A method and system for dynamic capacity expansion of high-voltage cables based on multi-parameter fusion and edge-cloud collaboration
By using multi-parameter fusion and edge-cloud collaboration, the problems of resource allocation fixation and communication disconnection in high-voltage cable dynamic capacity expansion technology have been solved, realizing safe capacity expansion control and resource optimization under high load and communication blind spots.
Patent Information
- Application Number
- CN202610956165.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-30
- Publication Date
- 2026-11-10
AI Technical Summary
Existing dynamic capacity expansion technologies for high-voltage cables suffer from problems such as fixed resource allocation, single thermodynamic constraints, and loss of capacity expansion calculation capabilities and protection mechanisms when communication is interrupted. These issues lead to insulation degradation under high loads and the inability to maintain safe cable operation in communication dead zones.
By employing a multi-parameter fusion and edge-cloud collaboration approach, multi-dimensional operational status data is collected synchronously, the sampling frequency is adjusted, insulation degradation parameters are introduced, and a lightweight one-dimensional convolutional neural network model is used to perform local dynamic current carrying capacity prediction when communication is interrupted. Combined with the solution of state machine penalty coefficients, capacity expansion control in offline state is achieved.
It enables on-demand allocation of hardware resources under high load and communication outage conditions, adaptive compression of temperature margin, maintenance of cable transmission potential and protection of insulation medium safety, avoiding waste of hardware resources and safe operation in communication blind spots.
Smart Images

Figure CN123074097A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of cable operation monitoring technology, specifically to a method and system for dynamic capacity expansion of high-voltage cables based on multi-parameter fusion and edge-cloud collaboration. Background Technology
[0002] High-voltage cables are the physical carriers of power transmission and distribution networks. Traditionally, the rated current carrying capacity determined based on static environmental parameters limits the transmission capacity of the line. High-voltage cable dynamic capacity expansion technology collects real-time environmental data and cable operating parameters, substitutes them into thermodynamic equations to dynamically calculate the allowable current carrying capacity of the line, thereby increasing the transmission power. Existing cable dynamic capacity expansion schemes adopt a centralized computing architecture and rely on the cloud master station system to run transient thermal circuit models for numerical solutions.
[0003] Existing dynamic capacity expansion technologies have limitations in practical applications. Most calculation models only use the thermodynamic temperature of the cable as a single constraint boundary, failing to incorporate the operating state of the insulation material and the characteristics of high-frequency partial discharge into the calculation framework. Under high-load capacity expansion operation conditions, the thermal stress inside the insulation medium accelerates material degradation, and simply increasing the output current based on the thermodynamic margin directly leads to insulation breakdown. At the same time, monitoring partial discharge requires the acquisition of high-frequency pulse signals, which limits the computing and storage resources of edge devices. Furthermore, existing hardware devices use a fixed reference sampling frequency and a uniform computing power distribution mechanism, which cannot dynamically adjust the sampling configuration of the underlying sensors according to the degree to which the load approaches the limit. This results in the system consuming excessive computing power resources during low-load normal operation and lacking high-density insulation sampling characteristics during high-risk operation phases approaching the safety threshold.
[0004] Furthermore, centralized diagnostic architectures require real-time connectivity of communication links. High-voltage cables are laid in complex geographical areas such as underground enclosed pipe corridors, and data monitoring networks face data packet loss or physical disconnection. When communication links are interrupted, existing edge devices lack local independent thermal evolution inference algorithms and offline protection mechanisms, and cannot obtain dynamic calculation results from the cloud. At this time, the system passively reverts to static rated current carrying capacity operation or suspends state machine decision logic. This centralized dependency mode makes it impossible for the system to use local real-time data for transient thermal circuit calculations in communication blind spots, and it cannot maintain the power transmission efficiency and hardware operation safety of cables in offline states. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides a method and system for dynamic capacity expansion of high-voltage cables based on multi-parameter fusion and edge-cloud collaboration. This solves the technical problems of existing technologies, such as fixed resource allocation, single thermodynamic constraints, and loss of capacity expansion computing capabilities and protection mechanisms when communication is interrupted.
[0006] To achieve the above objectives, the present invention provides the following technical solution:
[0007] The first aspect of this invention provides a method for dynamic capacity expansion of high-voltage cables based on multi-parameter fusion and edge-cloud collaboration, comprising the following steps: Simultaneously collect multi-dimensional operating status data of the target high-voltage cable. The multi-dimensional operating status data includes at least the cable surface thermodynamic temperature, power frequency load current, and high-frequency partial discharge signal collected by the underlying high-frequency pulse current sensor. Data preprocessing is performed on the multidimensional operating status data, and the heat load approximation rate parameter issued by the data monitoring network is extracted to determine the system computing power scheduling mode. Based on this, the physical sampling frequency of the underlying high-frequency pulse current sensor is adjusted, and the insulation degradation parameter is extracted based on the high-frequency partial discharge signal after the sampling frequency is adjusted. When the data monitoring network is connected, the insulation degradation parameter is introduced as a negative penalty term into the thermodynamic temperature rise equation, the global maximum dynamic current carrying capacity is output, and a new thermal load approximation rate parameter is generated synchronously based on the power frequency load current and returned to the edge side. When the data monitoring network is disconnected, the continuous power frequency load current sequence and the cable surface thermodynamic temperature sequence are extracted as input features. A lightweight one-dimensional convolutional neural network model is called to output the predicted operating temperature. A preset state machine penalty coefficient is introduced to solve the local dynamic current carrying capacity. The global maximum dynamic load capacity or the local dynamic load capacity is used as the safe threshold boundary for system operation. Combined with the new heat load approximation rate parameter, the output power of the underlying power generation equipment is adjusted and the tie switch is driven to perform load transfer action.
