A highway tunnel power quality intelligent monitoring method and system integrating cloud detection and terminal control

Abstract

The present invention relates to the technical field of intelligent monitoring of power quality, and discloses a method and system for intelligent monitoring of power quality in highway tunnels that integrates cloud detection and terminal control, collects data from tunnel electromechanical, lighting, fan, fire protection, and monitoring components, establishes a GRU+Encoder+XGboost model, inputs the real-time collected data into the GRU+Encoder+XGboost model, judges the power quality of different circuits, and reports abnormal data if the circuit is abnormal. The method receives the data collected by the intelligent terminal control device through a cloud detection center, pre-processes and standardizes the data, and calibrates the data for normal data and various abnormal situations, and then trains the GRU+Encoder+XGboost model with the calibrated data, and finally uses the trained model to detect the real-time data transmitted by the terminal control device, so as to detect whether the tunnel power quality is abnormal, and if abnormal, the abnormal circuit is located and the corresponding circuit is cut off through the intelligent terminal control device.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring of power quality, and specifically to a method and system for intelligent monitoring of power quality in a highway tunnel integrating cloud detection and terminal control. Background Art

[0002] In recent years, with the rapid development of highway construction, by the end of 2023, the mileage of highways in my country has reached 177,000 kilometers, ranking first in the world. The huge highway network has formed an efficient and fast road transportation network, but it has also caused a lot of energy consumption;

[0003] By monitoring the power data of the tunnel, energy waste and power quality problems can be discovered and solved in a timely manner. However, tunnel energy consumption monitoring is difficult in the existing technology, with low execution efficiency and poor accuracy. Summary of the invention

[0004] 1. Technical issues to be solved

[0005] In view of the shortcomings of the existing technology, the present invention provides a method and system for intelligent monitoring of highway tunnel power quality that integrates cloud detection and terminal control, which has the advantages of accurate detection of tunnel power quality and solves the above-mentioned technical problems.

[0006] (II) Technical solution

[0007] To achieve the above object, the present invention provides the following technical solution: a method for intelligent monitoring of power quality in highway tunnels integrating cloud detection and terminal control, comprising the following steps:

[0008] S1. Collect the three-phase voltage, three-phase line voltage, three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate, three-phase current 3rd, 5th, 7th harmonics and frequency of the tunnel electromechanical, lighting, fan, fire protection and monitoring components;

[0009] S2, transmitting and storing the collected three-phase voltage, three-phase line voltage, three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate, three-phase current 3rd, 5th, 7th harmonics and frequency;

[0010] S3, preprocessing and calibrating the data stored in step S2, the preprocessing step includes:

[0011] S3.1, three-phase voltage V, three-phase line voltage V A 、V B 、V C 、V AB 、V AC 、V BC, three-phase voltage harmonic distortion rate HV, three-phase current harmonic distortion rate HI, three-phase current 3rd, 5th, 7th harmonics HI3, HI5, HI7, frequency F data are preprocessed and data standardized by removing noise data and filling missing values, and then the data is divided into 80% training set and 20% test set;

[0012] S3.2, arranging the sampled data at fixed time intervals to obtain input data, and performing data calibration on the input data;

[0013] S4. Establish a GRU+Encoder+XGboost model, and use the three-phase voltage, three-phase line voltage, three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate, three-phase current 3rd, 5th, 7th harmonics and frequency after preprocessing and data calibration in step S3 as input data to train the GRU+Encoder+XGboost model. The specific process is:

[0014] S4.1. Establish a GRU+Encoder+XGboost model, wherein the GRU+Encoder+XGboost model includes a GRU layer, an Encoder layer and an XGBoost layer, wherein the GRU layer is composed of a reset gate and an update gate, the Encoder layer is used to abstract and encode the data passing through the GRU layer, and input the encoded data into the XGBoost layer, and the XGBoost layer is used to output the classification result of the power quality, and judge the power quality according to the result;

[0015] S4.2, input the input data into the GRU+Encoder+XGboost model, and train the GRU+Encoder+XGboost model;

[0016] S4.3. Store the trained GRU+Encoder+XGboost model;

[0017] S5. Input the real-time collected data into the GRU+Encoder+XGboost model to judge the power quality of different circuits. If an abnormality occurs in the circuit, the abnormal data is reported.

