Method for monitoring operation state of power station

Abstract

The invention discloses a power station operation state monitoring method, and relates to the technical field of power stations, and the method comprises the steps: deploying an equipment operation state sensor and an indoor environment state sensor in a power station, and obtaining the sensing data in real time; deploying a local memory in the power station, storing data of each sensor, and presetting each alarm threshold value; deploying an edge data processing system in the power station, analyzing the sensing data, predicting hidden dangers, and warning the sensing data exceeding an alarm threshold and the hidden dangers to a far-end server deployed in a central monitoring room; and the far-end server pushes the sensing data exceeding the alarm threshold value and the hidden danger to the user terminal. The safety and the operation efficiency of the power station are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power stations, and in particular to a method for monitoring the operating status of a power station. Background Art

[0002] With the development of the power industry, power stations, as important places for power production, their safe and stable operation is crucial to the reliability of the power grid.

[0003] The widely implemented power station monitoring currently relies on personnel to regularly check the power station sensor data. However, the existing research on information and automation monitoring technology focuses on the monitoring of equipment status, without considering the monitoring of the indoor environment of the power station. At the same time, it focuses on the research of threshold trigger alarm mechanism, while ignoring the importance of prevention. Therefore, it is particularly important to design a method that can monitor the operation status of the power station in real time and prevent it before it happens. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a method for monitoring the operating status of a power station, which solves the problems of incomplete monitoring and untimely early warning in the existing monitoring method for the operating status of a power station.

[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0006] A method for monitoring the operating status of a power station comprises the following steps:

[0007] S1. Deploy equipment operation status sensors and indoor environment status sensors in the power station to obtain sensor data in real time;

[0008] S2. Deploy local storage in the power station to store the data of each sensor and preset each alarm threshold;

[0009] S3. Deploy an edge data processing system in the power station to analyze various sensor data, predict hidden dangers, and alert the sensor data that has exceeded the alarm threshold and hidden dangers to the remote server deployed in the central monitoring room;

[0010] S4. The remote server pushes the sensor data that has exceeded the alarm threshold and the hidden dangers to the user terminal.

[0011] Furthermore, the edge data processing system includes:

[0012] Trend analysis module, used to predict the trend value of each sensor data through a high-order smoothing algorithm;

[0013] A threshold judgment module is used to judge whether each sensor data and its trend value exceed each alarm threshold;

[0014] The fault prediction module is used to analyze various sensor data and predict equipment failures through pre-trained classifiers.

[0015] The hidden dangers described in steps S3 and S4 refer to the sensor data trend value exceeding the alarm threshold and the equipment failure predicted by the fault prediction module.

[0016] Furthermore, the equipment operation status sensor includes: a voltage sensor, an electric power sensor, a frequency meter, an electric field strength sensor and a first temperature sensor.

[0017] The indoor environment state sensor includes: a second temperature sensor and a humidity sensor.

[0018] Furthermore, the sensing data include: AC voltage effective value, electric power value, AC frequency, electric field vector three-dimensional matrix, equipment temperature value, ambient temperature value and ambient humidity value;

[0019] The alarm thresholds include: undervoltage threshold, overvoltage threshold, power upper limit, frequency jitter limit, device temperature upper limit, ambient temperature lower limit, ambient temperature upper limit and ambient humidity upper limit.

[0020] Further, the trend analysis module predicts the device temperature trend value, the ambient temperature trend value and the ambient humidity trend value through a high-order smoothing algorithm according to the device temperature value, the ambient temperature value and the ambient humidity value;

[0021] The expression of the high-order smoothing algorithm is:

[0022] ,

[0023] Among them, y(t+1) is the predicted trend value at time t+1, y(tk) is the trend value at time tk predicted in the past, x(tk) is the true value at time tk, and α k is the k-th order true value weighting coefficient, β k is the kth order error weighting coefficient, and N is the order.

[0024] Furthermore, the fault prediction module analyzes the three-dimensional matrix of electric field vectors through a pre-trained classifier to predict partial discharge faults.