[0008] The above method switches the calculation mode according to the communication network status: when the network is connected, the insulation degradation parameter is introduced into the thermodynamic temperature rise equation to correct the dynamic current carrying capacity limit; when the network is disconnected, the edge side lightweight model is called to perform local calculations to maintain the capacity expansion control in the offline state.
[0009] Preferably, the steps of extracting the heat load approximation rate parameter issued by the data monitoring network to determine the system computing power scheduling mode and adjusting the physical sampling frequency according to the system computing power scheduling mode include: The extracted heat load approximation rate parameter is compared with the preset safety judgment threshold in the local non-volatile memory to calculate the approximation deviation. When the approximation deviation is greater than zero, a high-risk collaborative working mode is entered, and a hardware register configuration instruction based on the serial communication protocol is sent to increase the physical sampling frequency of the underlying high-frequency pulse current sensor from the reference frequency to a preset high-frequency threshold, and the priority of computing power resource allocation of the internal logic operation unit is reconstructed.
[0010] The above steps dynamically allocate the underlying hardware computing power based on the degree of thermal load approximation, and control the occupation of computing resources by normalized sampling.
[0011] Preferably, the step of extracting insulation degradation parameters based on the high-frequency partial discharge signal after adjusting the sampling frequency includes: In the high-risk collaborative working mode, the underlying high-frequency pulse current sensor, after increasing the physical sampling frequency, acquires a partial discharge dense sampling sequence, and performs partial discharge phase distribution feature extraction calculation on the partial discharge dense sampling sequence. The extracted high-frequency pulse amplitude and phase power frequency period coordinates are mapped and accumulated in a two-dimensional matrix to output a statistical probability distribution density matrix. Numerical dimensionality reduction is performed based on the statistical probability distribution density matrix to calculate the partial discharge severity index, which is then used as the insulation degradation parameter.
[0012] Preferably, the step of introducing the insulation degradation parameter as a negative penalty term into the thermodynamic temperature rise equation and outputting the global maximum dynamic current carrying capacity includes: The partial discharge severity index, which is the insulation degradation parameter, is introduced as a linear negative penalty term into the dynamic allowable temperature rise margin equation of the cable insulation material. The dynamic allowable temperature rise margin is calculated and output by combining the ultimate physical temperature withstand capability of the target high-voltage cable with the current real-time thermodynamic temperature. The dynamic allowable temperature rise margin is substituted into the comprehensive thermodynamic constraint equation, which includes AC equivalent physical resistance and system comprehensive equivalent physical thermal resistance, and the square root iterative solution is performed to output the global maximum dynamic current carrying capacity.
[0013] The above steps quantify the insulation degradation characteristics into negative constraints for thermodynamic calculations, limiting the allowable temperature rise under damaged insulation conditions.
[0014] Preferably, the step of synchronously calculating and generating a new heat load approximation rate parameter based on the power frequency load current and driving the tie switch to perform load transfer operation includes: Extract the power frequency load current and combine it with the global maximum dynamic current carrying capacity to perform ratio mathematical calculations to generate the new heat load approximation rate parameter; When it is determined that the new heat load approximation rate parameter falls into the emergency load shedding range, the internal active power redistribution algorithm is activated. Based on the node power balance equation of the distribution network topology where the target high-voltage cable is located and the global maximum dynamic current carrying capacity, a load transfer adjustment command carrying the target power reduction amount is generated, driving the tie switch on the substation side to perform hardware load transfer action.
[0015] Preferably, the determination step when the data monitoring network is disconnected includes: Send communication heartbeat packets to the cloud physical address of the data monitoring network, and count the frequency of continuous loss of heartbeat packets according to the preset timeout packet loss threshold; When the packet loss rate exceeds the preset timeout packet loss threshold, an interrupt signal is sent to the edge side and the backup wireless cellular network baseband chip is activated to perform network re-registration. When both primary and backup physical links are determined to be disconnected, the broadcast communication is marked as offline (Boolean), all state machine decision logic that depends on cloud parameters is forcibly suspended and the system switches to offline operation.
[0016] Preferably, the step of extracting the continuous power frequency load current sequence and the cable surface thermodynamic temperature sequence as input features includes: In the offline operation state, the continuous power frequency load current sequence and the cable surface thermodynamic temperature sequence of a predetermined time step number are extracted from the multi-dimensional operation state data; The partial discharge severity index of the last successful interaction cached before the communication disconnection is used as the static compensation dimension. The continuous power frequency load current sequence, the cable surface thermodynamic temperature sequence, and the static compensation dimension are combined to form a three-channel input feature and spliced into a two-dimensional time-series feature input matrix. The internal fixed-point arithmetic unit is invoked to perform a maximum-minimum-extreme value normalization preprocessing operation on the two-dimensional time-series feature input matrix, constraining all elements of the input tensor matrix to a dimensionless numerical space.
[0017] The above steps use the insulation degradation parameters cached before disconnection as static compensation to supplement the feature dimensions required for the model input in the offline state.
