Superconducting cable safety assessment method and system based on stacked self-encoding network

Abstract

The invention relates to a stacked self-encoding network-based superconducting cable safety assessment method and system. The method comprises the following steps of: obtaining a multi-dimensional actual measurement parameter value sequence of indexes associated with a superconducting cable; based on the actual measurement parameter value sequence, utilizing a pre-trained stacked self-encoding network based on a restricted Boltzmann machine to obtain a reconstruction parameter value sequence; based on the actual measurement parameter value sequence and the reconstruction parameter value sequence, calculating a reconstruction error sequence and carrying out sliding time window processing; dynamically updating a risk threshold value of the superconducting cable based on a sliding time window processing result; and based on the current reconstruction error and the risk threshold value, whether an alarm signal is sent is judged, and superconducting cable safety assessment is realized. Compared with the prior art, the method has the advantages that the influence of multiple factors on the safety of the superconducting cable is fully considered, and the negative influence of a fixed threshold value on safety evaluation is relieved.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a superconducting cable safety assessment method and system based on a stacked autoencoding network. Background Art

[0002] Compared with traditional cables, superconducting cables have advantages such as low loss and large capacity, and their operational safety has a significant impact on the power system. With the development of holographic power systems, the number of non-electrical data acquisition and monitoring devices for superconducting cables has gradually increased, and the non-electrical operating data of superconducting cables has also increased dramatically. These data provide new possibilities for superconducting cable safety assessment, but traditional safety assessment is only based on a single monitoring signal and a single threshold, and the assessment has a delay.

[0003] Chinese patent application publication number CN119001319A proposes a superconducting cable monitoring system and monitoring method. The method monitors the key operating parameters of the superconducting cable based on high-precision sensors and data acquisition technology. However, the method does not associate multi-dimensional monitoring variables together, and only performs safety assessment based on single monitoring data and fixed thresholds, which has certain limitations.

[0004] In summary, there is currently a lack of a superconducting cable safety assessment method and system to solve or partially solve the aforementioned problems. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a superconducting cable safety assessment method and system based on a stacked autoencoder network, so as to solve or partially solve the problems of incomplete assessment considerations, inaccurate safety assessment of a single monitoring parameter, and the impact of delay of a single parameter threshold on the assessment.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] One aspect of the present invention provides a superconducting cable safety assessment method based on a stacked autoencoder network, comprising the following steps:

[0008] Obtaining a multi-dimensional measured parameter value sequence of indicators associated with the superconducting cable;

[0009] Based on the historical measured parameter value sequence, a stacked autoencoder network based on a restricted Boltzmann machine is trained to obtain a stacked autoencoder network model of a superconducting cable, and the parameter value sequence is reconstructed according to the real-time measured parameter value sequence;

[0010] Based on the measured parameter value sequence and the reconstructed parameter value sequence, a reconstruction error sequence is calculated and a sliding time window processing is performed;

[0011] Based on the sliding window processing results, the risk threshold of the superconducting cable is dynamically updated;

[0012] Based on the current reconstruction error and risk threshold, it is determined whether to issue an alarm signal to achieve superconducting cable safety assessment.

[0013] As a preferred technical solution, the multiple indicators associated with the superconducting cable include superconductor cable body operation indicators, cooling system indicators and superconducting cable channel environment indicators, wherein the superconductor cable body operation indicators include one or more of current, liquid level, pressure, flow rate and temperature, the cooling system indicators include one or more of liquid level, pressure, flow rate and temperature, and the superconducting cable channel environment indicators include one or more of vibration signals, sedimentation signals and temperature parameters.

[0014] As a preferred technical solution, after obtaining the measured parameter value sequence, the method further includes:

[0015] Performing missing value filling processing on the measured parameter value sequence;

[0016] Based on the pre-acquired historical parameter values ​​of the superconducting cable under normal operating conditions, the measured parameter value sequence after the missing value filling process is normalized.

