Substation inspection robot inspection control method and device, electronic equipment and medium

Abstract

Embodiments of the present disclosure disclose a substation inspection robot inspection control method and device, electronic equipment and a medium. A specific embodiment of the method includes: obtaining satellite remote sensing information, unmanned aerial vehicle inspection information and substation ground collection information, performing flow feature extraction, and then performing fingerprint flow anomaly identification processing to obtain an inspection anomaly identification information set; performing data channel transmission on the substation inspection flow feature information set, and then performing multi-modal data fusion to obtain substation inspection fusion information; performing substation fault detection processing on the substation inspection fusion information to obtain a substation fault information set; generating substation inspection path information, and controlling the substation inspection robot to perform inspection. This embodiment can improve the substation inspection efficiency, improve the accuracy of substation fault detection, reduce substation fault loss and improve the stability of the substation system operation by integrating all-around substation field inspection data of space-air-ground integration.

Description

TECHNICAL FIELD

[0001] Embodiments of the present disclosure relate to the technical field of computer technology, and particularly to a substation inspection robot inspection control method and device, electronic equipment and a medium. BACKGROUND

[0002] With the continuous development of the power system, as the core hub of power transmission and distribution, the safe and stable operation of the substation is directly related to the reliable power supply of the entire power system. At present, the environment of the substation is becoming increasingly complex, and it is facing many challenges such as equipment aging, harsh natural environment, and increasing network attack risks. For the inspection control of the substation inspection robot, the commonly used way is: using an encryption algorithm to encrypt the transmission of the collected substation ground level data and the spatial level data collected by the unmanned aerial vehicle to obtain an encrypted multi-level substation data set. Then, using a traditional end-to-end target detection model, the encrypted multi-level substation data set is decrypted and then subjected to substation fault detection processing to obtain a substation fault information set, and according to the substation fault information set, the inspection robot is controlled to perform substation inspection.

[0003] However, in practice, it is found that when the above method is used to control the inspection of the substation inspection robot, the following technical problems often exist: Since only the ground level and spatial level data of the substation are collected, the overall environmental information of the substation in the sky level and the equipment thermal map are lacking, and the traditional end-to-end target detection model cannot effectively identify the similar substation fault problems in the complex environment of the substation, resulting in missed detection and false detection of the substation fault identification, reducing the fault identification accuracy, in addition, since the amount of data collected is large, the encryption algorithm needs to consume a large amount of resources, resulting in a high load of the substation edge device, reducing the stability of the substation system operation and increasing the substation equipment loss.

[0004] The above information disclosed in this BACKGROUND section is only for the purpose of enhancing the understanding of the background of the present disclosure and, therefore, can include information that does not form the prior art that is already known to those of ordinary skill in the art in this country. SUMMARY

[0005] The summary of the present disclosure is intended to introduce the concepts in a simplified form, which will be described in detail in the specific embodiments section. The summary of the present disclosure is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to be used to limit the scope of the claimed technical solutions.

[0006] Some embodiments of the present disclosure propose a substation inspection robot inspection control method and device, electronic equipment and a medium to solve one or more of the technical problems mentioned in the background section.

[0007] In a first aspect, some embodiments of the present disclosure provide a substation inspection robot inspection control method, comprising: acquiring satellite remote sensing information, unmanned aerial vehicle inspection information and substation ground collection information of a substation space-air-ground integration; controlling a substation edge device set to extract traffic features of the satellite remote sensing information, the unmanned aerial vehicle inspection information and the substation ground collection information, and obtaining a substation inspection traffic feature information set; performing fingerprint traffic anomaly identification processing on the substation inspection traffic feature information set, and obtaining an inspection anomaly identification information set; performing data channel encryption transmission on the substation inspection traffic feature information set according to the inspection anomaly identification information set, and obtaining a substation inspection encrypted information set; performing multi-modal data fusion processing on the substation inspection encrypted information set, and obtaining substation inspection fusion information; performing substation fault detection processing on the substation inspection fusion information, and obtaining a substation fault information set; generating substation inspection path information of a substation inspection robot according to the substation fault information set, and controlling the substation inspection robot to inspect the substation according to the substation inspection path information.

[0008] In a second aspect, some embodiments of the present disclosure provide a substation inspection robot inspection control device, comprising: an acquisition unit configured to acquire satellite remote sensing information, unmanned aerial vehicle inspection information and substation ground collection information of a substation space-air-ground integration; a first control unit configured to control a substation edge device set to extract traffic features of the satellite remote sensing information, the unmanned aerial vehicle inspection information and the substation ground collection information, and obtain a substation inspection traffic feature information set; a fingerprint traffic anomaly identification unit configured to perform fingerprint traffic anomaly identification processing on the substation inspection traffic feature information set, and obtain an inspection anomaly identification information set; a channel encryption transmission unit configured to perform data channel encryption transmission on the substation inspection traffic feature information set according to the inspection anomaly identification information set, and obtain a substation inspection encrypted information set; a multi-modal data fusion unit configured to perform multi-modal data fusion processing on the substation inspection encrypted information set, and obtain substation inspection fusion information; a substation fault detection unit configured to perform substation fault detection processing on the substation inspection fusion information, and obtain a substation fault information set; a second control unit configured to generate substation inspection path information of a substation inspection robot according to the substation fault information set, and control the substation inspection robot to inspect the substation according to the substation inspection path information.

[0009] In a third aspect, some embodiments of the present disclosure provide an electronic device, comprising: one or more processors; a storage device having stored thereon one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement the method as described in any implementation manner of the first aspect.

[0010] In a fourth aspect, some embodiments of the present disclosure provide a computer readable medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements the method as described in any implementation manner of the first aspect.

[0011] The above various embodiments of the present disclosure have the following beneficial effects: the substation inspection robot inspection control method of some embodiments of the present disclosure can improve the substation inspection efficiency and quality, improve the accuracy of substation fault detection, reduce the substation fault loss and improve the stability of the substation system operation by integrating the all-around substation field inspection data of space-ground integration. Specifically, the reason for the high load of the related substation edge device, reducing the stability of the substation system operation and increasing the substation equipment loss is that: because only the data of the ground level and the space level of the unmanned aerial vehicle of the substation are collected, the overall environmental information of the substation in the sky level and the equipment thermal map are lacking, and the traditional end-to-end target detection model cannot effectively identify the similar substation fault problems in the complex environment of the substation, resulting in the problems of missed detection and false detection of substation fault identification, reducing the fault identification accuracy, in addition, because the amount of collected data is large, encryption algorithm is used for encryption, a large amount of resources is consumed, resulting in high load of the substation edge device, reducing the stability of the substation system operation and increasing the substation equipment loss. Based on this, the substation inspection robot inspection control method of some embodiments of the present disclosure can first obtain satellite remote sensing information, unmanned aerial vehicle inspection information and substation ground collection information of the substation space-ground integration. Here, obtaining substation data from multiple aspects of space-ground can improve the comprehensiveness of data acquisition. Secondly, control the substation edge device set to extract the flow characteristics of the above satellite remote sensing information, the above unmanned aerial vehicle inspection information and the above substation ground collection information to obtain a set of substation inspection flow characteristic information. Here, preliminary processing by the edge device can reduce the amount of transmission data and waste of transmission resources, improve the efficiency and quality of flow characteristic extraction. Thirdly, the above substation inspection flow characteristic information set is subjected to fingerprint flow anomaly identification processing to obtain a set of inspection anomaly identification information. Here, anomaly detection by fingerprint information can improve the characteristics of captured power data, improve the accuracy of anomaly identification, and improve the security and reliable transmission of power data. Subsequently, according to the above inspection anomaly identification information set, the above substation inspection flow characteristic information set is subjected to data channel encryption transmission to obtain a set of substation inspection encryption information. Here, the security and concealment of power data transmission can be improved, and the data attacks suffered can be reduced. Then, the above substation inspection encryption information set is subjected to multi-modal data fusion processing to obtain substation inspection fusion information. Here, the perception and integration capability of data from different modalities and different sources can improve the accuracy of subsequent multi-source collaborative fault detection. Then, the above substation inspection fusion information is subjected to substation fault detection processing to obtain a set of substation fault information. Here, due to the comprehensiveness of the fused power data, the detection accuracy of fault detection in complex backgrounds can be improved, and the false detection and missed detection situations can be reduced.Finally, based on the aforementioned substation fault information set, substation inspection path information for the substation inspection robot is generated. Based on this path information, the robot is controlled to inspect the substation. This improves the robot's safety and operational efficiency, enhances substation stability, and reduces substation damage rates. Therefore, this substation inspection robot control method, by integrating comprehensive substation inspection data from air, space, and ground, can improve substation inspection efficiency and quality, increase the accuracy of fault detection, reduce substation fault losses, and enhance the stability of the substation system. Attached Figure Description

[0012] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0013] Figure 1 This is a flowchart of some embodiments of the substation inspection robot inspection control method according to the present disclosure;

[0014] Figure 2 This is a structural schematic diagram of some embodiments of the substation inspection robot inspection control device according to the present disclosure;

[0015] Figure 3 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0016] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0017] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0018] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0019] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0020] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0021] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0022] Figure 1 A flowchart 100 is shown, illustrating some embodiments of a substation inspection robot inspection control method according to the present disclosure. The substation inspection robot inspection control method includes the following steps:

[0023] Step 101: Obtain integrated satellite remote sensing information, UAV inspection information, and ground-based data collection information of the substation.