[0018] Preferably, the step of calling the output of the lightweight one-dimensional convolutional neural network model to predict the operating temperature includes: The preprocessed two-dimensional temporal feature input matrix is input into a cascaded one-dimensional causal convolutional layer, a max pooling layer, and a fully connected layer. Local temporal dependency features are extracted along the discrete time dimension through the one-dimensional causal convolutional layer, and data leakage at future time points is avoided through the causal padding mechanism. The max-pooling layer performs a local spatial dimension downsampling feature compression operation along the feature channel; The fully connected layer maps the high-dimensional temporal features output by pooling to single-node scalar values, and forward inference outputs the predicted operating temperature of the target high-voltage cable.
[0019] The above steps utilize a causal padding mechanism to maintain the unidirectionality of time-series data processing.
[0020] Preferably, the step of introducing a preset state machine penalty coefficient to solve for the local dynamic current carrying capacity includes: Extract the cable's factory-set maximum withstand temperature and subtract the state machine penalty coefficient to derive the offline physical protection temperature reference. The predicted operating temperature and the offline physical protection temperature reference are combined to perform a sign difference calculation to generate a dynamic temperature rise margin. When the predicted operating temperature exceeds the limit, a high-level Boolean instruction is output through the offline emergency trip control function. The dynamic temperature rise margin is substituted as the target constraint into the locally solidified lightweight transient thermal circuit model to perform an inverse solution that deduces the load current from the upper temperature limit, and the local dynamic current carrying capacity is output.
[0021] The second aspect of this invention provides a high-voltage cable dynamic capacity expansion system based on multi-parameter fusion and edge-cloud collaboration, using the high-voltage cable dynamic capacity expansion method based on multi-parameter fusion and edge-cloud collaboration provided in the first aspect of this invention. The system includes: The data synchronization acquisition module is deployed at the target high-voltage cable node to synchronously acquire multi-dimensional operating status data of the target high-voltage cable. The multi-dimensional operating status data includes at least the cable surface thermodynamic temperature, power frequency load current, and high-frequency partial discharge signal acquired by the underlying high-frequency pulse current sensor. The edge computing module is used to perform data preprocessing on the multi-dimensional operating status data, extract the heat load approximation rate parameter sent by the data monitoring network to determine the system computing power scheduling mode, adjust the physical sampling frequency of the underlying high-frequency pulse current sensor accordingly, and extract insulation degradation parameters based on the high-frequency partial discharge signal after adjusting the sampling frequency; and extract the continuous power frequency load current sequence and the cable surface thermodynamic temperature sequence as input features when the data monitoring network is disconnected, call the lightweight one-dimensional convolutional neural network model to output the predicted operating temperature, and introduce a preset state machine penalty coefficient to solve the local dynamic current carrying capacity. The cloud-based diagnostic capacity expansion module is used to introduce the insulation degradation parameter as a negative penalty term into the thermodynamic temperature rise equation when the data monitoring network is connected, output the global maximum dynamic current carrying capacity and calculate and generate a new heat load approximation rate parameter and return it to the edge computing module. The dual-network heterogeneous communication module is used to connect the edge computing module and the cloud diagnostic capacity expansion module to build a data monitoring network, and broadcasts a communication offline Boolean flag to the local bus when a physical network disconnection is detected. The instruction takeover scheduling module is used to adjust the output power of the underlying power generation equipment and drive the tie switch to perform load transfer actions, using the global maximum dynamic load or the local dynamic load as the safety threshold boundary for system operation.
[0022] This invention provides a method and system for dynamic capacity expansion of high-voltage cables based on multi-parameter fusion and edge-cloud collaboration. It offers the following advantages: 1. This invention determines the system's computing power scheduling mode by extracting the thermal load approximation rate parameter sent from the cloud. When the approximation deviation is greater than zero, the physical sampling frequency of the high-frequency pulse current sensor is increased to the high-frequency threshold and the computing power priority is reconstructed. This mechanism avoids the waste of hardware resources of edge devices during normal low-load operation. At the same time, it actively increases the sampling density of partial discharge signals when the cable approaches the thermal limit condition, realizing the on-demand dynamic allocation of computing power and sampling resources of the underlying devices.
[0023] 2. This invention introduces the partial discharge severity index as a linear negative penalty term into the dynamic allowable temperature rise margin equation of cable insulation material. Combined with real-time thermodynamic temperature iteration to output the global maximum dynamic current carrying capacity, this scheme overcomes the defect of traditional dynamic capacity expansion relying solely on thermodynamic boundaries as a single constraint. It enables the system to adaptively compress the temperature margin when sensing high-frequency insulation degradation characteristics, thus tapping the cable's transmission potential while taking into account the physical safety of the insulation medium.
[0024] 3. When the data monitoring network is detected to be disconnected, this invention uses a lightweight one-dimensional convolutional neural network model to predict the operating temperature by combining the insulation degradation parameters cached before the disconnection with the real-time temperature and current sequence. It also solves the local dynamic current carrying capacity in reverse. This design ensures that when the system faces communication blind spots or physical link interruptions, the edge side still has independent thermal evolution prediction and over-limit tripping protection capabilities, avoiding the cable from being forced to return to static parameter operation due to communication damage, and maintaining the capacity enhancement performance under offline conditions. Attached Figure Description
[0025] Figure 1 This is a schematic diagram of the module structure of the high-voltage cable dynamic capacity expansion system with multi-parameter fusion and edge-cloud collaboration according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the overall process of the high-voltage cable dynamic capacity expansion method with multi-parameter fusion and edge-cloud collaboration according to an embodiment of the present invention. Figure 3 This is a schematic diagram illustrating the process of the data synchronization acquisition module of this invention acquiring and transmitting multi-dimensional operating status data. Figure 4 This is a schematic diagram of sensor sampling configuration and partial discharge feature distribution in an embodiment of the present invention; wherein, (a) shows the sampling frequency jump state of the sensor before and after switching to high-risk collaborative mode; (b) shows the two-dimensional distribution features of partial discharge phase extracted under high-risk collaborative working mode; Figure 5This is a schematic diagram of the communication offline degradation and temperature prediction protection timing in an embodiment of the present invention; wherein, (a) shows the degradation evolution of the state machine of the multi-network heterogeneous communication link and the timing of the offline interruption determination; (b) shows the normalized feature amplitude fluctuation and static compensation trajectory extracted from the edge side in the offline state; (c) shows the execution timing of the advance prediction of the operating temperature exceeding the limit and the triggering of the offline emergency trip control command; Figure 6 This is a schematic diagram of the system execution status assessment, early warning, and load transfer adjustment process according to an embodiment of the present invention; Figure 7 This is a schematic diagram showing the comparison between the dynamic calculated current carrying capacity of the experimental group and the static current limiting of the control group in an embodiment of the present invention. Figure 8 This is a schematic diagram comparing the control physical latency of the system in a network-connected state and a local offline computing state according to an embodiment of the present invention.