[0025] Furthermore, the pre-trained classifier includes:

[0026] The modulo operation layer is used to perform a modulo operation on each element in the three-dimensional matrix of the electric field vector to obtain a three-dimensional matrix of the electric field scalar;

[0027] The differential operation layer is used to obtain the difference between each element in the current electric field scalar three-dimensional matrix and the corresponding element in the electric field scalar three-dimensional matrix at the previous moment, and obtain the electric field difference component three-dimensional matrix;

[0028] Reshape layer, used to rearrange the three-dimensional matrix of electric field difference components into a sequence of electric field difference components;

[0029] At least one fully connected layer, used for processing the electric field difference component sequence through a fully connected neural network operation to obtain an electric field feature vector;

[0030] The output layer is used to analyze the electric field feature vector and predict whether a partial discharge fault occurs.

[0031] Furthermore, the output layer expression of the pre-trained classifier is:

[0032] ,

[0033] Among them, f out is the partial discharge fault prediction value obtained by the output layer of the classifier. The data type is a 1-bit unsigned integer. Its value 1 represents that a partial discharge fault will occur, and its value 0 represents that a partial discharge fault will not occur. tanh(∙) is the hyperbolic tangent function. w i is the i-th weight coefficient of the output layer; f in (i) is the i-th element of the electric field eigenvector; b ias is the output layer bias coefficient; M is the length of the electric field eigenvector.

[0034] Furthermore, the pre-trained classifier is pre-trained by using a plurality of three-dimensional matrices of electric field vectors labeled with whether a partial discharge fault occurs as samples, and the loss function used is:

[0035] ,

[0036] Wherein, r is the label of whether a partial discharge fault occurs in the sample, and the data type is a 1-bit unsigned integer, wherein a value of 1 represents that a partial discharge fault will occur, and a value of 0 represents that a partial discharge fault will not occur; f out is the partial discharge fault prediction value obtained by the output layer of the classifier. The data type is a 1-bit unsigned integer. Its value 1 represents that a partial discharge fault will occur, and its value 0 represents that a partial discharge fault will not occur. L is the loss function. log2(∙) is the logarithmic function with base 2.

[0037] The beneficial effects of the present invention are:

[0038] (1) The present invention realizes real-time monitoring of the operating status and environmental status of the equipment in the power station. Through automated data collection, storage and processing, it reduces human operation errors and improves the accuracy of monitoring. It not only has a threshold trigger alarm mechanism, but also can predict alarm conditions and take precautions. It also has the functions of remote alarm push and remote storage to prevent data loss due to damage to the power station. It effectively improves the safety and operating efficiency of the power station.

[0039] (2) A variety of data monitoring systems including voltage, current, frequency, temperature, humidity and electric field are set up to cover the power station operation status in three dimensions, especially to be able to promptly detect and transmit warnings of undervoltage, overvoltage, power limit and frequency jitter limit to avoid danger.

[0040] (3) For data with time-domain continuity and accumulation such as equipment temperature, ambient temperature and ambient humidity, the present invention sets up a high-order smoothing algorithm based on smoothing iteration to predict development trends and provide early warning.

[0041] (4) Partial discharge is a discharge phenomenon in which only part of the insulation system discharges under the action of an electric field, but no through-discharge channel is formed. It will destroy the molecular structure of the equipment material, cause overheating, corrosion, sludge accumulation, and accelerate the aging of the insulation device, causing accidents. The present invention measures the electric field vectors at various locations in the three-dimensional space of the power station through an electric field strength sensor, and sets a classifier. Through modulo operation, the amplitude of the vector is extracted, and the amplitude change information is obtained through differential operation. The three-dimensional matrix is ​​then rearranged into a sequence, and finally, through full connection operation and binary logic output, the prediction result of partial discharge is obtained, thereby more comprehensively monitoring the safety of the power station.