[0018] As a preferred technical solution of the present invention, the specific expression of the input data in step S3.2 is as follows:

[0019]

[0020]

[0021] in, Respectively represent the three-phase line voltage of phases A, B, and C at time 1 to t, Respectively represent the three-phase line voltages of AB, AC, and BC phases at time 1 to t, They represent the three-phase voltage harmonic distortion rates of phases A, B, and C at time 1 to t, respectively. They represent the harmonic distortion rate of the three-phase current of phases A, B, and C at time 1 to t, respectively. Respectively represent the 3rd, 5th, and 7th harmonics of the three-phase current of phases A, B, and C at time 1 to t, It represents the frequency at time 1 to time t. When the input data is input, each column is regarded as a time series and inputted column by column.

[0022] As a preferred technical solution of the present invention, the calculation expression of the reset gate in the GRU layer is as follows:

[0023] r t =σ(W r [h t-1 ,x t ]+b)

[0024]

[0025] Among them, r t represents the reset gate, σ represents the sigmoid activation function, represents the hidden state at time t, W r and b represent the coefficient and bias coefficient of the linear relationship respectively, x t represents the input matrix at time t, * represents the Hadamard product, tanh represents the hyperbolic sine function, h t-1 represents the hidden state at time t-1, and W represents the weight matrix of the candidate hidden state at the current time;

[0026] The expression for updating the gate processing data in the GRU layer is as follows:

[0027] Z t =σ(W z [h t-1 ,x t ]+b t )

[0028]

[0029] Among them, * represents the Hadamard product, σ represents the sigmoid activation function, and b t represents the bias coefficient, h t-1 represents the hidden state at time t-1, h t Represents the output value at the current moment, W z Indicates the hidden state information h at the last moment in the update gate t-1 and the current input x tThe weight ratio of the information to be output by the update gate, Z t Represents the update gate.

[0030] As a preferred technical solution of the present invention, the expression calculated in the Encoder layer is as follows:

[0031]

[0032] in, They represent the output values ​​of the D1~Ds input matrix at the current moment, that is, the time series features of the three-phase line voltages A, B, and C extracted after passing through the GRU layer. D1~Ds represent the three-phase line voltages. They represent the output values ​​of the previous moment under the D1~Ds input matrix, that is, the time series characteristics of the three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate and the 3rd, 5th, and 7th harmonics of the three-phase current of A, B, and C at time 1~t, W e and W a They represent the time series characteristic weight matrices of the three-phase voltage, the three-phase voltage harmonic distortion rate, the three-phase current harmonic distortion rate, and the three-phase current 3rd, 5th, and 7th harmonics, respectively. c and b d Indicates the corresponding different bias coefficients in the Encoder layer.

[0033] As a preferred technical solution of the present invention, the tree model used in the XGBoost layer is a CART classification tree model based on the Gini index, and the calculation expression is as follows:

[0034]

[0035] Among them, Gini(D) represents the probability of two samples randomly selected from the data set D having inconsistent category labels, p k represents the probability of taking out a sample, N represents the number of feature values ​​output by the Encoder layer, Gini_index(D,a) represents the uncertainty of the data set D after segmentation by attribute a, |*| represents a positive value, D v Represents the vth subset among V subsets divided by V values.

