A transformer substation intelligent sensor cluster system adaptive to isolated network operation environment

Abstract

This invention relates to a substation intelligent sensor cluster system adapted to islanded grid operation environments, belonging to the field of power system monitoring technology. It includes: a physical sensing layer comprising multiple intelligent sensing units deployed on key primary equipment in the substation; an edge intelligence layer with a built-in local processing unit configured to analyze locally collected data in real time, identify the islanded grid's system operation mode, dynamically adjust monitoring strategies, and run a lightweight diagnostic model to output preliminary diagnostic conclusions with confidence; and a cluster collaboration layer connected to each intelligent sensing unit via a multi-mode communication interface to form a decentralized peer-to-peer network, configured to execute a distributed consensus algorithm to achieve multi-source information fusion and high-reliability decision-making, supporting event-driven collaborative actions. The beneficial effects of this invention are: by using a distributed consensus algorithm to achieve multi-source information fusion and high-reliability decision-making, and executing event-driven collaborative actions, adaptive monitoring of islanded grid operation environments is realized.

Description

Technical Field

[0001] This invention belongs to the field of power system monitoring technology, and specifically relates to a substation intelligent sensor cluster system adapted to islanded grid operation environment. Background Technology

[0002] Isolated substations (in remote mining areas, island microgrids, and critical infrastructure) are disconnected from the main power grid. They have low system inertia and experience severe frequency and voltage fluctuations, placing high demands on the speed, adaptability, and reliability of the monitoring and protection system.

[0003] Existing substation monitoring systems mostly adopt a data acquisition-centralized processing architecture: sensors collect raw data, which is then uploaded to the station control layer / back-end master station via a communication network for unified analysis and decision-making. This model has significant drawbacks in isolated network scenarios.

[0004] Response lag: Communication delays lead to slow response to rapid disturbances, making it easy to miss the best control opportunity;

[0005] Poor adaptability: Fixed threshold alarms cannot adapt to drastic load fluctuations, resulting in frequent invalid alarms and missed alarms;

[0006] Weak coordination: Sensors such as electrical quantity, temperature, partial discharge, and video work independently without information exchange, making it difficult to form a comprehensive judgment;

[0007] Communication dependency: When communication between the central node and stable communication is interrupted, the intelligent functions fail and the system has poor robustness. Summary of the Invention

[0008] This invention provides an axial flow fan with noise reduction function to solve the technical problem of real-time adaptive monitoring of isolated network operating environment. It connects with each intelligent sensing unit through a multi-mode communication interface to form a decentralized peer-to-peer network, and is configured to execute a distributed consensus algorithm to achieve multi-source information fusion and high-reliability decision-making. It supports event-driven collaborative actions and realizes adaptive monitoring of isolated network operating environment.

[0009] To achieve the above objectives, the present invention is implemented through the following technical solution:

[0010] A substation intelligent sensor cluster system adapted to isolated grid operation environment, comprising:

[0011] The physical sensing layer comprises multiple intelligent sensing units deployed on key primary equipment in the substation; each intelligent sensing unit includes an electrical quantity monitoring module, an auxiliary status quantity monitoring module, a local processing unit, and a multi-mode communication interface supporting at least two communication methods.

[0012] The edge intelligence layer has a built-in local processing unit configured to analyze locally collected data in real time, identify the system operation mode of isolated networks, dynamically adjust monitoring strategies, and run a lightweight diagnostic model to output preliminary diagnostic conclusions with confidence.

[0013] The cluster collaboration layer connects with each intelligent sensing unit through a multi-mode communication interface to form a decentralized peer-to-peer network. It is configured to execute a distributed consensus algorithm to achieve multi-source information fusion and high-reliability decision-making, supports event-driven collaborative actions, and has autonomous reconstruction and survivability capabilities.

[0014] Optionally, the multimode communication interface supports fiber optic Ethernet and wireless ad hoc networking to ensure communication redundancy and maintain the internal network of the cluster in the event of communication interruption.

[0015] Optionally, the auxiliary state monitoring module includes one or more of the following: temperature sensor, vibration sensor, partial discharge sensor, acoustic fingerprint sensor, visible light image sensor, and infrared image sensor.

[0016] Optionally, the local processing unit incorporates a lightweight AI processing chip, including an NPU, for extracting key fault features and running lightweight diagnostic models locally.