[0042] (5) The output layer of the classifier uses the hyperbolic tangent function as the activation function. Compared with the sigmoid function, it is less likely to experience gradient vanishing and is easier to train the parameters of the classifier. The present invention translates and scales the result of the hyperbolic tangent function to the interval of 0 and 1, and sets the bit width to a 1-bit unsigned integer, so that the value is limited to 0 and 1. Then, using the loss function designed by the present invention, through training with samples labeled with the same rule, it becomes a logical indication of whether a partial discharge fault occurs, and can accurately predict partial discharge faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flow chart of a method for monitoring the operating status of a power station provided by an embodiment of the present invention;

[0044] Figure 2 Schematic diagram of device operation status sensors, indoor environment status sensors, local storage, edge data processing system, remote server and user terminal deployed in an embodiment of the present invention;

[0045] Figure 3 This is a structural diagram of a pre-trained classifier according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The specific implementation modes of the present invention are described below to facilitate those skilled in the art to understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0047] like Figure 1 As shown, in one embodiment of the present invention, a method for monitoring the operating status of a power station includes the following steps:

[0048] S1. Deploy equipment operation status sensors and indoor environment status sensors in the power station to obtain various sensor data in real time.

[0049] Equipment operating status sensors such as Figure 2 As shown, it includes: a voltage sensor, an electric power sensor, a frequency meter, an electric field strength sensor and a first temperature sensor. The indoor environment state sensor includes: a second temperature sensor and a humidity sensor.

[0050] The sensing data include: effective value of AC voltage, electric power value, AC frequency, three-dimensional matrix of electric field vectors formed by electric field vectors at various locations in the three-dimensional space of the power station room, equipment temperature value measured by the first temperature sensor, and ambient temperature and ambient humidity value measured by the second temperature sensor.

[0051] S2. Deploy local storage in the power station to store the data of each sensor and preset each alarm threshold.

[0052] The alarm thresholds include: undervoltage threshold, overvoltage threshold, power upper limit, frequency jitter limit, device temperature upper limit, ambient temperature lower limit, ambient temperature upper limit and ambient humidity upper limit.

[0053] S3. Deploy an edge data processing system in the power station to analyze various sensor data, predict hidden dangers, and alert the sensor data and hidden dangers that have exceeded the alarm threshold to the remote server deployed in the central monitoring room.

[0054] Edge data processing systems, such as Figure 2 Shown include:

[0055] The trend analysis module is used to predict the device temperature trend value, the ambient temperature trend value and the ambient humidity trend value according to the device temperature value, the ambient temperature value and the ambient humidity value through a high-order smoothing algorithm.

[0056] A threshold judgment module is used to judge whether each sensor data and its trend value exceed each alarm threshold;

[0057] The fault prediction module is used to analyze the three-dimensional matrix of electric field vectors through a pre-trained classifier to predict partial discharge faults.

[0058] The hidden dangers in this embodiment refer to the trend value of the sensor data exceeding the alarm threshold and the partial discharge fault predicted by the fault prediction module.

[0059] The expression of the high-order smoothing algorithm used by the trend analysis module is:

[0060] ,

[0061] Among them, y(t+1) is the predicted trend value at time t+1, y(tk) is the trend value at time tk predicted in the past, x(tk) is the true value at time tk, and α k is the k-th order true value weighting coefficient, β k is the kth order error weighting coefficient, and N is the order.

[0062] The present invention is equipped with multiple data monitoring including voltage, current, frequency, temperature, humidity and electric field, which covers the operation status of the power station in three dimensions, and is especially capable of timely detecting and transmitting warnings of undervoltage, overvoltage, power exceeding limit and frequency jitter exceeding limit, so as to avoid danger.

[0063] For data with time-domain continuity and accumulation such as equipment temperature, ambient temperature and ambient humidity, the present invention sets up a high-order smoothing algorithm based on smooth iteration to predict development trends and provide earlier warnings.

[0064] In this embodiment, the initial value y(0) of the predicted trend value at time 0 is set to 0 to start smooth iteration and perform subsequent iterative updates.

[0065] It is worth noting that the true value weighting coefficient of each order must be greater than the error weighting coefficient; the sum of the true value weighting coefficients of each order must be equal to 1.