[0024] In some embodiments, the execution entity (e.g., electronic device) of the above-mentioned substation inspection robot inspection control method can acquire integrated space-air-ground satellite remote sensing information, UAV inspection information, and substation ground-based data collection information of the substation through wired or wireless connections. The integrated space-air-ground substation system can be a substation safety protection and inspection system that integrates the sensing and communication resources of satellite-based, UAV-based, and ground-based sensor systems to construct multi-level collaborative communication resources. The satellite remote sensing information can be substation information collected at the satellite-based level. This satellite remote sensing information may include, but is not limited to, at least one of the following: substation remote sensing images, substation geographical location information, substation geographical environment information, and equipment spectral information. The UAV inspection information can be substation information collected by UAV-based or other aerial platforms. This UAV inspection information may include, but is not limited to, at least one of the following: substation equipment appearance images, equipment infrared thermal imaging images, and substation 3D point cloud data. The substation ground-based data collection information can be substation information collected by ground-based sensor systems. The information collected from the ground at the substation may include, but is not limited to, at least one of the following: electrical data of substation equipment, operating status information of substation equipment, monitoring video of substation, environmental information such as temperature and humidity of substation.

[0025] Step 102: Control the substation edge device set to extract flow characteristics from satellite remote sensing information, UAV inspection information and substation ground acquisition information to obtain the substation inspection flow characteristic information set.

[0026] In some embodiments, the aforementioned executing entity can control a set of substation edge devices to extract flow characteristics from the aforementioned satellite remote sensing information, the aforementioned UAV inspection information, and the aforementioned substation ground-based data acquisition information, thereby obtaining a substation inspection flow characteristic information set. The substation edge devices in the substation edge device set can be edge devices used for receiving and preprocessing satellite remote sensing information, UAV inspection information, and substation ground-based data acquisition information. The substation inspection flow characteristic information in the substation inspection flow characteristic information set can characterize key features of space-based, air-based, and ground-based systems. The substation inspection flow characteristic information set may include, but is not limited to, at least one of the following: data packet size, timing information, and image / video visual information.

[0027] In some optional implementations of certain embodiments, the above-mentioned control substation edge device set, which extracts flow features from the above-mentioned satellite remote sensing information, the above-mentioned UAV inspection information, and the above-mentioned substation ground-collected information to obtain a substation inspection flow feature information set, may include the following steps:

[0028] The first step involves controlling at least one substation edge device corresponding to the aforementioned substation ground-based data acquisition information to perform transmission traffic packet parsing processing on the information, obtaining a parsed ground transmission traffic packet set. The at least one substation edge device can be an edge device of the substation system used to receive and initially process the aforementioned substation ground-based data acquisition information. The transmission traffic packet parsing processing can involve identifying the communication protocol format of the data packets encapsulated from the substation ground-based data acquisition information and splitting them according to corresponding rules to obtain the parsed substation ground-based data acquisition information. This transmission traffic packet parsing processing can be performed using the Wireshark parsing tool.

[0029] The second step involves partitioning the parsed ground transmission traffic packets to obtain ground inspection traffic clusters. Each ground inspection traffic cluster can consist of a predetermined number of data packets (predetermined bytes). Both the predetermined number and the predetermined bytes can be pre-set values. For example, the predetermined number could be 28, and the predetermined bytes could be 128 bytes. In practice, the executing entity can first extract a predetermined number of bytes from each parsed ground transmission traffic packet in the parsed ground transmission traffic packet set, obtaining a set of extracted transmission packets. This extraction can involve directly extracting parsed ground transmission traffic packets of 128 bytes or more, and appending 0x00 to the end of parsed ground transmission traffic packets smaller than 128 bytes until they reach 128 bytes. Then, the extracted transmission packets are partitioned into traffic clusters, with each predetermined number of extracted transmission packets forming a traffic cluster, to obtain the ground inspection traffic clusters. Among them, the aforementioned ground inspection flow cluster can be a 28*128 byte matrix.

[0030] The third step involves extracting the temporal features of the aforementioned ground inspection traffic clusters to obtain a set of ground inspection temporal feature vectors. These vectors can capture the periodicity and burstiness of the ground inspection traffic clusters over time by using their timestamps and interaction sequences. In practice, the executing entity can first input the ground inspection traffic clusters into a data packet temporal feature extraction model to obtain the set. This model can consist of a first convolutional layer, a first max-pooling layer, a second convolutional layer, a second max-pooling layer, a third convolutional layer, a fourth convolutional layer, a fifth convolutional layer, and three fully connected regularization layers. The first convolutional layer can be a convolutional neural network layer with an 11x11 kernel, a stride of 4, 64 channels, and a ReLU (Rectified Linear Unit) activation function. The second convolutional layer can be a convolutional neural network layer with a 5x5 kernel, a stride of 1, 128 channels, and a ReLU activation function. The third, fourth, and fifth convolutional layers can all have a 3x3 kernel, a stride of 1, 256, 256, and 128 channels respectively, and a ReLU activation function. The first and second max-pooling layers can be pooling layers with a 2x2 kernel. The fully connected regularization layers in the three layers can include both fully connected layers and Dropout layers.

[0031] The fourth step involves inputting the aforementioned ground inspection traffic clusters into the traffic cluster association feature extraction model to obtain a set of ground inspection cluster association feature vectors. These ground inspection cluster association feature vectors characterize the logical associations between clusters captured through field associations of different data packets within the ground inspection traffic clusters. Field associations can be source IP (Internet Protocol Address) / destination IP, port correspondences, or matching relationships between control and response commands. The traffic cluster association feature extraction model can be a neural network model that encodes the features of the input ground inspection traffic clusters into the encoder of a Transformer model, calculates the association weights of the ground inspection traffic clusters through eight parallel multi-head attention mechanisms, extracts global association features, and then fuses the features extracted by the encoder with the association weights through an "Add&Norm" layer.

[0032] The fifth step is to perform feature fusion on the above-mentioned ground inspection time sequence feature vector set and the above-mentioned ground inspection cluster correlation feature vector set to obtain the ground inspection fused flow feature vector set, which serves as the substation inspection flow feature information set.

[0033] The sixth step involves controlling at least one substation edge device corresponding to the aforementioned satellite remote sensing information to extract daily-level flow characteristics from the satellite remote sensing information, thereby obtaining a satellite inspection flow characteristic information set, which serves as the substation inspection flow characteristic information set. The specific implementation method can refer to the implementation methods of steps one through five above, and will not be described further here.

[0034] Step 7: Control at least one substation edge device corresponding to the aforementioned drone inspection information to extract air-level flow features from the drone inspection information, obtaining a drone inspection flow feature information set, which serves as the substation inspection flow feature information set. The specific implementation method can refer to the implementation methods of steps 1 to 5 above, and will not be described further here.

[0035] Step 103: Perform fingerprint flow anomaly identification processing on the substation inspection flow characteristic information set to obtain the inspection anomaly identification information set.

[0036] In some embodiments, the aforementioned executing entity can perform fingerprint flow anomaly identification processing on the aforementioned substation inspection flow characteristic information set to obtain an inspection anomaly identification information set. The inspection anomaly identification information in the aforementioned inspection anomaly identification information set can be information used to determine whether data is abnormal. In practice, the aforementioned executing entity can utilize a flow anomaly identification model to perform fingerprint flow anomaly identification processing on the aforementioned substation inspection flow characteristic information set to obtain the inspection anomaly identification information set. The aforementioned flow anomaly identification model can be an XGBoost (eXtreme Gradient Boosting) model.

[0037] In some optional implementations of certain embodiments, the above-mentioned fingerprint flow anomaly identification processing of the substation inspection flow characteristic information set to obtain the inspection anomaly identification information set may include the following steps:

[0038] The first step involves extracting bilateral fingerprints from the aforementioned substation inspection traffic characteristic information set to obtain a substation bilateral fingerprint information set. This bilateral fingerprint information can be obtained from fingerprint identification of both the communication network traffic side and the substation physical side of the aforementioned substation inspection traffic characteristic information set. The substation physical side can refer to information related to the substation system's operational status. In practice, the executing entity can first, based on the aforementioned substation inspection traffic characteristic information set, segment the data packets from satellite remote sensing information, UAV inspection information, and substation ground-collected information to obtain a substation data stream set. This data packet segmentation can be performed according to preset segmentation rules. These preset segmentation rules may include: data packets with the same source IP, destination IP, source port, destination port, and protocol belong to the same stream. Secondly, communication behavior fingerprints are extracted from the aforementioned substation data stream set to obtain a communication-side fingerprint information set. This communication fingerprint information may include: the number of data packets included in the substation data stream, the substation stream length, the average data packet delay, the data packet delay difference, and the number of connections with the same data source address and destination address. Then, using wavelet transform and step function, time-series features of the equipment voltage and current corresponding to the aforementioned substation inspection flow characteristic information set are extracted and discretized to obtain discretized power time-series information sets and discretized current time-series information sets. Next, time-section combinations are performed on the aforementioned discretized power information sets and the aforementioned discretized current information sets to obtain the physical-side fingerprint information set. The aforementioned time sections can be sections formed by data transmitted within the time period corresponding to the substation data stream. Finally, the aforementioned physical-side fingerprint information and the aforementioned communication-side fingerprint information set are standardized and normalized to determine the substation dual-side fingerprint information set.

[0039] The second step involves encoding the flow characteristics of the aforementioned dual-side fingerprint information set of the substation to obtain a hidden feature vector set of inspection flow. The hidden feature vectors in this set represent the linkage information used to determine abnormal flow through the interconnection of physical and communication fingerprint information. For example, if the physical fingerprint information indicates normal operation while the communication fingerprint information indicates abnormal operation, it means that the data transmitted within the corresponding time period of the substation data flow is abnormal. This is because the physical and communication fingerprint information together constitute the dual-side fingerprint information of the substation; that is, when the physical fingerprint information is operating normally, the matching communication fingerprint information will also necessarily meet the periodic flow characteristic and operate normally. The correspondence between the physical and communication fingerprint information ensures that the dual-side fingerprint information of the substation has unique association and is tamper-proof. Through linkage, fault flow in the substation can be avoided to a certain extent, improving the accuracy of anomaly detection and reducing false alarms and missed alarms. The aforementioned flow characteristic encoding can be performed using ELM (Extreme Learning Machine).

[0040] The third step is to determine the mean and standard deviation of the inspection flow for each hidden feature vector in the inspection flow hidden feature vector set, thus obtaining the set of inspection flow mean and the set of inspection flow standard deviation.

[0041] The fourth step involves performing feature sampling on the hidden feature vector set of the inspection traffic flow based on the aforementioned mean and standard deviation sets of the inspection traffic flow, thereby obtaining a latent feature vector set of the inspection traffic flow. The latent feature vectors in this set can be Gaussian-distributed. For example, the execution entity can first map the hidden feature vectors of the inspection traffic flow to the latent space using the encoder in a trained VAE (Variational Auto Encoder). Then, it can randomly sample latent variables from the normal distribution in the latent space using the mean and standard deviation sets of the inspection traffic flow. Finally, it can perform a nonlinear transformation on the latent variables using the ReLU function to obtain the latent feature vector set of the inspection traffic flow.

[0042] The fifth step involves performing feature decoding and reconstruction on the aforementioned latent feature vector set of the inspection traffic to obtain a reconstructed feature vector set of the inspection traffic. The reconstructed feature vectors in this set can be feature vectors generated by the decoder in a trained VAE model, which maps the latent feature vector set of the inspection traffic to the original data space. This original data space can be the space containing the aforementioned hidden feature vector set of the inspection traffic.

[0043] Step 6: Determine the fingerprint traffic feature similarity between each reconstructed feature vector in the above-mentioned inspection traffic reconstructed feature vector set and each hidden feature vector in the above-mentioned inspection traffic hidden feature vector set, thus obtaining a fingerprint traffic feature similarity set. The fingerprint traffic feature similarity can be the Euclidean distance between the reconstructed feature vector and the corresponding hidden feature vector.

[0044] Step 7: Based on the aforementioned fingerprint flow feature similarity set, perform fingerprint flow anomaly identification processing on the aforementioned substation inspection flow feature information set to obtain an inspection anomaly identification information set. In practice, the executing entity can first determine at least one fingerprint flow feature similarity that is not within the preset flow feature similarity range in the aforementioned fingerprint flow feature similarity set as the target fingerprint flow feature similarity set. The preset flow feature similarity range can be a pre-defined range of feature similarity values, which can be determined according to specific circumstances and is not limited here. Then, at least one substation inspection flow feature information corresponding to the aforementioned target fingerprint flow feature similarity set is determined as the inspection anomaly identification information set.

[0045] Step 104: Based on the inspection anomaly identification information set, the substation inspection flow characteristic information set is encrypted and transmitted via the data channel to obtain the substation inspection encrypted information set.

[0046] In some embodiments, the executing entity can perform encrypted data channel transmission of the substation inspection flow characteristic information set based on the aforementioned inspection anomaly identification information set to obtain a substation inspection encrypted information set. The substation inspection encrypted information in the substation inspection encrypted information set can be substation information obtained after encrypting the aforementioned substation inspection flow characteristic information set. As an example, the executing entity can first filter out at least one substation inspection flow characteristic information from the aforementioned substation inspection flow characteristic information that indicates the absence of anomalies. Then, it can perform encrypted data channel transmission of the at least one substation inspection flow characteristic information using the SSL (Secure Sockets Layer) / TLS (Transport Layer Security) protocol to obtain the substation inspection encrypted information set.

[0047] In addressing the technical problems mentioned in the background section, the following challenges often arise: the emergence of random quantum computers and the increase in computing power have led to a surge in attacks on encryption technologies. Electricity data is crucial and fundamental national infrastructure data, making its security a critical issue. A conventional solution to these problems typically involves uniformly encrypting the flow characteristics information from substation inspections and then filling it into the payload of network protocol messages for transmission. However, this conventional solution still suffers from the following issues: the payload of network protocol messages is a key area for anomaly detection, and it compromises the original communication effect of the messages. The amount of data that can be filled into the payload is limited, resulting in a small amount of data transmitted via covert channels. Multiple transmissions are required, leading to low power data security, the possibility of covert channels being exposed, low transmission efficiency and capacity, and increased waste of transmission resources. Considering the shortcomings of conventional solutions and leveraging the advantages and current state of covert channel encryption technology possessed by our company, we have decided to adopt the following solution:

[0048] In some optional implementations of certain embodiments, the process of encrypting and transmitting the substation inspection flow characteristic information set through a data channel based on the aforementioned inspection anomaly identification information set to obtain an encrypted substation inspection information set may include the following steps:

[0049] The first step is to determine the inspection data importance set of at least one substation inspection flow characteristic information that represents the absence of anomalies, selected from the aforementioned substation inspection flow characteristic information. The inspection data importance in this set can characterize the privacy level and importance of the substation inspection flow characteristic information. In practice, the executing entity can determine the inspection data importance set of at least one substation inspection flow characteristic information according to preset data importance rules and an entropy weight algorithm. The preset data importance rules can be data importance assessment information determined through various data importance assessment text information. For example, these various data importance assessment text information may include, but are not limited to, at least one of the following: internal enterprise data importance assessment rules, national power industry data security classification guidelines, and data association value assessment information.

[0050] The second step involves generating a concealed channel key in response to the inspection flow characteristic information of each substation whose inspection data importance is determined to be first, and whose concealed channel type meets the preset concealed channel type conditions. The concealed channel type can characterize the concealment type of the concealed channel for each substation's inspection flow characteristic information. This concealed channel type can include, but is not limited to, at least one of the following: storage-type concealed channel, time-type concealed channel, and sequence-type concealed channel. The storage-type concealed channel can be a channel that uses techniques such as unused fields in communication protocol data packets, file storage locations, or database memory operations to achieve concealed information transmission. The concealed channel key can be a quantum-resistant key used to encrypt and decrypt the inspection flow characteristic information of each substation. The inspection flow characteristic information of each substation with first-importance status can be the most sensitive and important information. As an example, the executing entity can generate the concealed channel key using a quantum key distribution protocol.

[0051] The third step involves encrypting the substation inspection flow characteristic fragment information set obtained by segmenting the substation inspection flow characteristic information into fragments based on the aforementioned concealed channel key, thus obtaining a concealed encrypted inspection information set. As an example, the executing entity can first perform the following determination steps based on the concealed channel key and the initial fragment information located at the initial position in the aforementioned substation inspection flow characteristic fragment information set: First, using a post-quantum algorithm, the initial fragment information is post-quantum encrypted according to the concealed channel key to obtain initial encrypted inspection fragment information. Second, the initial encrypted inspection fragment information is used as the second symmetric encryption key to post-quantum encrypt subsequent inspection flow characteristic fragment information, resulting in subsequent encrypted inspection fragment information. The subsequent inspection flow characteristic fragment information can be fragment information located at positions other than the initial position in the aforementioned substation inspection flow characteristic fragment information. Third, in response to determining that the target inspection flow characteristic fragment information is the fragment information located at the termination position, the initial encrypted inspection fragment information and the subsequent encrypted information are determined as the concealed encrypted inspection information set. Then, in response to determining that the target inspection traffic feature fragment information is not fragment information located at the termination position or the initial position, the target fragment information is determined as the initial fragment information, and the encrypted inspection fragment information corresponding to the previous fragment information of the target fragment information is determined as the covert channel key, so as to perform the determination step again. The aforementioned target fragment information may be fragment information located after the initial fragment information.

[0052] The fourth step involves embedding the aforementioned inspection covert encryption information set and transmission identifier information into the target field set of the communication protocol message to obtain the first inspection encryption information set. The transmission identifier information represents the correlation between the inspection covert encryption information and other inspection covert encryption information during transmission. The communication protocol message can be a data message of a transmission protocol used to transmit data to the receiving end. The target fields in the target field set can be fields in a message used to store encrypted data. The target field set can include, but is not limited to, at least one of the following: the Random field and the SessionID field. It should be noted that since the Random field and Session ID field in the original message store random fields generated by sending a set of files to be transmitted, modifying them will not affect the message's identification and can, to some extent, prevent it from being identified as an abnormal message.

[0053] The fifth step involves performing traffic anomaly detection processing on the aforementioned first inspection encrypted information set to obtain the anomaly detection probability. Based on this probability, the first inspection encrypted information set is then adjusted using a covert channel to obtain the adjusted encrypted information set. In practice, the executing entity can first input the first inspection encrypted information set into a message anomaly detection model trained using a heuristic algorithm to obtain the anomaly detection probability. This message anomaly detection model can be a model with a gated recurrent unit network added after a convolutional neural network. The heuristic algorithm can be one or more heuristic algorithms. Then, in response to determining that the anomaly detection probability is greater than or equal to a preset anomaly probability threshold, the embedded covert encrypted information set is re-determined to be embedded into the communication protocol message. The preset anomaly probability threshold can be a pre-set value used to determine whether an anomaly exists; for example, the preset anomaly probability threshold could be 0.7.

[0054] Step 6: In response to the inspection flow characteristic information of each substation with the second-highest importance of the inspection data, a flow encryption key and an integrity authentication key are generated. Based on these keys, the inspection flow characteristic information of each substation with the second-highest importance is encrypted and authenticated to obtain an encrypted inspection information set. The second-highest importance substation inspection flow characteristic information can be feature information with both medium importance and sensitivity. As an example, the execution entity can first use the HKDF (HMAC-based Key Derivation Function) algorithm to generate the flow encryption key and integrity authentication key. Then, the SM4-CTR algorithm is used to encrypt the inspection flow characteristic information of each substation to obtain the encrypted inspection information set. Finally, the integrity authentication key and the encrypted inspection information set are sent to the receiving end so that the receiving end can use the SM3 algorithm to perform ciphertext authentication on the encrypted inspection information set to ensure that the decrypted information has not been tampered with.

[0055] Step 7: In response to the inspection flow characteristic information of each substation with the inspection data importance level of the third, the inspection flow characteristic information of each substation with the inspection data importance level of the third is asymmetrically encrypted using an asymmetric encryption algorithm to obtain the inspection asymmetric encrypted information set.

[0056] Step 8 involves concatenating the adjusted encrypted information set, the aforementioned inspection encrypted information set, and the aforementioned inspection asymmetric encrypted information set to obtain the substation inspection encrypted information set, which is then sent to the receiving end via a dynamic interface. The dynamic interface can be an interface that maps the current timestamp and corresponding key to a hash function to the receiving port of the receiving end. Sending the information to the receiving end via a dynamic interface reduces the probability of data exposure on fixed ports.

[0057] The above-mentioned technical solution and related content, as an inventive point of this disclosure, solve the aforementioned technical problem: "Since the payload of network protocol messages is the key detection part for message anomaly detection and destroys the original communication effect of the messages, the amount of data that can be filled in the payload is small, the amount of data transmitted through the covert channel is small, and multiple transmissions are required, resulting in low power data security, the possibility of exposure of the covert channel, low transmission efficiency and transmission capacity, and increased waste of transmission resources." If the above factors are solved, the security of power data can be improved, transmission efficiency and transmission capacity can be increased, and the waste of transmission resources can be reduced. To achieve this effect, this disclosure first determines the privacy importance of substation inspection flow characteristic information that is not identified as abnormal, so that different levels of data encryption transmission can be performed according to different privacy and importance levels. Second, using a post-quantum algorithm, non-impact transmission fields are embedded into the covert transmission of each substation inspection flow characteristic information of the first importance level before anomaly detection is performed to determine the possibility of being identified as abnormal data, thereby further improving the covert channel's concealment and the concealment and security of power data. Then, the flow characteristic information of each substation inspection, classified as second-important, is encrypted and authenticated using a method that, compared to the encryption method used for the first-important level, ensures the security of the power data while reducing costs. Next, the characteristic information of the third-important level is encrypted using asymmetric encryption. This more efficient encryption method for less important and sensitive power data ensures data security while reducing costs and improving encryption timeliness. Finally, the characteristic information of all three levels is sent to the receiving end via a dynamic interface, avoiding the probability of power data exposure through fixed ports and improving data security. Therefore, by encrypting power data of different importance levels during transmission, power data security is improved, encryption costs and transmission resource waste are reduced, and transmission efficiency is increased.

[0058] Step 105: Perform multimodal data fusion processing on the encrypted information set of substation inspection to obtain substation inspection fusion information.

[0059] In some embodiments, the aforementioned executing entity may perform multimodal data fusion processing on the aforementioned substation inspection encrypted information set to obtain substation inspection fused information. This substation inspection fused information may be information that fuses data from different modalities included in the substation inspection encrypted information set to improve the comprehensiveness and accuracy of the data.

[0060] In some optional implementations of certain embodiments, the above-mentioned multimodal data fusion processing of the encrypted substation inspection information set to obtain substation inspection fused information may include the following steps:

[0061] The first step is to perform spatiotemporal alignment on the aforementioned encrypted substation inspection information set, resulting in spatiotemporally aligned satellite remote sensing information, spatiotemporally aligned UAV inspection information, and spatiotemporally aligned substation ground-based data collection information. In practice, the implementing entity can first use the WGS84 (World Geodetic System-1984 Coordinate System) coordinate system of the substation GIS (Geographic Information System) as a reference to decrypt the encrypted substation inspection information set and then convert its format to obtain a substation inspection information set in a standard format. Secondly, multimodal data registration is performed on the substation inspection information set in the standard format to obtain registered satellite remote sensing information, registered UAV inspection information, and registered substation ground-based data collection information. The standard format substation inspection information set includes satellite remote sensing information, which is directly correlated with latitude and longitude and converted into geographic coordinates. UAV inspection information utilizes GPS trajectory and camera intrinsic parameters, and is converted into geographic coordinates using the PNP algorithm to eliminate flight attitude deviations. Substation ground-acquired information is converted into geographic coordinates through an established substation equipment-geographic coordinate mapping table. Then, using a multi-scale pyramid model, the converted satellite remote sensing information, UAV inspection information, and substation ground-acquired information are mapped to the same spatial index structure (e.g., a quadtree), resulting in matched satellite remote sensing information, registered UAV inspection information, and registered ground-acquired information. Subsequently, time interpolation and resampling algorithms are used to synchronize the matched satellite remote sensing information, registered UAV inspection information, and registered ground-acquired information to eliminate time differences, resulting in time-synchronized satellite remote sensing information, time-synchronized UAV inspection information, and time-synchronized ground-acquired information. Finally, noise was removed from the time-synchronized satellite remote sensing information, time-synchronized UAV inspection information, and time-synchronized ground acquisition information to obtain the time-space aligned satellite remote sensing information, time-space aligned UAV inspection information, and time-space aligned substation ground acquisition information.