[0017] Optionally, the edge intelligence layer can be configured to automatically switch to high-speed sampling mode and lower the alarm threshold for transient disturbances when an increase in the rate of frequency change is detected.

[0018] Optionally, the distributed consensus algorithm used in the cluster collaboration layer is a simplified variant of the practical Byzantine fault-tolerant algorithm, which achieves consensus on the final diagnostic conclusion within the cluster through multiple rounds of voting and verification.

[0019] Optionally, the cluster collaboration layer supports event-driven collaborative actions: when the cluster reaches a consensus on a certain event, it automatically organizes relevant units to execute preset collaborative tasks, including instructing video monitoring units to capture images of suspected fault points in order to form a multi-dimensional chain of evidence.

[0020] Optionally, the cluster collaboration layer has autonomous reconfiguration capabilities: when a sensor unit fails or some communication links are interrupted, the cluster can automatically detect and reconfigure the network topology, and other units can take over some of the key monitoring responsibilities of the failed unit.

[0021] Optionally, the cluster coordination layer is configured with ultimate security rules: in the extreme case of complete loss of connection with the station control layer, the cluster will negotiate to perform minimum local protection actions, including voting to cut off confirmed critically faulty lines.

[0022] Optionally, after the intelligent sensing unit is powered on, it discovers neighboring nodes through broadcasting and automatically forms a Mesh self-organizing network;

[0023] The edge intelligence layer performs feature extraction and preliminary diagnosis at the data source.

[0024] The cluster collaboration layer broadcasts diagnostic information to the target of interest and runs a consensus algorithm. After multiple rounds of interaction, a consistent cluster decision is formed, triggering collaborative actions between units, and the final decision result is sent to the substation control system.

[0025] The beneficial effects of this invention are:

[0026] 1. The edge intelligence layer of the present invention is configured to analyze locally collected data in real time, identify the system operation mode of the isolated network, dynamically adjust the monitoring strategy, and push the intelligent analysis and decision-making capabilities down to the network edge, so as to realize millisecond-level perception and preliminary autonomous response to the vulnerable conditions of the isolated network, buy valuable time for the core control system, effectively suppress the expansion of the accident, and enhance the overall stability of the isolated network.

[0027] 2. The edge intelligence layer of this invention achieves true edge autonomy, does not rely on continuous instructions from the central node, and can still maintain key monitoring and protection functions through internal cluster negotiation when communication is restricted or even interrupted, significantly enhancing system robustness and survivability.

[0028] 3. The cluster collaboration layer of this invention connects with each intelligent sensing unit through a multi-mode communication interface to form a decentralized peer-to-peer network. It is configured to execute a distributed consensus algorithm to achieve multi-source information fusion and high-reliability decision-making, supports event-driven collaborative actions, and utilizes multi-sensor information fusion and a distributed consensus mechanism. The diagnostic conclusions are based on cross-validation of multiple evidence, which significantly reduces false alarms and false negatives caused by single-point sensor failures, interference, or algorithm limitations, and the output results have high reliability.

[0029] 4. The cluster collaboration layer of this invention possesses high scalability and self-healing capabilities. Its peer-to-peer network architecture and modular design allow for plug-and-play addition of new sensors. The network's self-organizing and self-reconstructing capabilities ensure continuous system operation even when some nodes fail, resulting in high operational flexibility.

[0030] 5. The entire system of this invention empowers intelligent operation and maintenance. The system outputs deeply processed "diagnostic conclusions", "fault evidence chains" and "handling suggestions" instead of massive amounts of raw data, which greatly reduces the data analysis burden of operation and maintenance personnel and helps substations achieve precise and intelligent operation and maintenance. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a schematic diagram of the overall system architecture of the present invention;

[0033] Figure 2 This is a hardware configuration diagram of the intelligent sensing unit of the present invention;

[0034] Figure 3 This is a flowchart of the cluster consensus decision-making process of the present invention. Detailed Implementation

[0035] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0036] Example 1

[0037] like Figure 1 As shown, this embodiment provides a substation intelligent sensor cluster system adapted to an islanded grid operation environment, including: a physical sensing layer, an edge intelligence layer, and a cluster collaboration layer; applied to a 35kV substation operating in an islanded grid environment. A total of 15 intelligent sensing units are deployed on the main primary equipment within the substation, including main transformer units, incoming and outgoing line units, busbar units, and capacitor units.