[0017] As a preferred technical solution, the stacked autoencoder network based on restricted Boltzmann machine includes:

[0018] The encoding part includes a plurality of restricted Boltzmann machines connected in sequence, and is used for reducing the dimension of the measured parameter value sequence to obtain low-dimensional data;

[0019] The decoding part includes a plurality of restricted Boltzmann machines connected in sequence, which are used to restore the low-dimensional data sequence to obtain reconstruction parameter values.

[0020] As a preferred technical solution, the restricted Boltzmann machine is modeled as:

[0021] When the state of the visible layer is known, the probability of the kth neuron in the hidden layer being activated is:

[0022]

[0023] When the state of the hidden layer is known, the probability of the kth neuron in the visible layer being activated is:

[0024]

[0025] Among them, v and h represent the restricted Boltzmann machine composed of visible layer neurons and hidden layer neurons respectively, θ = [W, a, b] is the parameter of the restricted Boltzmann machine, a = (a 1 ,a 2 ,…,a nv )T and b=(b 1 ,b 2 ,…,b nh ) T , w a,b is the ath row and bth column element of the weight matrix W between the visible layer and the hidden layer, n v 、n h are the number of neurons contained in the visible layer v and the hidden layer h respectively.

[0026] As a preferred technical solution, the training process of the stacked autoencoder network based on the restricted Boltzmann machine includes the following steps:

[0027] Starting from the lowest layer of the network, fix the parameters of the previous layer, and use error backpropagation training to optimize the parameters between the two layers of the current restricted Boltzmann machine.

[0028] As a preferred technical solution, in the reconstruction error sequence, the reconstruction error is calculated using the following formula:

[0029]

[0030] in, Represents the measured parameter value x and the reconstructed parameter value The reconstruction error, x i is the ith measured parameter value, is the i-th reconstruction parameter value, n is the number of parameters, ξ i is the reconstruction error of the i-th parameter.

[0031] As a preferred technical solution, the process of dynamically updating the risk threshold of the superconducting cable based on the sliding window processing result includes the following steps:

[0032] Construct a filter window, calculate the average of the reconstruction errors in the window, and use the average as the data value of the center point of the window. Then the filter window moves backward by one time unit. There is overlap between the windows, and the data in half of the window at both ends of the monitoring time retains the original value.

[0033] The risk threshold of the superconducting cable is dynamically updated according to the reconstruction error within the filtering window.

[0034] As a preferred technical solution, the risk threshold is calculated using the following formula:

[0035]

[0036] Among them, U cd is the risk threshold, is the time window error value, S is the standard deviation of the window error value, h is the moving step length, t α / 2is the α / 2 quantile of the t distribution, t m ,t n are the left and right moments of a time window respectively, e i (t) is the reconstruction error A time window sequence divided.

[0037] Another aspect of the present invention provides a superconducting cable safety assessment system based on a stacked self-encoding network, which is used to implement the aforementioned superconducting cable safety assessment method based on a stacked self-encoding network, and the system includes:

[0038] An index collection module, used to obtain a multi-dimensional measured parameter value sequence of indicators associated with the superconducting cable;

[0039] An indicator reconstruction module is used to train a stacked autoencoder network based on a restricted Boltzmann machine based on a historical measured parameter value sequence, obtain a stacked autoencoder network model of a superconducting cable, and reconstruct a parameter value sequence based on a real-time measured parameter value sequence;

[0040] An indicator error calculation module, used to calculate a reconstruction error sequence and perform sliding time window processing based on the measured parameter value sequence and the reconstruction parameter value sequence;

[0041] A risk threshold dynamic update module is used to dynamically update the risk threshold of the superconducting cable based on the sliding window processing result;

[0042] The alarm module is used to determine whether to issue an alarm signal based on the current reconstruction error and risk threshold to achieve superconducting cable safety assessment.

[0043] Compared with the prior art, the present invention has at least one of the following beneficial effects:

[0044] (1) Fully consider the impact of various factors on the safety of superconducting cables: The superconducting cable model based on the improved stacked autoencoding network of the present invention comprehensively considers the electrical variables, non-electrical variables, cooling system monitoring parameters and weather factors of the superconducting cable, making the safety assessment of the superconducting cable more comprehensive.

[0045] (2) Constructing the relationship between parameters through a stacked autoencoder network based on a restricted Boltzmann machine: The present invention establishes the relationship between multi-dimensional monitoring parameters by constructing a superconducting cable stacked autoencoder network model, thereby avoiding the limitations of security assessment of a single monitoring parameter.

[0046] (3) Alleviating the negative impact of fixed thresholds on safety assessment: The present invention adopts a dynamic threshold method to perform safety assessment on the reconstruction error and its residual. The dynamic threshold changes dynamically according to the working conditions of the superconducting cable, avoiding the delay of a single threshold safety assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of a flow chart of a superconducting cable safety assessment method based on a stacked autoencoder network in an embodiment;

[0048] Figure 2 A schematic diagram of a restricted Boltzmann machine in an embodiment;

[0049] Figure 3 To improve the stacked autoencoder network diagram in the embodiment;

[0050] Figure 4 It is a stacked autoencoder network parameter training diagram in the embodiment;

[0051] Figure 5 It is an equivalent diagram of the stacked autoencoder network in the embodiment;

[0052] Figure 6 is a schematic diagram of a safety assessment system in an embodiment;

[0053] Figure 7 Schematic diagram of an electronic device in an embodiment. DETAILED DESCRIPTION

[0054] 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 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 should fall within the scope of protection of the present invention.

[0055] Example 1

[0056] In response to the problems existing in the aforementioned prior art, this embodiment provides a superconducting cable safety assessment method based on a stacked autoencoder network, which comprehensively considers the electrical and non-electrical variables of the cable system's multiple structures, effectively learns the intrinsic correlation between multi-dimensional data, and combines the dynamic threshold method to perform safety assessment on the superconducting cable system, thereby alleviating the limitations of the single monitoring variable and single threshold safety assessment method for superconducting cables.

[0057] This method integrates electrical variables and non-electrical variables for safety assessment. First, the key equipment of the superconducting cable system and the key monitoring variables that affect the safety of the equipment are selected, and the data of the selected key monitoring variables are preprocessed; then, a superconducting cable stacking autoencoder network model is established based on the impact of each equipment monitoring variable on the cable, and the real-time reconstruction error of the equipment is obtained; next, the dynamic threshold of the reconstruction error is obtained by processing the reconstruction error. If the dynamic threshold is exceeded, a safety alarm is issued to the cable system, realizing real-time safety assessment of the superconducting cable system and effectively improving the timeliness of the superconducting cable safety alarm.

[0058] For details, see Figure 1 , the method comprises the following steps:

[0059] Step S1, determining the superconducting cable system equipment and its influencing factors according to the internal structure of the superconducting cable, and preprocessing the data. Specifically, step S1 may include steps S101-S103.

[0060] Step S101, the superconducting cable system needs to work in a low temperature environment. The basic principle of cable cooling is to absorb system heat through supercooled liquid nitrogen, transfer the absorbed heat to the cooling device, and cool the supercooled liquid nitrogen in the refrigerator and then return it to the cooling system through the reflux pipe to achieve the circulation of supercooled liquid nitrogen, so as to meet the temperature requirements for the superconducting state operation of the cable. In addition to the cable terminals, cable bodies, and cable intermediate joints of the traditional cable system, the superconducting cable also has a unique supercooled liquid nitrogen circulation system. The supercooled liquid nitrogen circulation system includes: liquid nitrogen tanks, pump boxes, refrigerators, cooling towers and cooling units, reflux pipes and other equipment.

[0061] Step S102, select the representative evaluation indicators of the equipment in step S101. There are a large number of safety evaluation indicators for the safety evaluation of superconducting cable systems. In order to ensure the efficiency and simplicity of the model, this embodiment selects key influencing factors as safety evaluation indicators. The causes of superconducting cable failures are divided into two categories. One is the internal operation failure of the cable itself. The internal factors mainly refer to the operating parameters of the superconducting cable body and the cooling system. The operating parameters of the superconducting cable body mainly include: current, liquid level, pressure, flow, and temperature; the operating parameters of the cooling system mainly include: liquid level, pressure, flow, and temperature. The second is the external environment, which refers to the environmental parameters of the superconducting cable channel. This embodiment selects: vibration signals (obtained by intelligent ground nails), settlement signals, and temperature parameters for analysis.

[0062] Step S103, preprocessing the selected multi-dimensional monitoring parameters.

[0063] Since some monitoring parameter data are missing, the mean value before and after the missing value is used to fill in. In order to solve the problem that the value ranges of different monitoring parameters are quite different, the historical monitoring parameter data under normal operation of the superconducting cable are selected for normalization processing, as shown in the following formula:

[0064]

[0065] Where x i The monitoring parameter x 0i The normalized result, x 0i-max 、x 0i-mi The monitoring parameters x are 0i The maximum and minimum values ​​of , a total of n monitoring parameters.

[0066] Step S2, constructing a superconducting cable stacked autoencoding network model. Specifically, step S2 may include steps S201-S203.

[0067] Step S201, constructing a restricted Boltzmann machine.

[0068] The restricted Boltzmann machine consists of two layers of neurons: the visible layer v and the hidden layer h. There is no connection between neurons in each layer, and the neurons in each layer are fully connected. When the state of the visible layer is known, the conditions for the activation of different neurons in the hidden layer are independent of each other. The schematic diagram of the restricted Boltzmann machine model is shown in the figure below: Figure 2 As shown in the figure, the parameters of RBM are θ = [W, a, b], where the biases of the neurons in the visible layer and hidden layer are: a = (a 1 ,a 2 ,…,a nv ) T and b=(b 1 ,b 2 ,…,b nh ) T , the weight matrix W between the visible layer and the hidden layer.

[0069] Establish the connection probability between the two layers of RBM, and the energy function is:

[0070] E(v,h)=-a T vb T hh T Wv 2)

[0071] The joint probability distribution is calculated using the energy formula:

[0072]

[0073] The marginal probability distribution obtained according to the joint probability distribution is:

[0074]

[0075] Use Sigmoid(x)=1 / (1+e -x ) activation function, when the state of the visible layer is known, the probability of the kth neuron in the hidden layer being activated is:

[0076]

[0077] Similarly, it can be concluded that when the hidden layer state is known, the probability of the kth neuron in the visible layer being activated is:

[0078]

[0079] Step S202, constructing an improved stacked autoencoder network.

[0080] The stacked autoencoder network is an unsupervised deep learning network consisting of two parts, encoding and decoding, in a symmetrical structure. It is designed to prevent problems such as gradient diffusion and explosion caused by multi-layer stacking of autoencoders. The traditional stacked autoencoder network consists of multiple layers of autoencoders. The autoencoder network learns the feature representation of the input data through the encoder and decoder.

[0081] In order to prevent problems such as gradient diffusion and gradient explosion that may be caused by multi-layer stacking of autoencoders, this embodiment replaces the multi-layer encoding and decoding units in the stacked autoencoder network with restricted Boltzmann machines. Compared with the autoencoder network, the restricted Boltzmann machine of this application adds back propagation optimization and parameter update on the basis of forward propagation, so that the parameters between every two layers of neurons in the network are optimized. Figure 3 As shown in Figure 1. The improved stacked autoencoder network consists of an encoding part and a decoding part, each of which is composed of multiple restricted Boltzmann machines. The encoding part obtains the key features of the data by reducing the dimension of the input data into a hidden layer representation, and the decoding part restores the low-dimensional data to the original input data. The model training process is: starting from the lowest layer of the network, fixing the parameters of the previous layers, and using error back propagation training to optimize the parameters between the two layers of the current restricted Boltzmann machine. This pre-training process is as follows: Figure 4 As shown in Figure 2, the parameters between multiple layers of neurons obtained through RBM training are applied to the stacked autoencoder network.

[0082] The training of the parameters of each layer of restricted Boltzmann machines is completed and combined, and the superconducting cable stacking autoencoder network model is initially completed. The difference between the input data and the output data of each dimension of the superconducting cable stacking autoencoder network model is called the residual of the monitoring parameter, and the sum of the squares of all residuals is the reconstruction error of the superconducting cable stacking autoencoder network model. The Adam algorithm is used to iterate the training data and continuously adjust the model parameters to minimize the reconstruction error, that is, the reconstruction value of each monitoring parameter is as close to the original input value as possible, thereby achieving effective reconstruction of the data.

[0083] Step S203, using the multi-dimensional monitoring parameters related to the superconducting cable to construct a stacked autoencoder network model. The internal principle can be equivalent to the nonlinear function relationship between the monitoring parameters, such as Figure 5As shown. In the improved stacked autoencoder model, the electrical and non-electrical monitoring variables of multiple devices in the superconducting cable system are simultaneously used as inputs, aiming to associate these variables, predict the target parameters through other multi-dimensional monitoring parameters, and avoid the problem of a single monitoring parameter being abnormal and its monitoring device failing. The parameters obtained by training the superconducting cable stacked autoencoder network model can reflect the characteristics of multi-dimensional data and the relationship between different monitoring parameters. The model training data is the data of the normal operation of the superconducting cable, so the model can reflect the functional relationship and data characteristics between different monitoring variables of the superconducting cable in a non-fault state. When the superconducting cable fails or is about to fail, the functional relationship established by different monitoring parameters is destroyed, and the reconstruction error between the model input data and the output reconstruction data will continue to increase. Therefore, the reconstruction error can be used to conduct real-time safety assessment of the superconducting cable.

[0084] Step S3, based on the stacked autoencoder network model, calculate the reconstruction error, and perform safety assessment on the superconducting cable according to the dynamic threshold of the superconducting cable reconstruction error. Specifically, step S3 includes steps S301-S303.

[0085] Step S301, calculation of reconstruction error.

[0086] Reconstruction error E and univariate reconstruction error ξ i As shown in the following formula:

[0087]

[0088] Where x is the monitoring parameter vector, is the reconstruction vector of the monitoring parameter; x i is the ith monitoring parameter, is the reconstructed value of the i-th monitoring parameter.

[0089] Step S302: Preprocessing of residual.

[0090] Considering that the residuals have certain volatility, the obtained residuals are filtered to enhance the accuracy of the safety assessment. That is, a filtering window is set, the residual data in the window is averaged, and the average is used as the data value of the center point of the window. Then the filtering window is moved back one unit and the process is repeated. To ensure the integrity of the data, there is overlap between each window, and the data in the half window at both ends of the monitoring time retains the original value. When the superconducting cable is abnormal, the residual of one or several monitoring parameters will increase significantly, which will cause the overall reconstruction error of the superconducting cable to gradually increase. However, due to the volatility of the reconstruction error, this paper evaluates the safety status of the superconducting cable by judging the reconstruction error after filtering and its rate of change.

[0091] Step S303, since the operating state of the superconducting cable is affected by the working conditions, the monitoring parameters of each component in different load ranges are quite different, and the normally operating superconducting cable may face different environmental conditions in different time periods. Even if the superconducting cable is operating normally, its operating state may change randomly within a certain range, and it does not mean that the system is abnormal. In this case, the effect of analyzing the operating state of the superconducting cable only by the residuals under the normal historical state of the superconducting cable is not good. Therefore, a sliding time window is used to perform statistical analysis on the residuals, and the upper limit of the confidence interval with a confidence level of 1-α is:

[0092]

[0093] U cd To calculate the threshold, is the residual mean of the time window, S is the standard deviation of the window residual value, h is the moving step length, t α / 2 is the α / 2 quantile of the t distribution. As time goes by, the residuals in each window are statistically analyzed to obtain the dynamic threshold of the superconducting cable. The dynamic threshold can effectively consider the uncertainty and variability of the data and better adapt to the changes in the operating status.

[0094] When the reconstruction error obtained by the model exceeds the threshold, it is determined that the superconducting cable has a tendency to fail and an alarm signal is issued. When an alarm occurs in the superconducting cable, the error ξ between the input monitoring variable and its output reconstruction value is determined one by one. i , in order to determine the specific monitoring parameters that cause the fault, so that corresponding measures can be taken more quickly to avoid major power outages.