[0036] As a preferred technical solution of the present invention, the specific steps of data calibration are as follows:

[0037] S3.2.1. Set loop labels including total loop, loop 1, loop 2, loop 3, ..., loop k;

[0038] S3.2.2. Set power quality type labels including: qualified, voltage interruption, abnormal voltage drop, abnormal voltage rise, three-phase voltage imbalance, abnormal frequency fluctuation, and abnormal harmonics;

[0039] S3.2.3. Based on steps S3.2.1 and S3.2.2, data labels are formed. The specific expression is as follows:

[0040]

[0041] Among them, 00, 01, 02, 03, 04, 05, 06 respectively represent the total circuit qualified, the total circuit voltage interruption, the total circuit voltage drop abnormal, the total circuit voltage rise abnormal, the total circuit three-phase voltage unbalance, the total circuit frequency fluctuation abnormal, the total circuit harmonic abnormal, 10, 11, 12, 13, 14, 15, 16 respectively represent the circuit 1 qualified, the circuit 1 voltage interruption, the circuit 1 voltage drop abnormal, the circuit 1 voltage rise abnormal, the circuit 1 three-phase voltage unbalance, the circuit 1 frequency fluctuation abnormal, the circuit 1 harmonic abnormal Normal, 20, 21, 22, 23, 24, 25, 26 respectively indicate that loop 2 is qualified, loop 2 voltage is interrupted, loop 2 voltage drops abnormally, loop 2 voltage rises abnormally, loop 2 three-phase voltage is unbalanced, loop 2 frequency fluctuation is abnormal, loop 2 harmonic is abnormal, k0, k1, k2, k3, k4, k5, k6 respectively indicate that loop k is qualified, loop k voltage is interrupted, loop k voltage drops abnormally, loop k voltage rises abnormally, loop k three-phase voltage is unbalanced, loop k frequency fluctuation is abnormal, loop k harmonic is abnormal;

[0042] S3.2.3. Add data labels to each column of data.

[0043] The present invention also provides an intelligent monitoring system for power quality in highway tunnels integrating cloud detection and terminal control, which adopts the above-mentioned intelligent monitoring method for power quality in highway tunnels integrating cloud detection and terminal control, including a meter, an intelligent terminal control device and a cloud detection center, wherein the meter is used to collect data of different circuits in tunnel electromechanical, lighting, fan, fire protection and monitoring components, including three-phase voltage, three-phase line voltage, three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate, three-phase current 3rd, 5th, 7th harmonics and frequency, the meter sends the collected data to the intelligent terminal control device for storage, the intelligent terminal control device is used to realize data aggregation, temporary storage, power supply cut-off of equipment according to circuit and data forwarding to the cloud detection center, and the cloud detection center stores the trained GRU+Encoder+XGboost model;

[0044] As a preferred technical solution of the present invention, the meter adopts the communication interface of RS485 and the communication method of Modbus-RTU protocol to transmit the collected three-phase voltage, three-phase line voltage, three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate, three-phase current 3rd, 5th, 7th harmonics and frequency to the intelligent terminal control device. The intelligent terminal control device sends the real-time collected data to the cloud detection center, and receives the power quality results output by the cloud detection center. If an abnormality occurs, the circuit is cut off.

[0045] Compared with the prior art, the present invention provides a method and system for intelligent monitoring of power quality in highway tunnels that integrates cloud detection and terminal control, which has the following beneficial effects:

[0046] The present invention receives the data collected by the intelligent terminal control device through the cloud detection center, pre-processes and standardizes the data, and calibrates the data for normal data and various abnormal situations. The calibrated data is then used to train the GRU+Encoder+XGboost model. Finally, the trained model is used to detect the real-time data transmitted by the terminal control device to detect whether the tunnel power quality is abnormal. If an abnormality occurs, the abnormal circuit is located and the corresponding circuit is cut off through the intelligent terminal control device. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of the process of the present invention;

[0048] Figure 2 This is a schematic diagram of the workflow of the cloud-based testing center of the present invention;

[0049] Figure 3 This is a schematic diagram of the working process of the intelligent terminal control device of the present invention;