[0062] The second step involves performing cross-modal feature extraction on the spatiotemporally aligned satellite remote sensing information, UAV inspection information, and substation ground-based data acquisition information to obtain satellite inspection feature vectors, UAV inspection feature vectors, and ground-based inspection feature vectors. Specifically, the satellite inspection feature vector can be obtained by using a U-shaped convolutional neural network to extract cross-modal features from the spatiotemporally aligned satellite remote sensing information. The UAV inspection feature vector can be obtained by using a YOLOv8 model to extract cross-modal features from the spatiotemporally aligned UAV inspection information. The ground-based inspection feature vector can be obtained by using a long short-term memory neural network to extract cross-modal features from the spatiotemporally aligned substation ground-based data acquisition information.

[0063] The third step involves multi-level data fusion of the aforementioned satellite inspection feature vectors, UAV inspection feature vectors, and ground inspection feature vectors to obtain fused air-space-ground inspection information. In practice, the executing entity can first assign different spatial weights to the aforementioned satellite inspection feature vectors, UAV inspection feature vectors, and ground inspection feature vectors through an attention mechanism network, obtaining satellite weighted feature vectors, UAV weighted feature vectors, and ground weighted feature vectors. Then, through comparative learning, cross-modal semantic association is performed on the satellite weighted feature vectors, UAV weighted feature vectors, and ground weighted feature vectors to obtain satellite associated feature vectors, UAV associated feature vectors, and ground associated feature vectors. Finally, through a gated fusion network, feature fusion is performed on the satellite associated feature vectors, UAV associated feature vectors, and ground associated feature vectors to obtain fused air-space-ground inspection information.

[0064] The fourth step involves performing data conflict detection on the fused space-air-ground inspection information to obtain multi-level data conflict detection information. This multi-level data conflict detection information can be the result information characterizing whether conflicts exist in the fused space-air-ground inspection information. Data conflict detection can be a three-level conflict detection process involving spatial location alignment verification, attribute feature consistency verification, and time series trend correlation verification.

[0065] The fifth step involves performing data conflict disambiguation processing on the fused air-space-ground inspection information based on the aforementioned multi-level data conflict detection information, thereby obtaining substation inspection fused information. As an example, the executing entity can determine the conflict level information set of the multi-level conflicts based on the determination of the multi-level data conflict detection information. Then, using a confidence-weighted algorithm, the fused air-space-ground inspection information is processed to obtain substation inspection fused information. Specifically, the confidence-weighted algorithm can be an algorithm that, after determining the confidence values ​​of each conflicting air-space-ground inspection data point, uses the inspection data with the highest confidence value as the benchmark for conflict disambiguation processing.

[0066] Step 106: Perform substation fault detection processing on the substation inspection fusion information to obtain a substation fault information set.

[0067] In some embodiments, the aforementioned executing entity may perform substation fault detection processing on the aforementioned substation inspection fusion information to obtain a substation fault information set. The substation fault information in the aforementioned substation fault information set may be location information of fault conditions within the substation.

[0068] In some optional implementations of certain embodiments, the above-mentioned substation fault detection processing of the substation inspection fusion information to obtain a substation fault information set may include the following steps:

[0069] The first step involves performing negative sample feature enhancement processing on the aforementioned substation inspection fusion information to obtain a substation inspection feature enhancement information set. This enhanced information can be derived by enhancing background images and background interference types (e.g., metal reflections, cable shadows, equipment rust) within the fusion information, thereby increasing the information about background image categories. In practice, the implementing entity can utilize generative adversarial networks (GANs) to perform sample enhancement processing on the background regions included in the fusion information, thus obtaining the substation inspection feature enhancement information set. It should be noted that negative sample feature enhancement processing can explicitly introduce background negative samples, enabling the subsequent substation fault detection model to better distinguish interference anchor boxes similar to the fault background, significantly reducing false detections in complex substation backgrounds.

[0070] The second step involves processing the enhanced substation inspection feature information set for fault identification, resulting in a substation fault anchor frame information set. This set includes rectangular frames used to select faulty areas within the substation. The fault identification process utilizes a fault anchor frame localization model. This model is composed of a deep neural network consisting of eight MBConv (Mobile inverted Bottleneck Convolution) modules with SimAM (Similarity-Aware Activation Module) attention modules stacked in series, and a bidirectional feature pyramid network. The bidirectional feature pyramid network is a deep neural network model that fuses features from the multi-scale feature map output by the eight MBConv modules with SimAM attention modules stacked in series. This bidirectional feature pyramid network aggregates features at different scales, enriching the semantic information of the image and enhancing the perception of detailed features in the enhanced substation inspection feature information set.

[0071] The third step involves scaling and filtering the aforementioned substation fault anchor frame information set to obtain a scaled substation fault anchor frame information set. In practice, the executing entity can first filter the substation fault anchor frame information set using confidence level and anchor frame intersection-over-union (IoU) thresholds to obtain a filtered fault anchor frame information set. Specifically, the confidence level and IoU threshold filtering can be used to retain anchor frames with a confidence level greater than or equal to 0.6 and an IoU greater than or equal to 0.3 for deduplication. Then, the filtered fault anchor frame information set is scaled proportionally and edge-filled to obtain a scaled substation fault anchor frame information set. Proportional scaling and edge filling can dynamically adjust the size of the fault area to ensure the integrity of the fault features.

[0072] The fourth step involves inputting the scaled substation fault anchor frame information set and the preset location encoding vector into the multi-scale convolutional fault feature extraction backbone network included in the substation fault detection model. This yields a multi-scale fault identification feature vector set. The substation fault detection model further includes a multi-branch fault attention feature fusion network and a multi-scale fault detection network. The substation fault detection model can be a deep neural network model that identifies the fault type of the fault region from the input scaled substation fault anchor frame information and outputs the results. The multi-scale convolutional fault feature extraction backbone network can be a network model using a residual network for multi-scale feature extraction. For example, the multi-scale convolutional fault feature extraction backbone network can be a ResNet50 model. The multi-scale fault identification feature vector set can include four multi-scale fault identification feature vectors. Each multi-scale fault identification feature vector can be a feature vector output by the ResNet50 model at each stage. The preset location encoding vector can be the pre-defined location information of the fault point within the scaled substation fault anchor frame information.

[0073] The fifth step involves inputting the aforementioned multi-scale fault identification feature vector set into the aforementioned multi-branch fault attention feature fusion network to obtain the substation fault attention feature vector set. The aforementioned multi-branch fault attention feature fusion network can be a deep neural network composed of four layered Transformer models, fusing features at different scales from the input multi-scale fault identification feature vector set. The Transformer model can be a model formed by concatenating an overlapping compression and merging layer, an efficient self-attention layer, and a hybrid feedforward network layer. The aforementioned overlapping compression and merging layer can be a deep neural network that divides the corresponding input multi-scale fault identification feature vectors into overlapping windows, then concatenates the pixels within the divided windows along the channel dimension, performs normalization, and then performs linear projection. The efficient self-attention layer can be a self-attention layer where dimensionality reduction projection is replaced with projection using reduction_ratio. The hybrid feedforward network layer can be a model that performs different types of linear transformations on the feature vector set output by the efficient self-attention layer using a global feedforward neural network and a local feedforward neural network. The global feedforward neural network can be a feedforward neural network with a large receptive field, capable of better capturing global contextual information. Local feedforward neural networks are feedforward neural networks with smaller receptive fields, enabling them to better capture local details. There is a one-to-one correspondence between the substation fault attention feature vector set, the multi-scale fault identification feature vector set, and the four hierarchical Transformer models; that is, the aforementioned substation fault attention feature vectors are obtained by inputting the corresponding multi-scale fault identification feature vectors into the corresponding hierarchical Transformer models.