[0095] Example 2

[0096] Based on Example 1, see Figure 6 This embodiment provides a superconducting cable safety assessment system based on a stacked autoencoding network, which is used to implement the superconducting cable safety assessment method based on a stacked autoencoding network as described in Example 1. The system includes:

[0097] (1) an index collection module, used to obtain a multi-dimensional measured parameter value sequence of an index associated with a superconducting cable;

[0098] (2) an indicator reconstruction module, which is used to train a stacked autoencoder network based on a restricted Boltzmann machine based on a historical measured parameter value sequence, obtain a stacked autoencoder network model of a superconducting cable, and reconstruct a parameter value sequence based on a real-time measured parameter value sequence;

[0099] (3) an index error calculation module, used to calculate a reconstruction error sequence and perform sliding window processing based on a measured parameter value sequence and a reconstruction parameter value sequence;

[0100] (4) a risk threshold dynamic update module, used to dynamically update the risk threshold of the superconducting cable based on the sliding window processing results;

[0101] (5) An alarm module is used to determine whether to issue an alarm signal based on the current reconstruction error and risk threshold to achieve superconducting cable safety assessment.

[0102] The method and system have the following characteristics:

[0103] (1) The superconducting cable model based on the improved stacked autoencoder network comprehensively considers the electrical variables, non-electrical variables, cooling system monitoring parameters and weather factors of the superconducting cable, making the safety assessment of the superconducting cable more comprehensive.

[0104] (2) The superconducting cable stacking self-encoding network model establishes the relationship between multi-dimensional monitoring parameters, avoiding the limitations of security assessment of a single monitoring parameter.

[0105] (3) A dynamic threshold method is used to conduct safety assessment of the reconstruction error and its residual. The dynamic threshold changes dynamically according to the working conditions of the superconducting cable, thus avoiding the delay of a single threshold safety assessment.

[0106] Example 3

[0107] Based on the foregoing embodiments, this embodiment provides an electronic device, comprising: one or more processors and a memory, wherein the memory stores one or more programs, and the one or more programs include instructions for executing the superconducting cable safety assessment method based on a stacked autoencoding network as described in Example 1.

[0108] like Figure 2 As mentioned above, at the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, and may also include other hardware required for the business. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to achieve the above Figure 1 Of course, in addition to the software implementation, the present invention does not exclude other implementations, such as logic devices or a combination of software and hardware, etc., that is, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0109] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0110] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0111] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A superconducting cable safety assessment method based on a stacked autoencoder network, characterized in that: The steps include: Obtaining a multi-dimensional measured parameter value sequence of indicators associated with the superconducting cable; Based on the historical measured parameter value sequence, a stacked autoencoder network based on a restricted Boltzmann machine is trained to obtain a stacked autoencoder network model of a superconducting cable, and the parameter value sequence is reconstructed according to the real-time measured parameter value sequence; Based on the measured parameter value sequence and the reconstructed parameter value sequence, a reconstruction error sequence is calculated and a sliding time window processing is performed; Based on the sliding window processing results, the risk threshold of the superconducting cable is dynamically updated; Based on the current reconstruction error and risk threshold, it is determined whether to issue an alarm signal to achieve superconducting cable safety assessment.

2. A superconducting cable safety assessment method based on a stacked autoencoder network according to claim 1, characterized in that: The multiple indicators associated with the superconducting cable include superconductor cable body operation indicators, cooling system indicators and superconducting cable channel environment indicators, wherein the superconductor cable body operation indicators include one or more of current, liquid level, pressure, flow rate and temperature, the cooling system indicators include one or more of liquid level, pressure, flow rate and temperature, and the superconducting cable channel environment indicators include one or more of vibration signals, sedimentation signals and temperature parameters.

3. A superconducting cable safety assessment method based on a stacked autoencoder network according to claim 1, characterized in that: After obtaining the measured parameter value sequence, the method further includes: Performing missing value filling processing on the measured parameter value sequence; Based on the pre-acquired historical parameter values ​​of the superconducting cable under normal operating conditions, the measured parameter value sequence after the missing value filling process is normalized.