[0026] Among them, 10 is the data synchronization acquisition module; 20 is the edge computing module; 30 is the cloud diagnostic capacity expansion module; 40 is the dual-network heterogeneous communication module; and 50 is the instruction takeover and scheduling module. Detailed Implementation
[0027] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0028] Reference Figure 1 This invention provides a high-voltage cable dynamic capacity expansion system based on multi-parameter fusion and edge-cloud collaboration. The system includes: The system includes a data synchronization acquisition module 10, an edge computing module 20, a cloud diagnostic capacity expansion module 30, a dual-network heterogeneous communication module 40, and an instruction takeover and scheduling module 50. The data synchronization acquisition module 10 is deployed at the target high-voltage cable node. The data synchronization acquisition module 10 is equipped with a temperature sensor, a current transformer and a high-frequency pulse current sensor. The data synchronization acquisition module 10 synchronously acquires the thermodynamic parameters, electrical operating parameters and insulation degradation parameters of the target high-voltage cable. The edge computing module 20 is deployed at the cable operation site. The edge computing module 20 is connected to the data synchronization acquisition module 10. The edge computing module 20 has built-in data processing circuits and logic operation units. The edge computing module 20 performs low-level data noise reduction and edge control command calculation. The cloud-based diagnostic and capacity expansion module 30 is deployed on a remote server side. The cloud-based diagnostic and capacity expansion module 30 is equipped with a state assessment and diagnostic unit and a thermodynamic mathematical model solver. The cloud-based diagnostic and capacity expansion module 30 performs global insulation baseline assessment and comprehensive thermodynamic constraint calculation. The dual-network heterogeneous communication module 40 is configured with two physically isolated network link entities. The network link entities include a data monitoring network and a local control bus. The data monitoring network uses broadband communication baseband to connect the edge computing module 20 and the cloud diagnostic capacity expansion module 30. The local control bus uses industrial serial communication hardwired to connect the edge computing module 20 and the instruction takeover and scheduling module 50. The instruction takeover and scheduling module 50 is electrically connected to the underlying power generation equipment and the grid load nodes. The instruction takeover and scheduling module 50 is equipped with a signal source switching controller. The instruction takeover and scheduling module 50 receives current carrying capacity control instructions and outputs load regulation execution signals.
[0029] Reference Figure 2 This invention provides a method for dynamic capacity expansion of high-voltage cables based on multi-parameter fusion and edge-cloud collaboration, comprising the following steps: S100 During the equipment operation cycle, the data synchronization acquisition module 10 acquires the cable surface thermodynamic temperature, ambient temperature and humidity, power frequency load current, transient traveling wave and high frequency partial discharge signal in real time through the underlying sensor array. The data synchronization acquisition module 10 transmits the acquired multi-dimensional operating status data to the edge computing module 20.
[0030] S200, the edge computing module 20 performs timestamp alignment and bandpass filtering on the multi-dimensional operating status data, and performs independent queue caching on the extracted transient traveling waves for auxiliary judgment of fault location under trigger conditions. The edge computing module 20 extracts the heat load approximation rate parameter sent by the cloud diagnostic capacity expansion module 30 through the data monitoring network. The edge computing module 20 judges whether the heat load approximation rate parameter exceeds the internally set threshold. The edge computing module 20 controls the sampling frequency configuration register of the underlying high-frequency pulse current sensor according to the judgment result. The edge computing module 20 reconstructs the data feature extraction priority of the edge logic operation unit.
[0031] S300, when the dual-network heterogeneous communication module 40 monitors that the data monitoring network maintains communication connectivity, the cloud diagnostic capacity expansion module 30 receives the preprocessed data uploaded by the edge computing module 20. The cloud diagnostic capacity expansion module 30 introduces the insulation degradation parameter characterizing the severity of partial discharge as a negative penalty term into the thermodynamic temperature rise equation. The cloud diagnostic capacity expansion module 30 calculates and outputs the global maximum dynamic current carrying capacity. The cloud diagnostic capacity expansion module 30 synchronously calculates and generates a new thermal load approximation rate based on the current measured power frequency load current. The cloud diagnostic capacity expansion module 30 returns the new thermal load approximation rate to the edge computing module 20 along the data monitoring network.
[0032] S400, when the dual-network heterogeneous communication module 40 detects a communication interruption or excessive data packet delay in the data monitoring network, the edge computing module 20 terminates the data upload process and the instruction waiting process. The edge computing module 20 calls its internal lightweight transient thermal circuit model for calculation. The edge computing module 20 introduces a numerically amplified state machine penalty coefficient to derive the offline physical protection temperature reference and generate a dynamic temperature rise margin. The edge computing module 20 outputs the local dynamic current carrying capacity by inversely solving the load current from the upper temperature limit. The edge computing module 20 directly sends the local dynamic current carrying capacity through the physically isolated local control bus. During the connection of the data monitoring network, the command takeover and scheduling module 50 receives the global maximum dynamic load capacity issued by the cloud diagnostic capacity expansion module 30. During the disconnection of the data monitoring network, the command takeover and scheduling module 50 switches the physical information source channel. The command takeover and scheduling module 50 receives the local dynamic load capacity issued by the edge computing module 20 through the local control bus. The command takeover and scheduling module 50 uses the global maximum dynamic load capacity or the local dynamic load capacity as the safety threshold boundary for system operation to adjust the output power of the underlying power generation equipment and drive the tie switch to perform load transfer actions.
[0033] Reference Figure 3 In a specific implementation, step S100 of the present invention includes the following steps: S101, the temperature sensor array in the data synchronization acquisition module 10 is deployed at the joint of the target high-voltage cable and the densely laid section of the cable body. The temperature sensor array is equipped with a distributed optical fiber temperature measurement unit that fits into the outer sheath of the cable and an ambient temperature and humidity probe arranged in the cable trench. During the equipment operation cycle, the temperature sensor array relies on the optical fiber Raman scattering effect and the physical mechanism of heat conduction to obtain the thermodynamic temperature parameters of the surface of the target high-voltage cable and the temperature and humidity parameters of the surrounding laying environment in real time.
[0034] S102, the current transformer assembly configured in the data synchronization acquisition module 10 is snapped onto the main circuit of the cable body, and the high-frequency pulse current sensor is snapped onto the cross-interconnection grounding wire of the target high-voltage cable in a non-contact manner. The current transformer assembly contains a measuring coil and a wideband coil. The current transformer assembly collects the power frequency load current sequence and fault transient traveling wave signal flowing through the cable based on the principle of electromagnetic induction. The high-frequency pulse current sensor synchronously couples the high-frequency transient pulse signal generated when the cable insulation medium undergoes partial discharge. The temperature sensor array, the current transformer assembly, and the high-frequency pulse current sensor convert the acquired multi-dimensional physical sensing parameters into corresponding analog voltage signals for continuous output.
[0035] S103, the data synchronization acquisition module 10 has a built-in high-precision timing chip and a multi-channel analog-to-digital converter. The multi-channel analog-to-digital converter continuously converts analog voltage signals into discrete digital sequences. The high-precision timing chip generates a second pulse trigger signal based on an external satellite or network clock source. The second pulse trigger signal is a unified hardware sampling time reference superimposed with digital sequences of thermodynamic parameters, electrical operating parameters, and insulation degradation parameters. Multi-source heterogeneous data form a set of synchronized sampling sequences with physically aligned features in the time dimension.
[0036] After generating the synchronized sampling sequence set, the data synchronization acquisition module 10 performs frame encapsulation on the multidimensional operating status data with the same time reference and transmits it to the edge computing module 20 along the communication link. The discrete-time representation of the multidimensional operating status dataset output by the data synchronization acquisition module 10 is as follows: ; : This is a multi-dimensional operational status perception data set for the edge side; For discrete time node variables; The time-series acquisition sequence of thermodynamic temperature on the surface of the target high-voltage cable; The time-series data acquisition sequence of ambient temperature around the target high-voltage cable. The power frequency load current timing sequence of the target high-voltage cable is collected. The timing sequence of transient signals from traveling wave faults in the target high-voltage cable sheath is acquired. The timing sequence of high-frequency pulse signals for partial discharge at the target high-voltage cable joint is obtained. Regarding the analog signal sampling and quantization process of multi-channel analog-to-digital converters in hardware structures and the application of network time synchronization protocols for high-precision timing chips, those skilled in the art select commercially available mature integrated circuit components for underlying communication bus connection and driver firmware burning based on the accuracy constraints of actual engineering implementation. The implementation path of basic hardware circuit layout and wiring design and clock distribution alignment mechanism falls within the scope of well-known basic technologies in this field. The detailed implementation methods described in this specification do not further elaborate on the underlying electrical principles and the routine configuration and operation details of registers.
[0037] Reference Figure 4 In a specific implementation, step S200 of the present invention includes the following steps: S201, the digital signal processor built into the edge computing module 20 calls the finite impulse response filter to perform bandpass filtering on the high-frequency partial discharge signal, and performs independent queue buffering on the extracted transient traveling wave for auxiliary judgment of fault location under trigger conditions. The digital signal processor removes environmental white noise and carrier communication interference frequency components according to the preset frequency band stopband attenuation parameter. In this embodiment, the preset frequency band stopband attenuation parameter ranges from 40 decibels to 80 decibels. The edge computing module 20 interpolates and aligns the heterogeneous parameters of different sampling frequencies on the time reference axis according to the unified hardware timestamp carried in the data frame header. The edge computing module 20 outputs a standardized sensing sequence set.
[0038] S202, while performing data preprocessing, the edge computing module 20 continuously monitors the downlink communication port of the data monitoring network. The edge computing module 20 parses the downlink data packets and extracts the heat load approximation rate parameter calculated and fed back by the cloud diagnostic capacity expansion module 30. The edge computing module 20 compares the extracted heat load approximation rate parameter with the preset security judgment threshold in the local non-volatile memory. The state machine logic unit inside the edge computing module 20 determines the system computing power scheduling mode of the current running cycle based on the Boolean value of the numerical comparison result.