[0066] like Figure 3 As shown, the pre-trained classifiers used in the fault prediction module include:

[0067] The modulo operation layer is used to perform a modulo operation on each element in the three-dimensional matrix of the electric field vector to obtain a three-dimensional matrix of the electric field scalar;

[0068] The differential operation layer is used to obtain the difference between each element in the current electric field scalar three-dimensional matrix and the corresponding element in the electric field scalar three-dimensional matrix at the previous moment, and obtain the electric field difference component three-dimensional matrix;

[0069] Reshape layer, used to rearrange the three-dimensional matrix of electric field difference components into a sequence of electric field difference components;

[0070] At least one fully connected layer, used to process the electric field difference component sequence through a fully connected neural network operation to obtain an electric field feature vector. This embodiment uses one fully connected layer;

[0071] The output layer is used to analyze the electric field feature vector and predict whether a partial discharge fault occurs.

[0072] The output layer expression is:

[0073] ,

[0074] Among them, f out is the partial discharge fault prediction value obtained by the output layer of the classifier. The data type is a 1-bit unsigned integer. Its value 1 represents that a partial discharge fault will occur, and its value 0 represents that a partial discharge fault will not occur. tanh(∙) is the hyperbolic tangent function. w i is the i-th weight coefficient of the output layer; f in (i) is the i-th element of the electric field eigenvector; b ias is the output layer bias coefficient; M is the length of the electric field eigenvector.

[0075] The so-called pre-training is to perform parameter training for the fully connected layer and the output layer in advance. In this embodiment, multiple three-dimensional matrices of electric field vectors labeled with whether partial discharge faults occur are used as samples for pre-training, and the loss function used is:

[0076] ,

[0077] Wherein, r is the label of whether a partial discharge fault occurs in the sample, and the data type is a 1-bit unsigned integer, wherein a value of 1 represents that a partial discharge fault will occur, and a value of 0 represents that a partial discharge fault will not occur; f out is the partial discharge fault prediction value obtained by the output layer of the classifier. The data type is a 1-bit unsigned integer. Its value 1 represents that a partial discharge fault will occur, and its value 0 represents that a partial discharge fault will not occur. L is the loss function. log2(∙) is the logarithmic function with base 2.

[0078] Partial discharge is a discharge phenomenon in which only part of the insulation system discharges under the action of an electric field without forming a penetrating discharge channel. It will destroy the molecular structure of the equipment material, cause overheating, corrosion, sludge accumulation, and accelerate the aging of the insulation device, causing accidents. The present invention measures the electric field vectors at various locations in the three-dimensional space of the power station through an electric field strength sensor, and sets a classifier. Through modulo operation, the amplitude of the vector is extracted, and the amplitude change information is obtained through differential operation. The three-dimensional matrix is ​​rearranged into a sequence, and finally through full connection operation and binary logic output, the prediction result of partial discharge is obtained, thereby more comprehensively monitoring the safety of the power station.

[0079] The output layer uses the hyperbolic tangent function as the activation function. Compared with the sigmoid function, it is less likely to have a gradient disappearance and is easier to train the parameters of the classifier. The present invention translates and scales the result of the hyperbolic tangent function to the interval of 0 and 1, and sets the bit width to a 1-bit unsigned integer, so that the value is limited to 0 and 1. Then, using the loss function designed by the present invention, through the training of samples with the same rule label, it becomes a logical indication of whether a partial discharge fault occurs, and can accurately predict partial discharge faults.

[0080] S4. The remote server pushes the sensor data that has exceeded the alarm threshold and the hidden dangers to the user terminal.

[0081] In summary, the present invention realizes real-time monitoring of the operating status and environmental status of the equipment in the power station; and through automated data collection, storage and processing, it reduces human operation errors and improves the accuracy of monitoring; it not only has a threshold trigger alarm mechanism, but also can predict alarm conditions and take precautions; and it has the functions of remote alarm push and remote storage to prevent data loss due to damage to the power station. It effectively improves the safety and operation efficiency of the power station.

[0082] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

[0083] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A method for monitoring the operation status of a power station, characterized in that: The following steps are involved: S1. Deploy equipment operation status sensors and indoor environment status sensors in the power station to obtain sensor data in real time; S2. Deploy local storage in the power station to store the data of each sensor and preset each alarm threshold; S3. Deploy an edge data processing system in the power station to analyze various sensor data, predict hidden dangers, and alert the sensor data and hidden dangers that have exceeded the alarm threshold to the remote server deployed in the central monitoring room, wherein the edge data processing system includes: a trend analysis module, which is used to predict the trend value of each sensor data through a high-order smoothing algorithm, a threshold judgment module, which is used to judge whether each sensor data and its trend value exceed each alarm threshold, and a fault prediction module, which is used to analyze each sensor data and predict equipment failure through a pre-trained classifier; S4. The remote server pushes the sensor data that has exceeded the alarm threshold and the hidden dangers to the user terminal.