[0050] Figure 4 This is a schematic diagram of the internal structure of the intelligent terminal control device of the present invention;

[0051] Figure 5 It is a schematic diagram of the system of the present invention;

[0052] Figure 6 This is a schematic diagram of the GRU+Encoder+XGboost model training workflow of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] See also Figure 1-6 , a highway tunnel power quality intelligent monitoring method integrating cloud detection and terminal control, comprising the following steps:

[0055] S1. Collect the three-phase voltage, three-phase line voltage, three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate, three-phase current 3rd, 5th, 7th harmonics and frequency of the tunnel electromechanical, lighting, fan, fire protection and monitoring components;

[0056] S2, transmitting and storing the collected three-phase voltage, three-phase line voltage, three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate, three-phase current 3rd, 5th, 7th harmonics and frequency;

[0057] S3, preprocessing and calibrating the data stored in step S2, the preprocessing step includes:

[0058] S3.1, three-phase voltage V, three-phase line voltage V A 、V B 、V C 、V AB 、V AC 、V BC , three-phase voltage harmonic distortion rate HV, three-phase current harmonic distortion rate HI, three-phase current 3rd, 5th, 7th harmonics HI3, HI5, HI7, frequency F data are preprocessed and data standardized by removing noise data and filling missing values, and then the data is divided into 80% training set and 20% test set;

[0059] S3.2, arranging the sampled data at fixed time intervals to obtain input data, and performing data calibration on the input data;

[0060] S4. Establish a GRU+Encoder+XGboost model, and use the three-phase voltage, three-phase line voltage, three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate, three-phase current 3rd, 5th, 7th harmonics and frequency after preprocessing and data calibration in step S3 as input data to train the GRU+Encoder+XGboost model. The specific process is:

[0061] S4.1. Establish a GRU+Encoder+XGboost model, wherein the GRU+Encoder+XGboost model includes a GRU layer, an Encoder layer and an XGBoost layer, wherein the GRU layer is composed of a reset gate and an update gate, the Encoder layer is used to abstract and encode the data passing through the GRU layer, and input the encoded data into the XGBoost layer, and the XGBoost layer is used to output the classification result of the power quality, and judge the power quality according to the result;

[0062] S4.2, input the input data into the GRU+Encoder+XGboost model, and train the GRU+Encoder+XGboost model;

[0063] S4.3. Store the trained GRU+Encoder+XGboost model;

[0064] S5. Input the real-time collected data into the GRU+Encoder+XGboost model to judge the power quality of different circuits. If an abnormality occurs in the circuit, the abnormal data is reported.

[0065] Furthermore, the specific expression of the input data in step S3.2 is as follows:

[0066]

[0067] in, Respectively represent the three-phase line voltage of phases A, B, and C at time 1 to t, Respectively represent the three-phase line voltages of AB, AC, and BC phases at time 1 to t, They represent the three-phase voltage harmonic distortion rates of phases A, B, and C at time 1 to t, respectively. They represent the harmonic distortion rate of the three-phase current of phases A, B, and C at time 1 to t, respectively. Respectively represent the 3rd, 5th, and 7th harmonics of the three-phase current of phases A, B, and C at time 1 to t, It represents the frequency at time 1 to t. When inputting data, each column is regarded as a time series and inputted column by column.

[0068] The specific steps of data calibration are as follows:

[0069] S3.2.1. Set loop labels including total loop, loop 1, loop 2, loop 3, ..., loop k;

[0070] S3.2.2. Set power quality type labels including: qualified, voltage interruption, abnormal voltage drop, abnormal voltage rise, three-phase voltage imbalance, abnormal frequency fluctuation, and abnormal harmonics;

[0071] S3.2.3. Based on steps S3.2.1 and S3.2.2, data labels are formed. The specific expression is as follows:

[0072]

