-02-25 | 江西飞尚科技有限公司 | Water level prediction method, system, storage medium and equipment |
-
2022
- 2022-09-22 CN CN202211158932.XA patent/CN115240394B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108596221A (en) * | 2018-04-10 | 2018-09-28 | 江河瑞通（北京）技术有限公司 | The image-recognizing method and equipment of rod reading |
| WO2019224739A1 (en) * | 2018-05-25 | 2019-11-28 | University Of Johannesburg | System and method for real time prediction of water level and hazard level of a dam |
| CN109508630A (en) * | 2018-09-27 | 2019-03-22 | 杭州朗澈科技有限公司 | A method of water gauge water level is identified based on artificial intelligence |
| WO2021007812A1 (en) * | 2019-07-17 | 2021-01-21 | 深圳大学 | Deep neural network hyperparameter optimization method, electronic device and storage medium |
| KR102149495B1 (en) * | 2019-08-19 | 2020-08-28 | 고려대학교 산학협력단 | Optimization apparatus for training conditions of environmental prediction model and operating thereof |
| US20210216913A1 (en) * | 2020-01-14 | 2021-07-15 | Alibaba Group Holding Limited | Prediction method, apparatus, and system for performing an image search |
| CN112149648A (en) * | 2020-11-24 | 2020-12-29 | 武汉光谷信息技术股份有限公司 | Water line identification and estimation method and device based on gray variable threshold |
| CN113052239A (en) * | 2021-03-25 | 2021-06-29 | 山东大学 | Image classification method and system of neural network based on gradient direction parameter optimization |
| CN215524773U (en) * | 2021-09-26 | 2022-01-14 | 仇美 | Water level detection device for hydraulic engineering management |
| CN113971779A (en) * | 2021-10-29 | 2022-01-25 | 中国水利水电科学研究院 | Water gauge automatic reading method based on deep learning |
| CN114091362A (en) * | 2022-01-24 | 2022-02-25 | 江西飞尚科技有限公司 | Water level prediction method, system, storage medium and equipment |
Non-Patent Citations (1)
| Title |
|---|
| 张超: "基于图像的水位标尺检测技术研究", 《中国优秀硕士学位论文全文数据库（电子期刊）》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115865992A (en) * | 2023-03-02 | 2023-03-28 | 中国建材检验认证集团湖南有限公司 | Wisdom water conservancy on-line monitoring system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115240394B (en) | 2022-12-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113449680B (en) | Knowledge distillation-based multimode small target detection method | |
| CN106228125B (en) | Method for detecting lane lines based on integrated study cascade classifier | |
| CN111583198A (en) | Insulator picture defect detection method combining FasterR-CNN + ResNet101+ FPN | |
| CN113052876B (en) | Video relay tracking method and system based on deep learning | |
| CN107808133A (en) | Oil-gas pipeline safety monitoring method, system and software memory based on unmanned plane line walking | |
| CN110070530A (en) | A kind of powerline ice-covering detection method based on deep neural network | |
| CN111553201A (en) | Traffic light detection method based on YOLOv3 optimization algorithm | |
| CN111209832B (en) | Auxiliary obstacle avoidance training method, equipment and medium for substation inspection robot | |
| CN115240394B (en) | Method and system for monitoring and early warning water level of accident oil pool of transformer substation | |
| CN108229524A (en) | A kind of chimney and condensing tower detection method based on remote sensing images | |
| CN112084712A (en) | Flood submerging range dynamic simulation method fusing active and passive microwave remote sensing information | |
| CN107784290A (en) | A kind of method for the bit occupancy digital image recognition that stops | |
| CN115240075B (en) | Construction and training method of electric power vision multi-granularity pre-training large model | |
| CN114743128A (en) | Multimode northeast tiger re-identification method and device based on heterogeneous neural network | |
| CN113536944A (en) | Distribution line inspection data identification and analysis method based on image identification | |
| CN113963314A (en) | Rainfall monitoring method and device, computer equipment and storage medium | |
| CN115797411B (en) | Method for online recognition of hydropower station cable bridge deformation by utilizing machine vision | |
| CN113052103A (en) | Electrical equipment defect detection method and device based on neural network | |
| CN116884192A (en) | Power production operation risk early warning method, system and equipment | |
| CN113591773B (en) | Distribution room object detection method, device and equipment based on convolutional neural network | |
| CN115619719A (en) | Pine wood nematode infected wood detection method based on improved Yolo v3 network model | |
| CN115578691A (en) | Video anomaly detection method, system and equipment based on scene target | |
| CN112818836B (en) | Method and system for detecting personnel target of transformer substation scene | |
| CN114627421A (en) | Multi-memory video anomaly detection and positioning method and system based on scene target | |
| CN113657351A (en) | High-resolution remote sensing image forest and grass change detection device and method based on deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |