A method and system for intelligently identifying abnormal energy consumption of tunnel equipment based on cloud-edge collaboration

Abstract

The present invention relates to the technical field of power data processing of electrical equipment in highway tunnels, and discloses a method and system for intelligently identifying energy consumption anomalies of tunnel equipment based on cloud-edge collaboration, including a data acquisition module, a storage module, an edge detection module and a communication module, wherein the data acquisition module, the storage module and the edge detection module are all connected by a communication module. The method and system for intelligently identifying energy consumption anomalies of tunnel equipment based on cloud-edge collaboration adopts the method of LSTM-BP neural network model to capture the timing and three-phase relationship of power consumption data, which can greatly improve the accuracy of detection. The intelligent power anomaly recognition system has accurate anomaly detection, good data transmission quality assurance and a long transmission distance that meets the actual situation of long tunnel mileage. At the same time, the hardware cost is low and easy to implement, which alleviates the workload of operation and maintenance personnel, improves work efficiency, reduces the possibility of human errors, and reduces energy waste.

Description

Technical Field

[0001] The present invention relates to the technical field of power data processing for electrical equipment in highway tunnels, and specifically to a method and system for intelligently identifying energy consumption anomalies of tunnel equipment based on cloud-edge collaboration. Background Art

[0002] As a tubular semi-enclosed structure, highway tunnels also have unavoidable shortcomings. For example, when the driver enters the tunnel, the light becomes dim, and the lighting system is needed to illuminate in order to ensure driving safety. After an accident occurs in the tunnel, timely ventilation and smoke exhaust are required. Therefore, tunnels undertake multiple functions such as lighting, ventilation, smoke exhaust, and monitoring, and the normal operation of these functions is inseparable from a stable and reliable power supply and the normal operation of equipment. Once the power supply is abnormal, the equipment fails, or a short circuit occurs, the lighting, monitoring, ventilation and other systems inside the tunnel may fail, increasing the risk of accidents. The tunnel environment is complex and is greatly affected by factors such as terrain and climate. Power equipment is easily affected by moisture and pollution, which may lead to equipment failure, electrical short circuits, and increased energy consumption. Summary of the invention

[0003] 1. Technical issues to be resolved

[0004] In response to the shortcomings of the prior art, the present invention provides a method and system for intelligently identifying energy consumption anomalies of tunnel equipment based on cloud-edge collaboration, which has the advantages of reducing the possibility of human errors and reducing energy waste, and solves the above-mentioned technical problems.

[0005] (II) Technical solution

[0006] To achieve the above object, the present invention provides the following technical solution: a method for intelligently identifying abnormal energy consumption of tunnel equipment based on cloud-edge collaboration, comprising the following steps:

[0007] S1. Collect and sort out the electricity consumption data of tunnel equipment, where the types of electricity consumption data in the tunnel include three-phase current Three-phase voltage Three-phase active power and reactive power t represents the time, It represents the three phases A, B, and C, and the data expression after sorting is as follows:

[0008]

[0009] in, Respectively represent phase A t1, t2, ..., t n The current at the moment, Respectively represent B phase t1, t2, ..., t n The current at the moment, Respectively represent C phase t1, t2, ..., t n The current at the moment, Respectively represent C phase t1, t2, ..., t n Reactive power at the moment;

[0010] S2. Normalize the data obtained in step S1. Taking the current of phase A as an example, the formula is as follows:

[0011]

[0012] in, is the normalized data, and are the maximum and minimum values ​​in the data, respectively. is the actual electricity consumption data;

[0013] S3, set the input window length l and step length step, 2≤l≤12, 1≤step≤5, and establish an LSTM model, where the LSTM model consists of a forget gate, an input gate, and an output gate;

[0014] The expression of the forget gate is as follows:

[0015] f t =σ(W[h t-1 ,x t ]+b)

[0016] Among them, f t represents the output of the forget gate at time t, h t-1 represents the hidden state at the previous moment, [*,*] represents connecting two vectors into a matrix, x t represents the input at the current moment, W and b represent the coefficient and bias coefficient of the linear relationship respectively, and σ(*) represents the Sigmoid activation function;

[0017] The expression of the input gate is as follows:

[0018] i t =σ(W i [h t-1 ,x t ]+b t )

[0019]

[0020]

[0021] Among them, i t Represents the update information of the output gate at time t, Indicates the current input unit state, h t-1 Indicates the hidden state at the previous moment, xt represents the input at the current moment, W i and b t Respectively represent the update information i of the output gate at time t t The weight matrix and bias term, W c and b c They represent, tanh(*) represents the hyperbolic tangent function, σ(*) represents the Sigmoid activation function, C t Indicates the current unit state, C t-1 represents the unit state at the previous moment, f t Represents the output of the forget gate at time t;

[0022] The expression of the output gate is as follows:

[0023] o t =σ(W o [h t-1 ,x t ]+b o )

[0024] h t =o t tanh(C t )

[0025] Among them, h t-1 Indicates the hidden state at the previous moment, x t represents the input at the current moment, σ(*) represents the Sigmoid activation function, tanh(*) represents the hyperbolic tangent function, C t Indicates the unit state at the current moment, o t Indicates the current information output level, h t is the output hidden state vector;

[0026] S4. Input the data in step S1 into the established LSTM model according to the window length l and the moving window step length step, and obtain the output feature space expression as follows:

[0027]

[0028] in, They represent the characteristics of the tunnel device A phase current length 1, length 2, ..., up to length len, respectively. They represent the characteristics of the tunnel device B phase current length 1, length 2, ..., up to length len, respectively. Respectively represent tunnel equipment The features of the τth type of electricity consumption data are length 1, length 2, ..., up to length len, where τ represents the type of electricity consumption data and the superscript len ​​represents the length of the feature space;

[0029] S5, arranging the data in step S4 in order and inputting them into the BP neural network model to obtain the abnormal power consumption result. If the abnormal power consumption result is acceptable, the model training is completed. If it is unacceptable, the back propagation algorithm is used to adjust the model parameters until the result is acceptable, and finally the LSTM-BP model is obtained.

[0030] S6. Input actual tunnel equipment power consumption data and predict abnormal power consumption data.

[0031] As a preferred technical solution of the present invention, the l+1th data output in step S2 is consistent with the real data y t If the difference exceeds the threshold ε, the electricity consumption behavior is abnormal and an alarm is issued.

[0032] As a preferred technical solution of the present invention, the model training process in step S5 uses the SMOTE algorithm to balance the data, and the specific steps are as follows:

[0033] S5.1. Select abnormal electricity consumption data sample X min , x i ∈X min , x i is the abnormal power consumption data sample X min The i-th data in the i The remaining power consumption abnormal samples are y i , and calculate the Euclidean distance d(x,y) in the m-dimensional space. The specific expression is as follows:

[0034]

[0035] Among them, y i Indicates the removal of x i The remaining power consumption anomaly samples, d(x,y) represents the Euclidean distance in the m-dimensional space, x i Represents abnormal power consumption data sample X min The i-th data in It means to sum up m data and get the distance x i The most recent k X min Sample data in ;

[0036] S5.2, from the distance x i The most recent k X min Randomly select a sample from the sample data in i Combine to form a new sample. The specific expression is as follows:

[0037]

[0038] Among them, x new represents a new synthesized sample, Indicates the distance xi The most recent k X min A sample is randomly selected from the sample data in , δ represents the sampling ratio, and δ∈[0,1].

[0039] As a preferred technical solution of the present invention, the number of hidden layers in the BP neural network model is four, and the activation function in the hidden layer adopts the tanh function, and the activation function of the output layer in the BP neural network model adopts the softmax function.

[0040] An intelligent recognition system for abnormal energy consumption of tunnel equipment based on cloud-edge collaboration includes a data acquisition module, a storage module, an edge detection module and a communication module, wherein the data acquisition module, the storage module and the edge detection module are all connected by the communication module;

[0041] The data acquisition module is used to collect electricity consumption data, and send the collected electricity consumption data to the smart cloud data processing center and the edge detection module through the communication module. The smart cloud data processing center stores the above-mentioned LSTM-BP model constructed based on the cloud-edge collaborative tunnel equipment energy consumption anomaly intelligent identification method, and sends the LSTM-BP model to the edge detection module. The edge detection module reads real-time data for detection, and if an abnormality occurs, it sends alarm data to the tunnel management center. The storage module is used to store data in the acquisition module, the smart cloud data processing center and the edge detection module.

[0042] As a preferred technical solution of the present invention, the data acquisition module collects electricity consumption data including three-phase voltage, three-phase current, three-phase active power and reactive power in tunnel electrical equipment, and is composed of an IM3332 three-phase power quality monitoring meter and an RS485 to TCP / IP module. The edge detection module uses an embedded development board customized by Raspberry Pi 5.

[0043] Compared with the prior art, the present invention provides a method and system for intelligently identifying abnormal energy consumption of tunnel equipment based on cloud-edge collaboration, which has the following beneficial effects:

[0044] The present invention can greatly improve the accuracy of detection by adopting the LSTM-BP neural network model method to capture the time series and three-phase relationship of power consumption data. The power consumption intelligent anomaly recognition system has accurate anomaly detection, good data transmission quality assurance and a long transmission distance that meets the actual situation of long tunnel mileage. At the same time, the hardware cost is low and easy to implement, which relieves the workload of operation and maintenance personnel, improves work efficiency, reduces the possibility of human error, and reduces energy waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of the process of the present invention;

[0046] Figure 2 This is a schematic diagram of the classification of the tunnel system of the present invention;

[0047] Figure 3 This is a schematic diagram of the structure of the cloud-edge collaborative tunnel power consumption anomaly intelligent identification system of the present invention;

[0048] Figure 4 It is a schematic diagram of the abnormal power consumption detection model of the present invention. DETAILED DESCRIPTION

[0049] 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.

[0050] See also Figure 1-4 The tunnel power system can be divided into power supply and distribution system, lighting system, ventilation system, fire alarm system, traffic control system, environmental monitoring system, emergency telephone and broadcasting system and monitoring video system according to its function. Each system includes one or more tunnel equipment, which are specifically classified into categories such as Figure 2

[0051] A method for intelligently identifying abnormal energy consumption of tunnel equipment based on cloud-edge collaboration includes the following steps:

[0052] S1. Collect and sort out the electricity consumption data of tunnel equipment, where the types of electricity consumption data in the tunnel include three-phase current Three-phase voltage Three-phase active power and reactive power t represents the time, It represents the three phases A, B, and C, and the data expression after sorting is as follows:

[0053]

[0054] in, Respectively represent phase A t1, t2, ..., t n The current at the moment, Respectively represent B phase t1, t2, ..., t n The current at the moment, Respectively represent C phase t1, t2, ..., t n The current at the moment, Respectively represent C phase t1, t2, ..., t n Reactive power at the moment;

[0055] S2, normalizing the data obtained in step S1;

[0056] S3, set the input window length l and step length step, and establish an LSTM model, where the LSTM model consists of a forget gate, an input gate, and an output gate;

[0057] The expression of the forget gate is as follows:

[0058] f t =σ(W[h t-1 ,x t ]+b)

[0059] Among them, f t represents the output of the forget gate at time t, h t-1 represents the hidden state at the previous moment, [*,*] represents connecting two vectors into a matrix, x t represents the input at the current moment, W and b represent the coefficient and bias coefficient of the linear relationship respectively, and σ(*) represents the Sigmoid activation function;

[0060] The expression of the input gate is as follows:

[0061] i t =σ(W i [h t-1 ,x t ]+b t )

[0062]

[0063] Among them, i t Represents the update information of the output gate at time t, Indicates the unit status, Indicates the current input unit state, h t-1 Indicates the hidden state at the previous moment, x t represents the input at the current moment, W i and b t Respectively represent the update information i of the output gate at time t t The weight matrix and bias term, W c and b c Represent the input gate weight matrix and bias term respectively, tanh(*) represents the hyperbolic tangent function, σ(*) represents the Sigmoid activation function, C t Indicates the current unit state, C t-1 represents the unit state at the previous moment, f t Represents the output of the forget gate at time t;

[0064] The expression of the output gate is as follows:

[0065] o t =σ(W o [h t-1 ,x t ]+b o )

[0066] h t =o t tanh(C t )

[0067]

[0068] Among them, h t-1 Indicates the hidden state at the previous moment, x t represents the input at the current moment, σ(*) represents the Sigmoid activation function, tanh(*) represents the hyperbolic tangent function, C t Indicates the unit state at the current moment, o t Indicates the current information output level, h t is the output hidden state vector, Indicates output;

[0069] S4, according to the window length l and step size step moving window to step S1 t1, t2, ..., t n The data is input to the established LSTM1, LSTM2...LSTM m In the m LSTM models, the output feature space expression is as follows:

[0070]

[0071] in, They represent the characteristics of the tunnel device A phase current length 1, length 2, ..., up to length len, respectively. They represent the characteristics of the tunnel device B phase current length 1, length 2, ..., up to length len, respectively. Respectively represent tunnel equipment The features of the τth type of electricity consumption data are length 1, length 2, ..., up to length len, where τ represents the type of electricity consumption data and the superscript len ​​represents the length of the feature space;

[0072] S5. Arrange the data in step S4 in order and input them into the BP neural network model to obtain the power consumption anomaly result. If the obtained power consumption anomaly manual auxiliary judgment result is acceptable, the LSTM-BP model training is completed. If it is unacceptable, the back propagation algorithm is used to adjust the model parameters until the result is acceptable, and finally the LSTM-BP model is obtained; the columns are arranged as follows:

[0073]

[0074] The specific examples by column are as follows:

[0075]

[0076] S6. Input the actual electricity consumption data of the tunnel equipment and judge the abnormal electricity consumption data. That is, the result is judged by the trained model as abnormal electricity consumption category 1, abnormal electricity consumption category 2, or abnormal electricity consumption category n, and an alarm is issued.

[0077] Furthermore, the SMOTE algorithm is used in the model training process in step S5 to balance the data. The abnormal power consumption data belongs to small sample data, which may lead to inaccurate parameter training of the model. Therefore, the SMOTE algorithm is used to balance the data. The SMOTE algorithm does not repeatedly copy samples to expand the sample set, but is based on the interpolation method to sum a few samples into new samples to balance the data sample set. The main construction process is as follows:

[0078] S5.1. Select abnormal electricity consumption data sample X min , x i ∈X min , x i is the abnormal power consumption data sample X min The i-th data in the i The remaining power consumption abnormal samples are y i , and calculate the Euclidean distance d(x,y) in the m-dimensional space. The specific expression is as follows:

[0079]

[0080] Among them, y i Indicates the removal of x i The remaining power consumption anomaly samples, d(x,y) represents the Euclidean distance in the m-dimensional space, x i Represents abnormal power consumption data sample X min The i-th data in It means to sum up m data and get the distance x i The most recent k X min Sample data in ;

[0081] S5.2, from the distance x i The most recent k X min Randomly select a sample from the sample data in i Combine to form a new sample. The specific expression is as follows:

[0082]

[0083] Among them, x new represents a new synthesized sample, Indicates the distance xi The most recent k X min A sample is randomly selected from the sample data in , δ represents the sampling ratio and δ∈[0,1].

[0084] Furthermore, generally speaking, the more layers a neural network has, the smaller its error is. However, as the number of layers increases, it is also possible that the reverse gradient disappears and the gradient explodes. Therefore, for the electricity consumption data of tunnel equipment, the number of hidden layers in the BP neural network model is optionally four, and the activation function in the hidden layer uses the tanh function, and the output layer activation function in the BP neural network model uses the softmax function.

[0085] An intelligent recognition system for abnormal energy consumption of tunnel equipment based on cloud-edge collaboration includes a data acquisition module, a storage module, an edge detection module and a communication module, wherein the data acquisition module, the storage module and the edge detection module are all connected by the communication module;

[0086] The data acquisition module is used to collect electricity consumption data, and send the collected electricity consumption data to the smart cloud data processing center and the edge detection module through the communication module. The smart cloud data processing center stores the above-mentioned LSTM-BP model constructed based on the cloud-edge collaborative tunnel equipment energy consumption anomaly intelligent identification method, and sends the LSTM-BP model to the edge detection module. The edge detection module reads real-time data for detection. If an abnormality occurs, it sends alarm data to the tunnel management center. The storage module is used to store data in the acquisition module, the smart cloud data processing center and the edge detection module.

[0087] Furthermore, the data acquisition module collects power consumption data including three-phase voltage, three-phase current, three-phase active power and reactive power in the tunnel power equipment. It is composed of IM3332 three-phase power quality monitoring meter and RS485 to TCP / IP module. The edge detection module adopts Raspberry Pi 5 customized embedded development board. In addition, it also includes power supply, display, memory card, communication module, etc. It runs Ubuntu 23.10 operating system and loads PyTorch deep learning framework. The functions are as follows: receiving power data collected from the underlying communication model and temporarily storing it; receiving control information transmitted from the cloud, regularly receiving and storing LSTM-BP models transmitted from the cloud, and running updated LSTM-BP models; preprocessing, standardizing, and reconstructing the input data of the real-time collected power data, using the updated LSTM-BP model to perform real-time energy consumption anomaly detection on the collected power data, and storing the abnormal results and data; if an abnormality is found, sending an alarm message to the cloud and the tunnel management center; performing corresponding operations according to the instructions of the management center and the cloud and the security level to ensure the safety of electricity use in the tunnel; providing maintenance personnel with operation functions and interfaces for on-site inspection and maintenance;

[0088] The data acquisition module is installed in the power supply and distribution system, lighting system, ventilation system, fire alarm system, traffic control system, environmental monitoring system, emergency telephone and broadcasting system and monitoring video system loop. The IM3332 three-phase power quality monitoring meter collects power consumption data in real time. The meter has an RS485 interface, which can be directly connected to RS485. At the same time, the RS485 is connected to the internal RS485 to TCP / IP module, and the module outputs the data as the final output and transmits it to the gateway. In the edge detection module, the Raspberry Pi issues a command, connects to the gateway through the transmission medium, obtains the power consumption data in the gateway for detection. At the same time, the smart cloud data processing center regularly obtains the data in the gateway and updates the model, and returns the model to the edge detection module. Optionally, the gateway can use the "one master and multiple slaves" method to poll the data acquisition module to obtain data to reduce communication conflicts and delays.

[0089] 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 intelligently identifying abnormal energy consumption of tunnel equipment based on cloud-edge collaboration, characterized in that: The following steps are involved: S1. Collect and sort out the electricity consumption data of tunnel equipment, where the types of electricity consumption data in the tunnel include three-phase current Three-phase voltage Three-phase active power and reactive power t represents the time, It represents the three phases A, B, and C, and the data expression after sorting is as follows: in, Respectively represent phase A t1, t2, ..., t n The current at the moment, Respectively represent B phase t1, t2, ..., t n The current at the moment, Respectively represent C phase t1, t2, ..., t n The current at the moment, Respectively represent C phase t1, t2, ..., t n Reactive power at the moment; S2, normalizing the data obtained in step S1; S3, set the input window length l and step size step, and establish the LSTM model; S4, move the window of the data in step S1 according to the window length l and the step length step to convert t1, t2, ..., t n Data is input to the established LSTM1, LSTM2...LSTM m In the m LSTM models, the output feature space expression is as follows: in, They represent the characteristics of the tunnel device A phase current length 1, length 2, ..., up to length len, respectively. They represent the characteristics of the tunnel device B phase current length 1, length 2, ..., up to length len, respectively. Respectively represent tunnel equipment The features of the τth type of electricity consumption data are length 1, length 2, ..., up to length len, where τ represents the type of electricity consumption data and the superscript len ​​represents the length of the feature space; S5, arranging the data in step S4 in order and inputting them into the BP neural network model to obtain the abnormal power consumption result. If the abnormal power consumption result is acceptable, the model training is completed. If it is unacceptable, the back propagation algorithm is used to adjust the model parameters until the result is acceptable, and finally the LSTM-BP model is obtained. S6. Input actual tunnel equipment power consumption data and predict abnormal power consumption data.