[0073] Among them, 00, 01, 02, 03, 04, 05, 06 respectively represent the total circuit qualified, the total circuit voltage interruption, the total circuit voltage drop abnormal, the total circuit voltage rise abnormal, the total circuit three-phase voltage unbalance, the total circuit frequency fluctuation abnormal, the total circuit harmonic abnormal, 10, 11, 12, 13, 14, 15, 16 respectively represent the circuit 1 qualified, the circuit 1 voltage interruption, the circuit 1 voltage drop abnormal, the circuit 1 voltage rise abnormal, the circuit 1 three-phase voltage unbalance, the circuit 1 frequency fluctuation abnormal, the circuit 1 harmonic abnormal Normal, 20, 21, 22, 23, 24, 25, 26 respectively indicate that loop 2 is qualified, loop 2 voltage is interrupted, loop 2 voltage drops abnormally, loop 2 voltage rises abnormally, loop 2 three-phase voltage is unbalanced, loop 2 frequency fluctuation is abnormal, loop 2 harmonic is abnormal, k0, k1, k2, k3, k4, k5, k6 respectively indicate that loop k is qualified, loop k voltage is interrupted, loop k voltage drops abnormally, loop k voltage rises abnormally, loop k three-phase voltage is unbalanced, loop k frequency fluctuation is abnormal, loop k harmonic is abnormal;

[0074] Taking phase A voltage as an example, the input and output data formats of the GRU of phase A voltage are explained:

[0075]

[0076] For time t=1, the input voltage of phase A is The corresponding feature vector output is

[0077] Furthermore, the calculation expression of the reset gate in the GRU layer is as follows:

[0078] r t =σ(W r [h t-1 ,x t ]+b)

[0079]

[0080] Among them, r t represents the reset gate, σ represents the sigmoid activation function, represents the hidden state at time t, W r and b represent the coefficient and bias coefficient of the linear relationship respectively, x t represents the input matrix at time t, * represents the Hadamard product, tanh represents the hyperbolic sine function, h t-1 represents the hidden state at time t-1, and W represents the weight matrix of the candidate hidden state at the current time, which is responsible for determining how to get the current input information x t Construct candidate memory states in t The proportion of information output so that it can be subsequently combined with the output of the update gate;

[0081] r t The smaller the value of is, the closer it is to the previous hidden state h t-1 The smaller the Hadamard product is, the value W[r t *h t-1 ] is smaller, indicating that more things need to be forgotten and discarded at the previous moment; when r t When the value is close to 0, the value is also close to 0, indicating that the content of the previous moment is completely discarded, and only the input of the current moment is retained, which is used to discard historical information that is not related to the prediction;

[0082] r t The larger the value of W[r t *h t-1 ]The larger the value, the more information needs to be remembered at the previous moment. t The more it is combined with the previous memory; when r t When it is close to 1, it means that the hidden state h of the previous moment is retained. t-1

[0083] The expression for updating the gate processing data in the GRU layer is as follows:

[0084] Z t =σ(W z [h t-1 ,x t ]+b t )

[0085]

[0086] Among them, * represents the Hadamard product, σ represents the sigmoid activation function, and b t represents the bias coefficient, h t-1 represents the hidden state at time t-1, W z and Z t Represent the weight matrix and update gate respectively, h t Represents the output value at the current moment, W z Indicates the hidden state information h at the last moment in the update gate t-1 and the current input x t The weight ratio of the information to be output by the update gate determines how the current input information is combined with the previous memory state. t That is, the update gate, which is output by a sigmoid function with an output value between 0 and 1, which is used to control the proportion of retaining or discarding the current memory state, Z t The closer the elements in are to 1, the model will tend to retain more historical information and use less of the latest input x. t .

[0087] The update gate is used to control the state information h of the previous moment t-1 The update gate helps the model decide how much past information to carry into the future. t and h t-1 To determine the updated part of the GRU unit state information at the current moment, on the one hand, it is necessary to determine the proportion of the updated information; on the other hand, the specific content of the updated information is determined through the tanh function channel.