[0039] The mathematical expression for the threshold determination condition executed by the state machine logic unit built into the edge computing module 20 is as follows: ; : This is the approximation deviation between the heat load approximation rate and the safety warning threshold. The real-time thermal load approximation rate parameter for the target high-voltage cable; The preset threshold for the underlying hardware thermal load is approaching a warning threshold, and in this embodiment, the preset threshold for the underlying hardware thermal load approaching a warning threshold ranges from 0.85 to 0.90. S203, when the state machine logic unit determines that the approximation deviation is less than or equal to zero, the edge computing module 20 maintains the normal data pass-through mode, the high-frequency pulse current sensor maintains the reference sampling frequency, and the multi-dimensional operating status data is directly uploaded to the cloud diagnostic capacity expansion module 30. When the state machine logic unit determines that the approximation deviation is greater than zero, the edge computing module 20 automatically switches to the high-risk collaborative working mode. The edge computing module 20 sends a hardware register configuration instruction based on the serial communication protocol to the data synchronization acquisition module 10. The hardware register configuration instruction directly rewrites the value in the sampling frequency configuration register address of the underlying analog-to-digital conversion circuit of the high-frequency pulse current sensor. After the hardware register configuration instruction takes effect, the high-frequency pulse current sensor jumps its physical sampling frequency from the reference frequency to a preset high-frequency threshold. The preset high-frequency threshold is calibrated according to the Nyquist sampling theorem and matched with the nanosecond-level width of a typical partial discharge pulse. In this embodiment, the preset high-frequency threshold ranges from 50MHz to 100MHz. The high-frequency pulse current sensor acquires a dense sampling sequence of partial discharge containing high-order harmonic components.
[0040] S204, in the high-risk collaborative working mode, the edge computing module 20 reconstructs the priority of computing resource allocation of the internal logic operation unit, the edge computing module 20 terminates the remote pass-through process of the original waveform of broadband data and releases the corresponding underlying memory queue cache space, the edge computing module 20 calls local tensor calculation to accelerate the verification of the dense sampling sequence of partial discharge and perform partial discharge phase distribution feature extraction calculation, the logic operation unit performs two-dimensional matrix mapping and accumulation of the extracted high-frequency pulse amplitude and phase power frequency period coordinates, and the logic operation unit outputs the statistical probability distribution density matrix of the discharge pulse arranged according to the occurrence frequency in a specific phase interval.
[0041] The logic unit performs numerical dimensionality reduction extraction based on the probability distribution density matrix and calculates the partial discharge severity index: ; : The three-dimensional partial discharge severity index of the target high-voltage cable; The total number of discrete intervals of discharge charge in the phase sequence phase distribution spectrum of partial discharge detection; The total number of power frequency phase discrete intervals in the phase sequence phase distribution spectrum of partial discharge detection; This is the step index number for the discrete interval of the discharge charge; This is the step index number for the discrete interval of the power frequency phase; For the corresponding to the first The nominal value of the center charge of each discrete interval of discharge charge; For falling in the first Statistical frequency of discharge pulses within the two-dimensional grid at the intersection of the j-th discrete interval of discharge charge and the j-th discrete interval of power frequency phase; To be assigned to the The discharge hazard penalty weighting coefficient is set for the discrete interval of the power frequency phase, and in this embodiment, the value range of the discharge hazard penalty weighting coefficient is 1.0-3.0.
[0042] After extracting the partial discharge severity index from the three-dimensional map, the edge computing module 20 inserts the partial discharge severity index as a core insulation characterization parameter into the high-priority data frame header and transmits it back to the cloud diagnostic capacity expansion module 30 separately through the data monitoring network.
[0043] For the derivation process of the finite impulse response filtering algorithm called by the digital signal processor and the physical addressing operation of the underlying register of the serial communication protocol, those skilled in the art can consult the general integrated circuit firmware development guide to write the underlying driver code and map the address. The specific mathematical derivation of the signal difference equation and the timing logic of the underlying control bus are well-known technologies in the field of embedded system development. The basic code execution logic will not be elaborated in the specific implementation description of this specification.
[0044] In a specific implementation, step S300 of the present invention includes the following steps: S301, when the dual-network heterogeneous communication module 40 monitors that the data monitoring network maintains communication connectivity, the cloud diagnostic capacity expansion module 30 receives the standardized sensing sequence set or three-dimensional map partial discharge severity index uploaded by the edge computing module 20. The state assessment and diagnosis unit inside the cloud diagnostic capacity expansion module 30 extracts the absolute value of the partial discharge amplitude in the standardized sensing sequence set. The state assessment and diagnosis unit performs a hard numerical comparison between the absolute value of the partial discharge amplitude and the critical discharge threshold in the dielectric strength extreme value database. In this embodiment, the critical discharge threshold ranges from 500pC to 1000pC. The state assessment and diagnosis unit intercepts malicious discharge pulses that exceed the physical limits of material insulation tolerance. The state assessment and diagnosis unit outputs an immediate circuit breaker trip command. Within the stable operating range where the extreme value exceeds the limit, the cloud diagnostic capacity expansion module 30 calls the thermodynamic mathematical model solver to calculate the AC resistance of the inner conductor at the current operating temperature and the total system thermal resistance mapped according to the physical dimensions of the cable based on the real-time acquired thermodynamic temperature parameters.