[0074] Step 6: Input the aforementioned substation fault attention feature vector set into the aforementioned multi-scale fault detection network to obtain the substation fault information set. The multi-scale fault detection network can be a deep neural network that generates and outputs fault location information and fault category from the substation fault attention feature vector set. The multi-scale fault detection network can be a deep neural network composed of six fault detection decoder modules with identical structures but different inputs. These six fault detection decoder modules are divided into three groups based on the different input feature vectors. Each fault detection decoder module group can include two fault detection decoder modules. The input features of the first group of fault detection decoder modules can be feature vectors located at the third and fourth positions in the substation fault attention feature vector set. The input features of the second group of fault detection decoder modules can be feature vectors located at the second and third positions in the substation fault attention feature vector set. The input features of the third group of fault detection decoder modules can be feature vectors located at the first and second positions in the substation fault attention feature vector set. Each fault detection decoder module can be a decoder that includes a feature fusion enhancement module and a query update module. The feature fusion enhancement module in the first fault detection decoder module can be as follows: The fault attention feature vector from the third substation is downsampled and then concatenated with the fault attention feature vector from the fourth substation. This fusion is then input into a 1x1 convolutional layer for channel adjustment to obtain a channel-adjusted feature vector. Next, the channel-adjusted feature vector is input into a max-pooling layer and an average-pooling layer, respectively, and then into a multilayer perceptron for nonlinear learning and mapping to obtain a first feature vector and a second feature vector. Afterward, the first and second feature vectors are summed and input into a ReLU activation function to obtain feature weights. Finally, the feature weights and the channel-adjusted feature vector are multiplied to obtain an enhanced feature vector, which is then multiplied by the value weight matrix and the key weight matrix to obtain the key feature vector and the value feature vector. The query update module in the first fault detection decoder module can input the fault attention feature vector of the fourth substation into a multilayer perceptron to predict the spatial location of the 8 key points of each object query, and then sample the corresponding key point set from the fault attention feature vector of the fourth substation through bilinear interpolation; then, concatenate the feature vectors corresponding to the sampled key point set to obtain the updated query feature vector; at the same time, linearly project the previous query feature vector and input it into the ReLU activation function, and multiply it with the updated query feature vector to obtain the target query feature vector. The key feature vector and value feature vector output by the feature fusion enhancement module are then input into the multi-head self-attention to output a deep neural network of substation fault information set.

[0075] Step 107: Based on the substation fault information set, generate the substation inspection path information for the substation inspection robot, and control the substation inspection robot to inspect the substation based on the substation inspection path information.

[0076] In some embodiments, the aforementioned executing entity can generate substation inspection path information for the substation inspection robot based on the aforementioned substation fault information set, and control the substation inspection robot to inspect the substation based on the substation inspection path information. The aforementioned substation inspection path information can be the path information for the substation inspection robot to reach the location of the aforementioned substation fault information set for fault repair from its own location.

[0077] As an example, the aforementioned execution entity can use a dynamic programming algorithm to generate substation inspection path information for the substation inspection robot based on the aforementioned substation fault information set, and control the aforementioned substation inspection robot to perform inspections according to the aforementioned substation inspection path.

[0078] Furthermore, in the process of adopting technical solutions to address the technical problems mentioned in the background, the following technical issues often arise: Due to the complex environment of substations, it is difficult to accurately locate the position of the substation inspection robot and plan the path for the robot to reach the fault area of ​​the substation. It is also difficult to ensure accurate obstacle avoidance and generate the shortest path to adapt to the complex substation scenario, resulting in a high damage rate and low safety of the substation inspection robot, as well as low stability and high equipment damage rate of the substation system. A conventional solution to these technical problems is to use a fast search random tree algorithm to generate substation inspection paths and control the substation inspection robot to perform inspections according to these paths. However, this conventional solution still has the following problems: Because the fast search random tree algorithm performs a random search throughout the space, it generates some redundant paths, resulting in low planning efficiency and incompatibility with the complex substation scenario with a large number of devices. This leads to low accuracy of the generated paths, low safety of the substation inspection robot, and an inability to promptly detect and resolve faults, resulting in low stability and high equipment damage rate of the substation system. Considering the shortcomings of conventional solutions, and taking into account the advantages and current state of dynamic path planning technology possessed by the inventor's company, we decided to adopt the following solution:

[0079] In some optional implementations of certain embodiments, generating the substation inspection path information for the substation inspection robot based on the aforementioned substation fault information set may include the following steps:

[0080] The first step is to acquire the substation inspection image set captured by the substation inspection robot and the robot's own inspection pose information. The substation inspection images in the image set can be images of the robot's surrounding environment. The inspection pose information can be the robot's location and orientation information.

[0081] As an example, the aforementioned execution entity can use the ORB-SLAM (Oriented fast and Rotated Brief-Simultaneous Localization and Mapping) algorithm to generate the inspection robot's own inspection pose information based on the aforementioned substation inspection image set.

[0082] The second step involves sorting the aforementioned substation fault information set to obtain a substation fault information sequence. This sequence can be obtained by sorting the substation fault information set based on the distance set from the inspection robot, the fault urgency set corresponding to the substation fault information set, and the fault propagation speed set. In practice, the executing entity can utilize the Analytic Hierarchy Process (AHP) to sort the substation fault information set based on the distance set from the inspection robot, the fault urgency set corresponding to the substation fault information set, and the fault propagation speed set, thus obtaining the substation fault information sequence. The fault urgency can be quantified using the AHP to determine the importance of equipment, fault type, and fault safety risk. The fault propagation speed can be represented as the ratio of the number of devices connected to the fault location to the total number of devices in the substation.

[0083] The third step involves generating the following path for the inspection position information and substation fault information set:

[0084] Sub-step 1: Based on the aforementioned inspection pose information, construct a starting inspection location search tree. This starting inspection location search tree can be a random tree obtained by using the location corresponding to the position information included in the inspection pose information as the root node, and sampling within the region from the pose information to the substation fault location corresponding to the substation fault information using a preset sampling strategy. The preset sampling strategy can be a target bias sampling strategy.

[0085] Sub-step 2, based on the root node of the inspection position in the initial inspection position search tree, performs the following generation steps:

[0086] The first sub-step involves determining the starting random inspection position node and the target starting inspection position sampling node of the root inspection position node. The starting random inspection position node can be a subsequent node selected using the position information included in the pose information, based on a target bias sampling strategy. The target starting inspection position sampling node can be the node closest to the starting random inspection position node.

[0087] The second sub-step involves constructing the inspection force information for the substation inspection robot at the sampling node of the target initial inspection position, based on the aforementioned substation power equipment topology distribution map. This inspection force information can include the gravitational force exerted on the substation inspection robot by the resultant force of the gravitational potential field constructed at the location of the substation fault information set and the gravitational potential field constructed at the location of the random node at the initial inspection position; and the direction and magnitude of the resultant force of the repulsive potential field constructed at the location where obstacles hinder the movement of the substation inspection robot. It should be noted that if the substation inspection robot fails to approach the fault point for three consecutive steps (i.e., the distance change is less than or equal to 0.5 meters), it is determined to be trapped in a local minimum. In this case, the strength of the gravitational potential field is temporarily increased (e.g., increased by 1.5 times), while the range of the obstacle's repulsive force is reduced (e.g., from 2 meters to 1 meter), thus breaking the force balance.

[0088] As an example, the aforementioned execution entity can first construct a fault gravitational potential field set, a node gravitational potential field set, and a fault repulsive potential field set at the substation fault information set, the random node at the starting inspection position, and the location of the faulty object in the substation, respectively. Secondly, it can perform force decomposition and synthesis on the fault gravitational potential field set, the node gravitational potential field set, and the fault repulsive potential field set to obtain the vector resultant force information experienced by the substation inspection robot at the sampling node at the target starting inspection position.

[0089] The third sub-step involves determining the subsequent inspection position sampling nodes based on the inspection force information. These subsequent inspection position sampling nodes can be nodes obtained from the next sampling step corresponding to the target's initial inspection position sampling node.