[0088] Furthermore, the expression calculated in the Encoder layer is as follows:

[0089]

[0090] in, They represent the output values ​​of the D1~Ds input matrix at the current moment, that is, the time series features of the three-phase line voltages A, B, and C extracted after passing through the GRU layer. D1~Ds represent the three-phase line voltages. They represent the output values ​​of the previous moment under the D1~Ds input matrix, that is, the time series characteristics of the three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate and the 3rd, 5th, and 7th harmonics of the three-phase current of A, B, and C at time 1~t, W e and W a They represent the time series characteristic weight matrices of the three-phase voltage, the three-phase voltage harmonic distortion rate, the three-phase current harmonic distortion rate, and the three-phase current 3rd, 5th, and 7th harmonics, respectively. c and b d Indicates the corresponding different bias coefficients in the Encoder layer.

[0091] Furthermore, the tree model used in the XGBoost layer is the CART classification tree model based on the Gini index. The calculation expression is as follows:

[0092]

[0093]

[0094] Among them, Gini(D) represents the probability of two samples randomly selected from the data set D having inconsistent category labels, p k represents the probability of taking out a sample, N represents the number of feature values ​​output by the Encoder layer, Gini_index(D,a) represents the uncertainty of the data set D after segmentation by attribute a, |*| represents a positive value, D v Represents the vth subset among V subsets divided by V values.

[0095] S3.2.3. Add data labels to each column of data.

[0096] The present invention also provides an intelligent monitoring system for power quality in highway tunnels that integrates cloud detection and terminal control. The above-mentioned intelligent monitoring method for power quality in highway tunnels that integrates cloud detection and terminal control includes a meter, an intelligent terminal control device and a cloud detection center. The meter is used to collect data from different circuits in tunnel electromechanical, lighting, fan, fire protection and monitoring components, including three-phase voltage, three-phase line voltage, three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate, three-phase current 3rd, 5th, 7th harmonics and frequency. The meter sends the collected data to the intelligent terminal control device for storage. The intelligent terminal control device is used to realize data aggregation, temporary storage, power supply cut-off of equipment according to circuits and data forwarding to the cloud detection center. The cloud detection center stores a trained GRU+Encoder+XGboost model.

[0097] Furthermore, the meter uses the RS485 communication interface and the Modbus-RTU protocol communication method to transmit the collected three-phase voltage, three-phase line voltage, three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate, three-phase current 3rd, 5th, 7th harmonics and frequency to the intelligent terminal control device. The intelligent terminal control device sends the real-time collected data to the cloud detection center, and receives the power quality results output by the cloud detection center. If an abnormality occurs, the circuit is cut off.

[0098] The intelligent terminal control device comprises a data receiver, a processor, a power cut-off device and a communication module.

[0099] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent monitoring of power quality in highway tunnels integrating cloud detection and terminal control, characterized in that: The following steps are involved: S1. Collect the three-phase voltage, three-phase line voltage, three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate, three-phase current 3rd, 5th, 7th harmonics and frequency of the tunnel electromechanical, lighting, fan, fire protection and monitoring components; S2, transmitting and storing the collected three-phase voltage, three-phase line voltage, three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate, three-phase current 3rd, 5th, 7th harmonics and frequency; S3, preprocessing and calibrating the data stored in step S2, the preprocessing step includes: S3.1, three-phase voltage V, three-phase line voltage V A 、V B 、V C 、V AB 、V AC 、V BC , three-phase voltage harmonic distortion rate HV, three-phase current harmonic distortion rate HI, three-phase current 3rd, 5th, 7th harmonics HI3, HI5, HI7, frequency F data are preprocessed and data standardized by removing noise data and filling missing values, and then the data is divided into 80% training set and 20% test set; S3.2, arranging the sampled data at fixed time intervals to obtain input data, and performing data calibration on the input data; S4. Establish a GRU+Encoder+XGboost model, and use the three-phase voltage, three-phase line voltage, three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate, three-phase current 3rd, 5th, 7th harmonics and frequency after preprocessing and data calibration in step S3 as input data to train the GRU+Encoder+XGboost model. The specific process is: S4.