2. The method for monitoring the operation status of a power station according to claim 1, characterized in that: The equipment operation status sensor includes: a voltage sensor, an electric power sensor, a frequency meter, an electric field strength sensor and a first temperature sensor; The indoor environment state sensor includes: a second temperature sensor and a humidity sensor.

3. The method for monitoring the operation status of a power station according to claim 2, characterized in that: The sensor data include: AC voltage effective value, electric power value, AC frequency, electric field vector three-dimensional matrix, equipment temperature value, ambient temperature value and ambient humidity value; The alarm thresholds include: undervoltage threshold, overvoltage threshold, power upper limit, frequency jitter limit, device temperature upper limit, ambient temperature lower limit, ambient temperature upper limit and ambient humidity upper limit.

4. The method for monitoring the operation status of a power station according to claim 3, characterized in that: The trend analysis module predicts the device temperature trend value, the ambient temperature trend value and the ambient humidity trend value according to the device temperature value, the ambient temperature value and the ambient humidity value through a high-order smoothing algorithm; The expression of the high-order smoothing algorithm is: ， Among them, y(t+1) is the predicted trend value at time t+1, y(tk) is the trend value at time tk predicted in the past, x(tk) is the true value at time tk, and α k is the k-th order true value weighting coefficient, β k is the kth order error weighting coefficient, and N is the order.

5. The method for monitoring the operation status of a power station according to claim 3, characterized in that: The fault prediction module analyzes the three-dimensional matrix of electric field vectors through a pre-trained classifier to predict partial discharge faults.

6. The method for monitoring the operation status of a power station according to claim 5, characterized in that: The pre-trained classifier includes: The modulo operation layer is used to perform a modulo operation on each element in the three-dimensional matrix of the electric field vector to obtain a three-dimensional matrix of the electric field scalar; The differential operation layer is used to obtain the difference between each element in the current electric field scalar three-dimensional matrix and the corresponding element in the electric field scalar three-dimensional matrix at the previous moment, and obtain the electric field difference component three-dimensional matrix; Reshape layer, used to rearrange the three-dimensional matrix of electric field difference components into a sequence of electric field difference components; At least one fully connected layer, used for processing the electric field difference component sequence through a fully connected neural network operation to obtain an electric field feature vector; The output layer is used to analyze the electric field feature vector and predict whether a partial discharge fault occurs.

7. The method for monitoring the operation status of a power station according to claim 6, characterized in that: The output layer expression of the pre-trained classifier is: ， Among them, f out is the partial discharge fault prediction value obtained by the output layer of the classifier. The data type is a 1-bit unsigned integer. Its value 1 represents that a partial discharge fault will occur, and its value 0 represents that a partial discharge fault will not occur. tanh(∙) is the hyperbolic tangent function. w i is the i-th weight coefficient of the output layer; f in (i) is the i-th element of the electric field eigenvector; b ias is the output layer bias coefficient; M is the length of the electric field eigenvector.

8. The method for monitoring the operation status of a power station according to claim 7, characterized in that: The pre-trained classifier is pre-trained using a plurality of three-dimensional matrices of electric field vectors labeled with whether a partial discharge fault occurs as samples, and the loss function used is: ， Wherein, r is the label of whether a partial discharge fault occurs in the sample, and the data type is a 1-bit unsigned integer, wherein a value of 1 represents that a partial discharge fault will occur, and a value of 0 represents that a partial discharge fault will not occur; f out is the partial discharge fault prediction value obtained by the output layer of the classifier. The data type is a 1-bit unsigned integer. Its value 1 represents that a partial discharge fault will occur, and its value 0 represents that a partial discharge fault will not occur. L is the loss function. log2(∙) is the logarithmic function with base 2.

Images