[0090] As an example, the aforementioned execution entity can first determine the force direction of the aforementioned inspection force information as the search sampling direction of the aforementioned initial inspection position search tree. Secondly, it can decompose the aforementioned inspection force information to obtain the horizontal axis force value and the vertical axis force value. Then, it can determine the horizontal axis value by the ratio of the preset sampling step size to the aforementioned horizontal axis force value, and determine the vertical axis value by the product of the preset sampling step size and the aforementioned vertical axis force value. The preset sampling step size can be a pre-set sampling distance value. Finally, based on the aforementioned sampling direction, it can determine the coordinate values ​​corresponding to the aforementioned target initial inspection position sampling node and the aforementioned horizontal and vertical axis values, which will be used as subsequent inspection position sampling nodes. For example, if the aforementioned sampling direction is positive on both the horizontal and vertical axes, the sum of the horizontal coordinate of the aforementioned target initial inspection position sampling node and the aforementioned horizontal axis value will be determined as the horizontal coordinate of the subsequent inspection position sampling node, and the sum of the vertical coordinate of the aforementioned target initial inspection position sampling node and the aforementioned vertical axis value will be determined as the vertical coordinate of the subsequent sampling node.

[0091] The fourth sub-step involves generating local inspection path information for the substation based on the sampling nodes at subsequent inspection locations and the power equipment topology map. The power equipment topology map can represent the spatial geometric distribution of all equipment within the substation. The local inspection path information can be the path from the location of the substation inspection robot to the sampling nodes at subsequent inspection locations.

[0092] As an example, the aforementioned execution entity can first utilize the aforementioned power equipment topology map to perform collision checks on the local path formed from the location of the substation inspection robot to the sampling node of the subsequent inspection location, obtaining a path collision detection result. This collision check can be a check to determine whether the local path passes through an area containing an obstacle. Secondly, in response to determining that the path collision detection result indicates a successful detection, a local inspection path is generated from the location of the inspection pose information to the sampling node of the subsequent inspection location.

[0093] The fifth sub-step involves constructing a termination fault location search tree based on the aforementioned substation fault information sequence, and determining the termination fault location search tree and the fault location root node as the starting search random tree and the inspection location root node, respectively, to repeat the above generation steps and obtain the substation fault inspection path information. The termination fault location search tree can be a search tree constructed with the last position in the substation fault information sequence as its root node.

[0094] As an example, the aforementioned execution entity can first determine whether the parent node of the subsequent inspection location sampling node needs to be updated. This determination can be achieved as follows: First, determine the set of sampling points within a region centered on the subsequent inspection location sampling node and with a preset radius. Second, determine the parent sampling point of each sampling point in the aforementioned sampling point set, obtaining a parent sampling point set. Third, determine the first distance value from each parent sampling point in the aforementioned parent sampling point set to the pose information, obtaining a first distance value set. Furthermore, determine the subsequent sampling node as the parent sampling point of each sampling point in the aforementioned sampling point set, and determine the parent sampling point of the aforementioned sampling point as the second distance value from the subsequent sampling node to the pose information, as the second distance value set. Fourth, in response to determining that there exists a first distance value in the first distance value set that is less than the corresponding second distance value in the second distance value set, update the parent node of the subsequent sampling node, that is, determine the sampling point whose first distance value is less than the second distance value as the parent node of the subsequent sampling node, and sequentially backtrack to the determination of the location nodes included in the pose information. Secondly, in response to the determination that the parent node of the subsequent inspection location sampling node does not need to be updated, a directional search random tree is constructed for the position corresponding to the substation fault information at the termination position in the above substation fault information sequence to obtain the termination fault location search tree, and the termination fault location search tree is determined as the starting search random tree, and the position corresponding to the substation fault information at the termination position is determined as the inspection location root node, so as to execute the above generation steps again, and the obtained subsequent inspection location sampling node is used as the subsequent termination sampling node, and the termination path information from the position corresponding to the substation fault information at the termination position to the subsequent termination sampling node is generated as the substation fault inspection path information.

[0095] Sub-step 3: Generate substation inspection path information based on substation local inspection path information and substation fault inspection path information.

[0096] As an example, the aforementioned execution entity can first check whether the subsequent inspection location sampling node of the substation partial inspection path information and the subsequent inspection location sampling node of the substation fault inspection path information are the same node. If they are, the substation partial inspection path information and the substation fault inspection path information are connected to obtain the substation inspection path information. If not, the above generation steps are executed again.

[0097] The above-mentioned technical solution and related content, as an inventive point of this disclosure, solve the aforementioned technical problem: "Because the fast search random tree algorithm performs random searches throughout the entire space, it generates some redundant paths, resulting in low planning efficiency and unsuitability for complex substation scenarios with a large number of devices. This leads to low accuracy of the generated paths, low safety of the substation inspection robot, inability to promptly detect and resolve faults, and consequently, low stability of the substation system and high equipment damage rate." Solving these factors can improve the safety of the substation inspection robot and the stability of the substation system, while reducing the equipment damage rate. To achieve this, this disclosure firstly determines the inspection pose information of the substation inspection robot itself using a substation inspection image set, thereby improving the accuracy of the robot's positioning. Secondly, by sorting the substation fault information set according to distance, fault urgency, and fault propagation, the impact of different substation fault information sets can be quantified, thus controlling the substation inspection robot to start from the most urgent fault point to reduce equipment damage rate and fault impact range, thereby improving substation stability. Secondly, random trees are constructed by searching the inspection pose information and the fault information at the termination position in the substation fault information sequence. Bidirectional path construction improves path generation efficiency. Subsequently, by establishing gravitational and repulsive potential field functions to control the direction of subsequent sampling nodes, the space for random search is reduced, redundant paths are decreased, and the generation rate is increased. Then, obstacle collision checks are performed on the substation local inspection path information, improving the substation inspection robot's obstacle avoidance ability and generating a safe path with the highest fault repair efficiency. Finally, the substation local inspection path information and substation fault inspection path information generated through iterative processing are concatenated, and the substation inspection robot is controlled to inspect according to the substation inspection path information. This improves the safety of the substation inspection robot, reduces the damage rate, enhances the stability of the substation, and reduces equipment damage.

[0098] Further reference Figure 2 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a substation inspection robot inspection control device, which are similar to... Figure 1 Corresponding to the method embodiments shown, the substation inspection robot inspection control device can be specifically applied to various electronic devices.

[0099] like Figure 2As shown, a substation inspection robot control device 200 includes: an acquisition unit 201, a first control unit 202, a fingerprint flow anomaly identification unit 203, a channel encrypted transmission unit 204, a multimodal data fusion unit 205, a substation fault detection unit 206, and a second control unit 207. The acquisition unit 201 is configured to acquire integrated air-space-ground satellite remote sensing information, UAV inspection information, and substation ground-based data collection information from the substation. The first control unit 202 is configured to control a set of substation edge devices to extract flow features from the aforementioned satellite remote sensing information, UAV inspection information, and substation ground-based data collection information to obtain a substation inspection flow feature information set. The fingerprint flow anomaly identification unit 203 is configured to perform fingerprint flow anomaly identification processing on the substation inspection flow feature information set to obtain an inspection anomaly identification information set. The channel encrypted transmission unit 204 is configured to perform encrypted data channel transmission on the substation inspection flow feature information set based on the inspection anomaly identification information set to obtain a substation inspection encrypted information set. The multimodal data fusion unit 205 is configured to perform multimodal data fusion processing on the aforementioned substation inspection encrypted information set to obtain substation inspection fusion information. The substation fault detection unit 206 is configured to perform substation fault detection processing on the aforementioned substation inspection fusion information to obtain a substation fault information set. The second control unit 207 is configured to generate substation inspection path information for the substation inspection robot based on the aforementioned substation fault information set, and to control the substation inspection robot to perform inspections on the aforementioned substation based on the substation inspection path information.

[0100] It is understandable that the various units described in the substation inspection robot inspection control device 200 are related to the reference. Figure 1 The steps in the described method correspond accordingly. Therefore, the operations, features, and beneficial effects described above for the method also apply to the substation inspection robot inspection control device 200 and the units contained therein, and will not be repeated here.

[0101] The following is for reference. Figure 3 It shows a schematic diagram of the structure of an electronic device (e.g., an electronic device) 300 suitable for implementing some embodiments of the present disclosure. Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0102] like Figure 3As shown, the electronic device 300 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. The RAM 303 also stores various programs and data required for the operation of the electronic device 300. The processing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0103] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 308 including, for example, magnetic tapes, hard disks, etc.; and communication devices 309. Communication device 309 allows electronic device 300 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 An electronic device 300 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 3 Each box shown can represent a device or multiple devices as needed.

[0104] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 309, or installed from storage device 308, or installed from ROM 302. When the computer program is executed by processing device 301, it performs the functions defined in the methods of some embodiments of this disclosure.