2. According to claim 1, a method for intelligently identifying abnormal energy consumption of tunnel equipment based on cloud-edge collaboration is characterized by: The LSTM model consists of a forget gate, an input gate and an output gate; The expression of the forget gate is as follows: f t =σ(W[h t-1 ,x t ]+b) Among them, f t represents the output of the forget gate at time t, h t-1 represents the hidden state at the previous moment, [*,*] represents connecting two vectors into a matrix, x t represents the input at the current moment, W and b represent the coefficient and bias coefficient of the linear relationship respectively, and σ(*) represents the Sigmoid activation function; The expression of the input gate is as follows: i t =σ(W i [h t-1 ,x t ]+b t ) Among them, i t Represents the update information of the output gate at time t, Indicates the unit status, Indicates the current input unit state, h t-1 Indicates the hidden state at the previous moment, x t represents the input at the current moment, W i and b t Respectively represent the update information i of the output gate at time t t The weight matrix and bias term, W c and b c Represent the input gate weight matrix and bias term respectively, tanh(*) represents the hyperbolic tangent function, σ(*) represents the Sigmoid activation function, C t Indicates the current unit state, C t-1 represents the unit state at the previous moment, f t Represents the output of the forget gate at time t; The expression of the output gate is as follows: the t =σ(W o [h t-1 ,x t ]+b o ) h t =o t ·tanh(C t ) Among them, h t-1 Indicates the hidden state at the previous moment, x t represents the input at the current moment, σ(*) represents the Sigmoid activation function, tanh(*) represents the hyperbolic tangent function, C t Indicates the unit state at the current moment, o t Indicates the current information output level, h t is the output hidden state vector, W o and b o Represents the weight matrix and bias term of the output gate.

3. According to claim 1, a method for intelligently identifying abnormal energy consumption of tunnel equipment based on cloud-edge collaboration is characterized in that: The model training process in step S5 uses the SMOTE algorithm to balance the data, and the specific steps are as follows: S5.

1. Select abnormal electricity consumption data sample X min , x i ∈X min , x i is the abnormal power consumption data sample X min The i-th data in the i The remaining power consumption abnormal samples are y i , and calculate the Euclidean distance d(x,y) in the m-dimensional space. The specific expression is as follows: Among them, y i Indicates the removal of x i The remaining power consumption anomaly samples, d(x,y) represents the Euclidean distance in the m-dimensional space, x i Represents abnormal power consumption data sample X min The i-th data in It means to sum up m data and get the distance x i The most recent k X min Sample data in ; S5.2, from the distance x i The most recent k X min Randomly select a sample from the sample data in i Combine to form a new sample. The specific expression is as follows: Among them, x new represents a new synthesized sample, Indicates the distance x i The most recent k X min A sample is randomly selected from the sample data in , δ represents the sampling ratio, and δ∈[0,1].

4. According to claim 1, a method for intelligently identifying abnormal energy consumption of tunnel equipment based on cloud-edge collaboration is characterized in that: The number of hidden layers in the BP neural network model is four, and the activation function in the hidden layer adopts the tanh function, and the activation function of the output layer in the BP neural network model adopts the softmax function.

5. An intelligent identification system for abnormal energy consumption of tunnel equipment based on cloud-edge collaboration, characterized by: It includes a data acquisition module, a storage module, an edge detection module and a communication module, wherein the data acquisition module, the storage module and the edge detection module are all connected by the communication module; The data acquisition module is used to collect electricity consumption data, and send the collected electricity consumption data to the smart cloud data processing center and the edge detection module through the communication module. The smart cloud data processing center stores an LSTM-BP model constructed based on the intelligent identification method of energy consumption anomalies of cloud-edge collaborative tunnel equipment as described in any one of claims 1 to 4, and sends the LSTM-BP model to the edge detection module. The edge detection module reads real-time data for detection, and if an abnormality occurs, it sends alarm data to the tunnel management center. The storage module is used to store data in the acquisition module, the smart cloud data processing center and the edge detection module.

6. According to claim 5, a system for intelligently identifying abnormal energy consumption of tunnel equipment based on cloud-edge collaboration is characterized in that: The data acquisition module collects electricity consumption data including three-phase voltage, three-phase current, three-phase active power and reactive power in tunnel electrical equipment, and is composed of an IM3332 three-phase power quality monitoring meter and an RS485 to TCP / IP module. The edge detection module uses an embedded development board customized by Raspberry Pi5.

Images