1. Establish a GRU+Encoder+XGboost model, wherein the GRU+Encoder+XGboost model includes a GRU layer, an Encoder layer and an XGBoost layer, wherein the GRU layer is composed of a reset gate and an update gate, the Encoder layer is used to abstract and encode the data passing through the GRU layer, and input the encoded data into the XGBoost layer, and the XGBoost layer is used to output the classification result of the power quality, and judge the power quality according to the result; S4.2, input the input data into the GRU+Encoder+XGboost model, and train the GRU+Encoder+XGboost model; S4.

3. Store the trained GRU+Encoder+XGboost model; S5. Input the real-time collected data into the GRU+Encoder+XGboost model to judge the power quality of different circuits. If an abnormality occurs in the circuit, the abnormal data is reported.

2. According to claim 1, a highway tunnel power quality intelligent monitoring method integrating cloud detection and terminal control is characterized by: The specific expression of the input data in step S3.2 is as follows: in, Respectively represent the three-phase line voltage of phases A, B, and C at time 1 to t, Respectively represent the three-phase line voltages of AB, AC, and BC phases at time 1 to t, They represent the three-phase voltage harmonic distortion rates of phases A, B, and C at time 1 to t, respectively. They represent the harmonic distortion rate of the three-phase current of phases A, B, and C at time 1 to t, respectively. Respectively represent the 3rd, 5th, and 7th harmonics of the three-phase current of phases A, B, and C at time 1 to t, It represents the frequency at time 1 to time t. When the input data is input, each column is regarded as a time series and inputted column by column.

3. According to claim 2, a highway tunnel power quality intelligent monitoring method integrating cloud detection and terminal control is characterized by: The calculation expression of the reset gate in the GRU layer is as follows: r t =σ(W r [h t-1 ,x t ]+b) Among them, r t represents the reset gate, σ represents the sigmoid activation function, represents the hidden state at time t, W r and b represent the coefficient and bias coefficient of the linear relationship respectively, x t represents the input matrix at time t, * represents the Hadamard product, tanh represents the hyperbolic sine function, h t-1 represents the hidden state at time t-1, and W represents the weight matrix of the candidate hidden state at the current time; The expression for updating the gate processing data in the GRU layer is as follows: Z t =σ(W z [h t-1 ,x t ]+b t ) Among them, * represents the Hadamard product, σ represents the sigmoid activation function, and b t represents the bias coefficient, h t-1 represents the hidden state at time t-1, h t Represents the output value at the current moment, W z Indicates the hidden state information h at the last moment in the update gate t-1 and the current input x t The weight ratio of the information to be output by the update gate, Z t Represents the update gate.

4. According to claim 1, a highway tunnel power quality intelligent monitoring method integrating cloud detection and terminal control is characterized by: The expression calculated in the Encoder layer is as follows: in, They represent the output values ​​of the D1~Ds input matrix at the current moment, that is, the time series features of the three-phase line voltages A, B, and C extracted after passing through the GRU layer. D1~Ds represent the three-phase line voltages. They represent the output values ​​of the previous moment under the D1~Ds input matrix, that is, the time series characteristics of the three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate and the 3rd, 5th, and 7th harmonics of the three-phase current of A, B, and C at time 1~t, W e and W a They represent the time series characteristic weight matrices of the three-phase voltage, the three-phase voltage harmonic distortion rate, the three-phase current harmonic distortion rate, and the three-phase current 3rd, 5th, and 7th harmonics, respectively. c and b d Indicates the corresponding different bias coefficients in the Encoder layer.