[0038] The physical sensing layer consists of multiple intelligent sensing units, namely temperature sensors, humidity sensors, vibration sensors, and power quality sensors, deployed on key primary equipment such as transformers, circuit breakers, and busbars. Each intelligent sensing unit is a highly integrated hardware module, which includes at least: an electrical quantity monitoring module for high-precision measurement of voltage, current, frequency, and harmonic parameters; an auxiliary state quantity monitoring module for monitoring temperature, vibration, partial discharge, acoustic signature, and images (visible light / infrared), including temperature sensors, vibration sensors, partial discharge sensors, acoustic signature sensors, visible light image sensors, and infrared image sensors; a local processing unit with a built-in lightweight AI processing chip (such as an NPU); and a multi-mode communication interface supporting at least two communication methods (such as fiber optic Ethernet and wireless self-organizing network) to ensure communication redundancy.

[0039] The edge intelligence layer is a collection of software and algorithms embedded in each intelligent sensing unit and local processing unit; that is, a collection of edge computing gateways and local processing units. Its core function is to endow each sensor node with preliminary intelligence. Specifically, the edge intelligence layer can analyze locally collected data in real time, identify the current operating mode of the isolated network system, such as stable operation, load addition / removal transients, and fault disturbances, and dynamically adjust its sampling frequency, data filtering algorithm, feature extraction focus, and alarm threshold accordingly. For example, when an increase in the rate of change of frequency (ROCOF) is detected, it automatically switches to high-speed sampling mode and lowers the alarm threshold for transient disturbances. At the same time, the edge intelligence layer can extract key fault features (such as waveform distortion and frequency band vibration energy) locally, run a lightweight diagnostic model, and output a preliminary diagnostic conclusion with confidence (e.g., "suspected arc fault, confidence level 70%)".

[0040] The cluster collaboration layer is a logical layer that connects all the aforementioned intelligent sensing units through a multi-mode communication interface. It is a collection of cloud / central servers and data centers, forming a decentralized peer-to-peer (P2P) network.

[0041] The core of this layer lies in a distributed coordination mechanism:

[0042] First, a distributed consensus algorithm (e.g., a simplified variant of the Practical Byzantine Fault-Tolerant (PBFT) algorithm) is used to achieve multi-source information fusion and high-confidence decision-making. When it is necessary to determine the status of a device or area, multiple related sensing units (e.g., electrical quantity monitoring units, temperature monitoring units, and partial discharge monitoring units of the same transformer) exchange their preliminary diagnostic results and confidence levels. After multiple rounds of voting and verification, the cluster reaches a consensus on the final conclusion (e.g., "The transformer has internal mechanical loosening, and the anomaly level is level two"), greatly avoiding misjudgments by single-point sensors.

[0043] Secondly, it supports event-driven collaborative actions. When the cluster reaches a consensus on a certain event (such as "slight grounding of phase B of 10kV bus"), it will automatically organize relevant units to execute preset collaborative tasks. For example, the electrical quantity unit accurately locates the grounding branch, and then instructs the video monitoring unit deployed nearby to automatically rotate the pan-tilt unit to perform high-definition recording and infrared temperature measurement at the suspected fault point, forming a multi-dimensional chain of evidence.

[0044] Finally, the cluster possesses autonomous reconfiguration and survivability. When a sensor unit fails or some communication links are interrupted, the cluster can automatically detect and reconfigure the network topology, with other units taking over some of the critical monitoring responsibilities of the failed unit. In extreme cases where there is a complete loss of connection with the station control layer, the cluster can, according to preset ultimate security rules, negotiate and execute minimum local protection actions (such as voting to disconnect confirmed critically faulty lines) to ensure minimum system security.

[0045] Example 2

[0046] Based on Example 1, such as Figure 2 As shown, each intelligent sensing unit is based on an industrial-grade ARM processor and FPGA, and extends with a high-precision ADC chip for electrical quantity acquisition. It integrates miniature sensors such as infrared thermometer, partial discharge high-frequency CT, and vibration accelerometer. The local processing unit is equipped with the lightweight TensorFlow Lite Micro framework. The communication interface supports both RS-485 / fiber optic conversion access to the station's Ethernet and LoRa wireless self-organizing network.

[0047] After the system is powered on, each unit completes self-organizing networking via LoRa broadcast, forming a cluster. During daily operation, each unit works independently based on the edge intelligence layer algorithm. Taking the main transformer monitoring as an example, its oil temperature sensing unit detects that the temperature rise rate exceeds the steady-state threshold, and then adjusts the sampling interval from 1 minute to 10 seconds, and initially diagnoses that "cooling efficiency may decrease".

[0048] like Figure 3 As shown, when an intermittent arc-to-ground fault occurs on a 10kV feeder, the process is as follows:

[0049] The smart sensing unit (unit A) corresponding to the feeder locally detected an abnormal rise in zero-sequence voltage accompanied by high-frequency current pulses. The edge smart layer initially diagnosed it as "suspected intermittent arc grounding" with a confidence level of 80%, and broadcast this information through the LoRa network.

[0050] Upon receiving the broadcast, other feeder units (units B and C) and bus unit (unit D) on the same busbar segment immediately checked their own zero-sequence current data. Units B and C found no abnormalities, while unit D confirmed that the busbar zero-sequence voltage was abnormal.

[0051] Units A, B, C, and D initiate a rapid consensus protocol. Unit A provides detailed waveform characteristics, Unit D provides evidence of voltage anomalies, and Units B and C provide reverse evidence that they are "normal." After two rounds of voting, the cluster reaches a consensus with 92% confidence: "An intermittent grounding fault has occurred in the bus section, and the probability that the fault point is located on the feeder where Unit A is located is extremely high."

[0052] After consensus is reached, the cluster collaboration layer triggers collaborative actions: Unit A marks the fault recording data as key evidence; Unit D instructs the track inspection robot (as a mobile image sensing unit) deployed indoors in the substation to move to the vicinity of the busbar section and the corresponding switchgear of Unit A to perform infrared thermal imaging scanning and visible light video recording.

[0053] Finally, the cluster packages a complete report containing "consensus diagnostic conclusions," "fault waveform data index," and "infrared thermal image," and uploads it to the station control system via a more reliable fiber optic Ethernet connection. Based on this, the station control system immediately issues an alarm and alerts maintenance personnel to the faulty feeder and potential fault locations.

[0054] The entire process from the occurrence of a fault to the generation of a multi-dimensional diagnostic report is completed within 300 milliseconds, and the decision is made autonomously by the edge cluster, without relying on the analysis cycle of the station control system.

[0055] Example 3

[0056] Based on Examples 1 and 2, in a scenario where communication is partially interrupted, assume that the fiber optic network connecting the station control system is interrupted due to an unexpected construction accident. In this case, the cluster maintains internal communication through a LoRa self-organizing network. If a fault occurs within the station at this time, the cluster can still complete the diagnosis through a consensus mechanism. According to preset rules, when the cluster determines a "permanent metallic short circuit" with a consensus of over 95% confidence and the fault point is clearly defined, it has the right to output a signal through a hard contact to directly trip the corresponding circuit breaker unit (the circuit breaker must have a smart interface), achieving basic local protection and preventing the accident from escalating.

[0057] Example 4

[0058] Based on Embodiment 1, this embodiment provides a working method for a substation intelligent sensor cluster system adapted to an isolated grid operating environment. The specific steps are as follows:

[0059] Step S1: Self-organizing network initialization. After each intelligent sensing unit is powered on, it discovers neighboring nodes through broadcasting, automatically forms a Mesh self-organizing network, and can dynamically elect temporary coordinating nodes (non-fixed master nodes).

[0060] Step S2: Adaptive Monitoring and Local Diagnosis. Each unit works in parallel, with the edge intelligence layer continuously identifying operating conditions and dynamically adjusting its monitoring strategy. Simultaneously, feature extraction and preliminary diagnosis are completed at the data source.

[0061] Step S3: Cluster Consensus Decision. For specific targets of interest, relevant units broadcast diagnostic information through the cluster collaboration layer, run consensus algorithms, and reach a consistent cluster decision after multiple rounds of interaction.