[0045] S302, the edge computing module 20 of the status assessment and diagnosis unit independently calculates the reported three-dimensional partial discharge severity index in the high-risk collaborative working mode. The cloud diagnosis capacity expansion module 30 introduces the partial discharge severity index as a linear negative penalty term into the dynamic allowable temperature rise margin equation of the cable insulation material. The thermodynamic mathematical model solver combines the highest tolerance temperature threshold specified on the cable material's nameplate with the current real-time thermodynamic temperature to calculate and output the dynamic allowable temperature rise margin.
[0046] The built-in formula for calculating the dynamic allowable temperature rise margin in the thermodynamic mathematical model solver is as follows: ; : The dynamic allowable temperature rise margin of high-voltage cables after the introduction of a partial discharge penalty mechanism; The ultimate physical temperature that the insulation material of the target high-voltage cable is rated to withstand at the factory. The current real-time thermodynamic temperature scalar of the target high-voltage cable surface; The insulation degradation penalty conversion factor is issued from the cloud, and in this embodiment, the value range of the insulation degradation penalty conversion factor is 0.05-0.20; The partial discharge severity index of the target high-voltage cable is represented by a three-dimensional image.
[0047] S303, the thermodynamic mathematical model solver substitutes the dynamic allowable temperature rise margin into the comprehensive thermodynamic constraint equation for square root extraction. The solution process simultaneously references the AC resistance of the inner conductor and the total thermal resistance of the system. If the calculated dynamic allowable temperature rise margin is less than or equal to zero, the thermodynamic mathematical model solver forces the output of a zero current carrying capacity. The thermodynamic mathematical model solver triggers hardware defense lockout. When the dynamic allowable temperature rise margin is greater than zero, the cloud diagnostic capacity expansion module 30 outputs the global maximum dynamic current carrying capacity of the target high-voltage cable in the next scheduling cycle through closed-loop mathematical calculation.
[0048] The specific expression for the comprehensive thermodynamic constraint equation is as follows: ; : The cloud-based diagnostic capacity expansion module outputs the global maximum dynamic current carrying capacity based on the thermoelectric coupling model; The dynamic allowable temperature rise margin of high-voltage cables after the introduction of a partial discharge penalty mechanism; The AC equivalent physical resistance of the target high-voltage cable conductor at the current operating temperature; The comprehensive equivalent physical thermal resistance of the high-voltage cable insulation layer to the external environment.
[0049] S304, the cloud-based diagnostic capacity expansion module 30 extracts the power frequency load current recorded in the standardized sensing sequence set, performs ratio mathematical calculations in combination with the global maximum dynamic current carrying capacity, generates a new thermal load approximation rate parameter based on the physical ratio calculation results, encapsulates and reassembles the new thermal load approximation rate parameter using a communication protocol, and returns the new thermal load approximation rate parameter to the edge computing module 20 along the downlink physical broadband link of the data monitoring network.
[0050] The formula for generating the new heat load approximation rate parameter is: ; : The real-time thermal load approximation rate parameter for the target high-voltage cable; This is a scalar representation of the current power frequency load current collected in real time at the edge side. The cloud-based diagnostic capacity expansion module outputs the global maximum dynamic current carrying capacity based on the thermoelectric coupling model.
[0051] Regarding the hard-coded implementation logic for the comparison of conventional dielectric values within the condition assessment and diagnosis unit, and the basic derivation process for the calculation of the system's comprehensive thermal resistance based on the physical layer thickness, those skilled in the art can refer to the current international standard calculation guidance document for cable steady-state current carrying capacity published by the International Electrotechnical Commission to substitute conventional material constants and expand the formula. The temperature correlation calibration of the basic thermodynamic transfer conversion rules and static physical resistivity is within the scope of known technology in this field. The details of the general mathematical expansion calculation are not elaborated in the specific implementation description of this specification.
[0052] Reference Figure 5 In a specific implementation, step S400 of the present invention includes the following steps: S401, the dual-network heterogeneous communication module 40 continuously sends communication heartbeat packets with fixed time intervals to the cloud physical address of the data monitoring network. The dual-network heterogeneous communication module 40 waits to receive the response verification frame returned by the cloud diagnostic capacity expansion module 30. The link state machine inside the dual-network heterogeneous communication module 40 counts the continuous loss frequency of heartbeat packets according to the preset timeout packet loss threshold. The preset timeout packet loss threshold is pre-fixed and calibrated in the underlying firmware according to the normal channel delay and normal network jitter extreme value of the backbone optical fiber network. In this embodiment, the preset timeout packet loss threshold ranges from three to five consecutive heartbeat packets lost. When the packet loss rate of the backbone optical fiber Ethernet communication link exceeds the preset timeout packet loss threshold, the dual-network heterogeneous communication module 40 sends an interrupt signal to the edge computing module 20 through the underlying universal asynchronous transceiver interface. The dual-network heterogeneous communication module 40 synchronously cuts off the power supply circuit of the optical transceiver at the hardware radio frequency front end. The dual-network heterogeneous communication module 40 activates the backup wireless cellular network baseband chip to perform network re-registration and authentication network access operations.
[0053] S402, when the dual-network heterogeneous communication module 40 determines that both the primary and backup physical links have encountered unrecoverable hardware interruptions or channel blockages, the dual-network heterogeneous communication module 40 broadcasts a communication offline Boolean flag to the local bus where the edge computing module 20 is located. After the daemon process inside the edge computing module 20 captures the communication offline Boolean flag, it forcibly suspends all state machine decision logic that depends on cloud parameters. The edge computing module 20 loads a lightweight one-dimensional convolutional neural network model dedicated to offline state from non-volatile flash memory into the static random access memory of the tensor acceleration operation core. The lightweight one-dimensional convolutional neural network model combines the current change and hysteresis thermal response characteristics within the local time window to perform nonlinear fitting calculations.