5. According to claim 1, a highway tunnel power quality intelligent monitoring method integrating cloud detection and terminal control is characterized by: The tree model used in the XGBoost layer is a CART classification tree model based on the Gini index. The calculation expression is as follows: Among them, Gini(D) represents the probability of two samples randomly selected from the data set D having inconsistent category labels, p k represents the probability of taking out a sample, N represents the number of feature values ​​output by the Encoder layer, Gini_index(D,a) represents the uncertainty of the data set D after segmentation by attribute a, |*| represents a positive value, D v Represents the vth subset among V subsets divided by V values.

6. The method for intelligent monitoring of power quality in highway tunnels integrating cloud detection and terminal control according to claim 2 is characterized in that: The specific steps of the data calibration are as follows: S3.2.

1. Set loop labels including total loop, loop 1, loop 2, loop 3, ..., loop k; S3.2.

2. Set power quality type labels including: qualified, voltage interruption, abnormal voltage drop, abnormal voltage rise, three-phase voltage imbalance, abnormal frequency fluctuation, and abnormal harmonics; S3.2.

3. Based on steps S3.2.1 and S3.2.2, data labels are formed. The specific expression is as follows: Among them, 00, 01, 02, 03, 04, 05, 06 respectively represent the total circuit qualified, the total circuit voltage interruption, the total circuit voltage drop abnormal, the total circuit voltage rise abnormal, the total circuit three-phase voltage unbalance, the total circuit frequency fluctuation abnormal, the total circuit harmonic abnormal, 10, 11, 12, 13, 14, 15, 16 respectively represent the circuit 1 qualified, the circuit 1 voltage interruption, the circuit 1 voltage drop abnormal, the circuit 1 voltage rise abnormal, the circuit 1 three-phase voltage unbalance, the circuit 1 frequency fluctuation abnormal, the circuit 1 harmonic abnormal Normal, 20, 21, 22, 23, 24, 25, 26 respectively indicate that loop 2 is qualified, loop 2 voltage is interrupted, loop 2 voltage drops abnormally, loop 2 voltage rises abnormally, loop 2 three-phase voltage is unbalanced, loop 2 frequency fluctuation is abnormal, loop 2 harmonic is abnormal, k0, k1, k2, k3, k4, k5, k6 respectively indicate that loop k is qualified, loop k voltage is interrupted, loop k voltage drops abnormally, loop k voltage rises abnormally, loop k three-phase voltage is unbalanced, loop k frequency fluctuation is abnormal, loop k harmonic is abnormal; S3.2.

3. Add data labels to each column of data.

7. A highway tunnel power quality intelligent monitoring system integrating cloud detection and terminal control, adopting a highway tunnel power quality intelligent monitoring method integrating cloud detection and terminal control as claimed in any one of claims 1 to 5, characterized in that: It includes a meter, an intelligent terminal control device and a cloud detection center. The meter is used to collect data from different circuits in the tunnel electromechanical, lighting, fan, fire protection and monitoring components, including three-phase voltage, three-phase line voltage, three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate, three-phase current 3rd, 5th, 7th harmonics and frequency. The meter sends the collected data to the intelligent terminal control device for storage. The intelligent terminal control device is used to realize data aggregation, temporary storage, power supply cut-off of equipment according to circuits and data forwarding to the cloud detection center. The cloud detection center stores the trained GRU+Encoder+XGboost model.

8. The highway tunnel power quality intelligent monitoring system integrating cloud detection and terminal control according to claim 7 is characterized by: The meter adopts RS485 communication interface and Modbus-RTU protocol communication mode to transmit the collected three-phase voltage, three-phase line voltage, three-phase voltage harmonic distortion rate, three-phase current harmonic distortion rate, three-phase current 3rd, 5th, 7th harmonics and frequency to the intelligent terminal control device. The intelligent terminal control device sends the real-time collected data to the cloud detection center, and receives the power quality results output by the cloud detection center. If an abnormality occurs, the circuit is cut off.

Images