[0062] Step S4: Collaborative Execution and Reporting. Based on the consensus decision, trigger collaborative actions between units (such as data complementarity and joint inspection), and send the final decision result to the substation control system.

[0063] Step S5: Iterative Optimization. The cluster periodically, or with the assistance of the station control system, aggregates operational data and performs incremental learning and parameter optimization on the diagnostic model in the edge intelligence layer to achieve continuous performance evolution.

[0064] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope described in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A substation intelligent sensor cluster system adapted to islanded network operation environment, characterized in that, include: The physical sensing layer comprises multiple intelligent sensing units deployed on key primary equipment in the substation; each of the intelligent sensing units includes an electrical quantity monitoring module, an auxiliary status quantity monitoring module, a local processing unit, and a multi-mode communication interface supporting at least two communication methods. The edge intelligence layer has the local processing unit built in, configured to analyze locally collected data in real time, identify the system operation mode of the isolated network, dynamically adjust the monitoring strategy, and run a lightweight diagnostic model to output a preliminary diagnostic conclusion with confidence. The cluster collaboration layer connects with each intelligent sensing unit through the multi-mode communication interface to form a decentralized peer-to-peer network. It is configured to execute a distributed consensus algorithm to achieve multi-source information fusion and high-reliability decision-making, supports event-driven collaborative actions, and has autonomous reconstruction and survivability capabilities.

2. The intelligent sensor cluster system for substations adapted to isolated grid operation environments according to claim 1, characterized in that, The multimode communication interface supports fiber optic Ethernet and wireless ad hoc networks to ensure communication redundancy and maintain the internal network of the cluster in the event of communication interruption.

3. The substation intelligent sensor cluster system adapted to islanded network operation environment according to claim 1, characterized in that, The auxiliary state monitoring module includes one or more of the following: temperature sensor, vibration sensor, partial discharge sensor, acoustic fingerprint sensor, visible light image sensor, and infrared image sensor.

4. The substation intelligent sensor cluster system adapted to islanded network operation environment according to claim 1, characterized in that, The local processing unit has a built-in lightweight AI processing chip, including an NPU, for extracting key fault features and running lightweight diagnostic models locally.

5. A substation intelligent sensor cluster system adapted to islanded network operation environment according to claim 1, characterized in that, The edge intelligence layer is configured to automatically switch to high-speed sampling mode and lower the alarm threshold for transient disturbances when an increase in the frequency change rate is detected.

6. A substation intelligent sensor cluster system adapted to islanded network operation environment according to claim 1, characterized in that, The distributed consensus algorithm used in the cluster collaboration layer is a simplified variant of the practical Byzantine fault-tolerant algorithm, which achieves consensus on the final diagnostic conclusion within the cluster through multiple rounds of voting and verification.

7. A substation intelligent sensor cluster system adapted to islanded network operation environment according to claim 1, characterized in that, The cluster collaboration layer supports event-driven collaborative actions: when the cluster reaches a consensus on a certain event, it automatically organizes relevant units to execute preset collaborative tasks, including instructing video monitoring units to capture images of suspected fault points in order to form a multi-dimensional chain of evidence.

8. A substation intelligent sensor cluster system adapted to islanded network operation environment according to claim 1, characterized in that, The cluster collaboration layer has autonomous reconfiguration capabilities: when a sensor unit fails or some communication links are interrupted, the cluster can automatically detect and reconfigure the network topology, and other units can take over some of the key monitoring responsibilities of the failed unit.

9. A substation intelligent sensor cluster system adapted to islanded network operation environment according to claim 1, characterized in that, The cluster collaboration layer is configured with ultimate security rules: in the extreme case of complete loss of connection with the station control layer, the cluster will negotiate and perform minimum local protection actions, including voting to cut off confirmed critical fault lines.

10. A substation intelligent sensor cluster system adapted to islanded network operation environment according to claim 1, characterized in that, After being powered on, the intelligent sensing unit discovers neighboring nodes through broadcasting and automatically forms a Mesh self-organizing network. The edge intelligence layer completes feature extraction and preliminary diagnosis at the data source. The cluster collaboration layer broadcasts diagnostic information to the target of interest and runs a consensus algorithm. After multiple rounds of interaction, a consistent cluster decision is formed, triggering collaborative actions between units, and the final decision result is sent to the substation control system.

Images