[0105] It should be noted that, in some embodiments of this disclosure, the computer-readable medium described above may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0106] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and end-to-end networks (e.g., ad hoc end-to-end networks), as well as any currently known or future-developed networks.

[0107] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to: Steps 101-107.

[0108] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0109] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0110] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor may be described as including an acquisition unit, a first control unit, a fingerprint traffic anomaly identification unit, a channel encrypted transmission unit, a multimodal data fusion unit, a substation fault detection unit, and a second control unit. The names of these units do not necessarily limit the specific unit; for example, the acquisition unit may also be described as "a unit that acquires integrated air-space-ground satellite remote sensing information, UAV inspection information, and substation ground-based data collection information from the substation."

[0111] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0112] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A substation inspection robot inspection control method, comprising: Acquire integrated satellite remote sensing information, UAV inspection information, and ground-based data collection information of substations; The control substation edge device set extracts flow characteristics from the satellite remote sensing information, the UAV inspection information, and the substation ground acquisition information to obtain the substation inspection flow characteristic information set; The process involves performing fingerprint flow anomaly identification processing on the substation inspection flow feature information set to obtain an inspection anomaly identification information set. This includes: extracting bilateral fingerprints from the substation inspection flow feature information set to obtain a substation bilateral fingerprint information set; encoding flow features from the substation bilateral fingerprint information set to obtain an inspection flow hidden feature vector set; determining the inspection flow mean and inspection flow standard deviation for each inspection flow hidden feature vector in the inspection flow hidden feature vector set to obtain an inspection flow mean set and an inspection flow standard deviation set; and, based on the inspection flow mean set and the inspection flow standard deviation set, identifying the inspection flow... The hidden feature vector set is subjected to feature sampling processing to obtain the inspection flow potential feature vector set; the inspection flow potential feature vector set is subjected to feature decoding and reconstruction processing to obtain the inspection flow reconstruction feature vector set; the fingerprint flow feature similarity between each inspection flow reconstruction feature vector in the inspection flow reconstruction feature vector set and each inspection flow hidden feature vector in the inspection flow hidden feature vector set is determined to obtain the fingerprint flow feature similarity set; based on the fingerprint flow feature similarity set, the substation inspection flow feature information set is subjected to fingerprint flow anomaly identification processing to obtain the inspection anomaly identification information set; Based on the inspection anomaly identification information set, the substation inspection flow characteristic information set is encrypted and transmitted through the data channel to obtain the substation inspection encrypted information set. The encrypted information set of substation inspection is subjected to multimodal data fusion processing to obtain substation inspection fused information; The substation inspection fusion information is processed for substation fault detection to obtain a substation fault information set. Based on the substation fault information set, substation inspection path information for the substation inspection robot is generated, and based on the substation inspection path information, the substation inspection robot is controlled to inspect the substation.

2. The method according to claim 1, wherein, The control substation edge device set extracts flow characteristics from the satellite remote sensing information, the UAV inspection information, and the substation ground-based data acquisition information to obtain a substation inspection flow characteristic information set, including: Control at least one substation edge device corresponding to the substation ground acquisition information to perform transmission traffic packet parsing processing on the substation ground acquisition information to obtain a parsed ground transmission traffic packet set; The parsed ground transmission traffic packets are divided into transmission traffic clusters to obtain ground inspection traffic clusters. The ground inspection flow cluster is subjected to flow time series feature extraction processing to obtain a ground inspection time series feature vector set; The ground inspection flow clusters are input into the flow cluster association feature extraction model to obtain the ground inspection cluster association feature vector set; The ground inspection time-series feature vector set and the ground inspection cluster correlation feature vector set are fused to obtain the ground inspection fused flow feature vector set, which serves as the substation inspection flow feature information set. Control at least one substation edge device corresponding to the satellite remote sensing information to extract daily-level flow characteristics from the satellite remote sensing information, and obtain a satellite inspection flow characteristic information set as the substation inspection flow characteristic information set; Control at least one substation edge device corresponding to the UAV inspection information to extract air-level flow features from the UAV inspection information, and obtain a UAV inspection flow feature information set as the substation inspection flow feature information set.

3. The method according to claim 1, wherein, The process of performing multimodal data fusion processing on the encrypted substation inspection information set to obtain substation inspection fusion information includes: Spatiotemporal alignment is performed on the encrypted information set of substation inspection to obtain spatiotemporally aligned satellite remote sensing information, spatiotemporally aligned UAV inspection information, and spatiotemporally aligned substation ground-collected information. Cross-modal feature extraction is performed on the spatiotemporally aligned satellite remote sensing information, the spatiotemporally aligned UAV inspection information, and the spatiotemporally aligned substation ground acquisition information to obtain satellite inspection feature vector, UAV inspection feature vector, and ground inspection feature vector; Multi-level data fusion is performed on the satellite inspection feature vector, the UAV inspection feature vector, and the ground inspection feature vector to obtain fused air-space-ground inspection information. Data conflict detection is performed on the fused air-space-ground inspection information to obtain multi-level data conflict detection information; Based on the multi-level data conflict detection information, the fused air-space-ground inspection information is processed to disambiguate data conflicts, resulting in fused substation inspection information.

4. The method according to claim 1, wherein, The substation inspection fusion information is processed to perform substation fault detection, resulting in a substation fault information set, including: The substation inspection fusion information is subjected to negative sample feature enhancement processing to obtain a substation inspection feature enhancement information set; The substation inspection feature enhancement information set is processed for fault identification to obtain the substation fault anchor frame information set. The substation fault anchor frame information set is subjected to scaling and filtering processing to obtain a scaled substation fault anchor frame information set. The scaled substation fault anchor frame information set and the preset position encoding vector are input into the multi-scale convolutional fault feature extraction backbone network of the substation fault detection model to obtain a multi-scale fault identification feature vector set. The substation fault detection model also includes a multi-branch fault attention feature fusion network and a multi-scale fault detection network. The multi-scale fault identification feature vector set is input into the multi-branch fault attention feature fusion network to obtain the substation fault attention feature vector set; The substation fault attention feature vector set is input into the multi-scale fault detection network to obtain the substation fault information set.

5. A substation inspection robot inspection control device, comprising: The acquisition unit is configured to acquire integrated air-space-ground satellite remote sensing information, UAV inspection information, and ground-based data collection information of the substation. The first control unit is configured to control the substation edge device set and extract flow characteristics from the satellite remote sensing information, the UAV inspection information and the substation ground acquisition information to obtain the substation inspection flow characteristic information set. The fingerprint flow anomaly identification unit is configured to perform fingerprint flow anomaly identification processing on the substation inspection flow feature information set to obtain an inspection anomaly identification information set. This includes: extracting bilateral fingerprints from the substation inspection flow feature information set to obtain a substation bilateral fingerprint information set; encoding flow features from the substation bilateral fingerprint information set to obtain an inspection flow hidden feature vector set; determining the inspection flow mean and inspection flow standard deviation for each inspection flow hidden feature vector in the inspection flow hidden feature vector set to obtain an inspection flow mean set and an inspection flow standard deviation set; and applying the inspection flow mean set and the inspection flow standard deviation set... The system performs feature sampling on the hidden feature vector set of the inspection flow to obtain a potential feature vector set of the inspection flow; it then performs feature decoding and reconstruction on the potential feature vector set of the inspection flow to obtain a reconstructed feature vector set of the inspection flow; it determines the fingerprint flow feature similarity between each reconstructed feature vector in the reconstructed feature vector set of the inspection flow and each hidden feature vector in the hidden feature vector set of the inspection flow to obtain a fingerprint flow feature similarity set; and based on the fingerprint flow feature similarity set, it performs fingerprint flow anomaly identification processing on the substation inspection flow feature information set to obtain an inspection anomaly identification information set. The channel encryption transmission unit is configured to perform data channel encryption transmission on the substation inspection flow characteristic information set according to the inspection anomaly identification information set, so as to obtain the substation inspection encrypted information set. The multimodal data fusion unit is configured to perform multimodal data fusion processing on the encrypted information set of substation inspection to obtain substation inspection fusion information; The substation fault detection unit is configured to perform substation fault detection processing on the substation inspection fusion information to obtain a substation fault information set. The second control unit is configured to generate substation inspection path information for the substation inspection robot based on the substation fault information set, and to control the substation inspection robot to inspect the substation based on the substation inspection path information.

6. An electronic device, comprising: One or more processors; Storage device, on which one or more programs are stored, When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-4.

7. A computer-readable medium having a computer program stored thereon, wherein, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-4.

Images