4. A superconducting cable safety assessment method based on a stacked autoencoder network according to claim 1, characterized in that: The stacked autoencoder network based on restricted Boltzmann machine includes: The encoding part includes a plurality of restricted Boltzmann machines connected in sequence, and is used for reducing the dimension of the measured parameter value sequence to obtain low-dimensional data; The decoding part includes a plurality of restricted Boltzmann machines connected in sequence, which are used to restore the low-dimensional data sequence to obtain reconstruction parameter values.

5. A superconducting cable safety assessment method based on a stacked autoencoder network according to claim 1, characterized in that: The restricted Boltzmann machine is modeled as: When the state of the visible layer is known, the probability of the kth neuron in the hidden layer being activated is: When the state of the hidden layer is known, the probability of the kth neuron in the visible layer being activated is: Among them, v and h represent the restricted Boltzmann machine composed of visible layer neurons and hidden layer neurons respectively, θ = [W, a, b] is the parameter of the restricted Boltzmann machine, a = (a1, a2, …, a nv ) T and b=(b1,b2,…,b nh ) T , w a,b is the ath row and bth column element of the weight matrix W between the visible layer and the hidden layer, n v 、n h are the number of neurons contained in the visible layer v and the hidden layer h respectively.

6. A superconducting cable safety assessment method based on a stacked autoencoder network according to claim 1, characterized in that: The training process of the stacked autoencoder network based on the restricted Boltzmann machine includes the following steps: Starting from the lowest layer of the network, fix the parameters of the previous layer, and use error backpropagation training to optimize the parameters between the two layers of the current restricted Boltzmann machine.

7. A superconducting cable safety assessment method based on a stacked autoencoder network according to claim 1, characterized in that: In the reconstruction error sequence, the reconstruction error is calculated using the following formula: in, Represents the measured parameter value x and the reconstructed parameter value The reconstruction error, x i is the ith measured parameter value, is the i-th reconstruction parameter value, n is the number of parameters, ξ i is the reconstruction error of the i-th parameter.

8. A superconducting cable safety assessment method based on a stacked autoencoder network according to claim 1, characterized in that: Based on the sliding window processing results, the process of dynamically updating the risk threshold of the superconducting cable includes the following steps: Construct a filter window, calculate the average of the reconstruction errors in the window, and use the average as the data value of the center point of the window. Then the filter window moves backward by one time unit. There is overlap between the windows, and the data in half of the window at both ends of the monitoring time retains the original value. The risk threshold of the superconducting cable is dynamically updated according to the reconstruction error within the filtering window.

9. A superconducting cable safety assessment method based on a stacked autoencoder network according to claim 1, characterized in that: The risk threshold is calculated using the following formula: Among them, U cd is the risk threshold, is the time window error value, S is the standard deviation of the window error value, h is the moving step length, t α / 2 is the α / 2 quantile of the t distribution, t m ,t n are the left and right moments of a time window respectively, e i (t) is the reconstruction error A time window sequence divided.

10. A superconducting cable safety assessment system based on a stacked self-encoding network, characterized in that: A method for superconducting cable safety assessment based on a stacked autoencoder network as described in any one of claims 1 to 9, the system comprising: An index collection module, used to obtain a multi-dimensional measured parameter value sequence of indicators associated with the superconducting cable; An indicator reconstruction module is used to train a stacked autoencoder network based on a restricted Boltzmann machine based on a historical measured parameter value sequence, obtain a stacked autoencoder network model of a superconducting cable, and reconstruct a parameter value sequence based on a real-time measured parameter value sequence; An indicator error calculation module, used to calculate a reconstruction error sequence and perform sliding time window processing based on the measured parameter value sequence and the reconstruction parameter value sequence; A risk threshold dynamic update module is used to dynamically update the risk threshold of the superconducting cable based on the sliding window processing result; The alarm module is used to determine whether to issue an alarm signal based on the current reconstruction error and risk threshold to achieve superconducting cable safety assessment.

Images