[0054] S403, the edge computing module 20 extracts the continuous power frequency load current sequence and the cable surface thermodynamic temperature sequence of the number of time steps prior to the current hardware clock node. The edge computing module 20 uses the partial discharge severity index of the three-dimensional spectrum cached before the communication disconnection as the static compensation dimension. The continuous power frequency load current sequence, the cable surface thermodynamic temperature sequence, and the static compensation dimension together constitute a three-channel input feature. In this embodiment, the value range of the number of time steps prior to the current node is 64-256 discrete time nodes. The edge computing module 20 splices the three-channel input features to form a two-dimensional time-series feature input matrix with a dimension size of the number of time steps multiplied by the number of physical feature channels. The edge computing module 20 calls the internal fixed-point number arithmetic unit to perform a maximum-minimum-extreme value normalization preprocessing operation on the two-dimensional time-series feature input matrix. The maximum-minimum-extreme value normalization preprocessing operation eliminates the distribution differences of different physical dimensions on the numerical scale. The maximum-minimum-extreme value normalization preprocessing operation converges and constrains all pixel elements of the input tensor matrix to a dimensionless numerical space of 0 to 1.
[0055] The S404 lightweight one-dimensional convolutional neural network model cascades three one-dimensional causal convolutional layers, one max pooling layer, and one fully connected layer in its internal network topology. The one-dimensional causal convolutional layers slide along the discrete time dimension to extract local temporal dependence features of load current and thermodynamic temperature. The one-dimensional causal convolutional layers are controlled by a causal filling mechanism to avoid data leakage at future time points. The max pooling layer performs downsampling feature compression operations along the feature channels in the local spatial dimension. The fully connected layer maps the high-dimensional temporal features output by pooling to single-node scalar values corresponding to physical business meanings. The lightweight one-dimensional convolutional neural network model outputs the predicted operating temperature of the target high-voltage cable after a preset time window in the future.
[0056] For the initial construction process of the lightweight one-dimensional convolutional neural network model, the cloud diagnostic capacity expansion module 30 continuously collects normal temperature and flow synchronous acquisition sequences during the historical normal operation phase as the model training sample set. During the model training phase, the cloud diagnostic capacity expansion module 30 extracts the real cable physical temperature readings after the corresponding preset time step as the target label for supervised learning. The cloud diagnostic capacity expansion module 30 uses mean square error as the loss function to evaluate the deviation between the network's forward prediction value and the real temperature label. The cloud diagnostic capacity expansion module 30 calls the adaptive moment estimation optimization algorithm to update the backpropagation gradient of the network's internal weight parameters. The cloud diagnostic capacity expansion module 30 compresses the floating-point network weight matrix after training convergence into an eight-bit integer data format using a static quantization algorithm. The cloud diagnostic capacity expansion module 30 pre-downloads and solidifies the eight-bit integer data format into the local storage medium of the edge computing module 20.
[0057] S405, the edge computing module 20 extracts the predicted operating temperature output by the forward inference of the lightweight one-dimensional convolutional neural network model, the edge computing module 20 extracts the cable's factory limit tolerance temperature and deducts the state machine penalty coefficient, and derives the offline physical protection temperature reference. In this embodiment, the value range of the state machine penalty coefficient is 5℃-10℃. The edge computing module 20 combines the predicted operating temperature with the offline physical protection temperature benchmark to perform a sign difference calculation to generate a dynamic temperature rise margin. Edge computing module 20 uses the dynamic temperature rise margin as the target constraint and substitutes it into its locally embedded lightweight transient thermal circuit model to perform an inverse solution that deduces the load current from the upper temperature limit. Edge computing module 20 calculates and outputs the local dynamic current carrying capacity, which is then sent to the instruction takeover scheduling module 50 via the local control bus. Simultaneously, when the predicted operating temperature exceeds the limit, edge computing module 20 constructs an offline emergency trip control function based on the sign difference calculation result. ; In the formula: Boolean output instruction for offline emergency trip control logic; For standard mathematical symbol functions; The predicted operating temperature is used to reduce the forward inference output of a lightweight one-dimensional convolutional neural network model. This is an offline physical protection temperature reference that is embedded inside the edge computing module 20.
[0058] For the physical addressing operations of the underlying registers of the Universal Asynchronous Transceiver Interface (UART) and the gradient descent derivation steps of the adaptive moment estimation optimization algorithm, those skilled in the art can refer to the general microcontroller underlying development manual or the official open-source documentation of mainstream deep learning frameworks to write conventional driver code and configure hyperparameters. The basic communication hardware interrupt response logic and the chain rule for backpropagation of neural networks are well-known technologies in the fields of deep learning and embedded systems. The standardized underlying firmware calling mechanism and basic mathematical derivation process will not be elaborated in the specific implementation descriptions in this specification.
[0059] Reference Figure 6 In a specific implementation, step S500 of the present invention includes the following steps: S501, the cloud-based diagnostic capacity expansion module 30 extracts new thermal load approximation rate parameters. The cloud-based diagnostic capacity expansion module 30 inputs the new thermal load approximation rate parameters into the built-in early warning status assessment machine. The cloud-based diagnostic capacity expansion module 30 retrieves the graded safety operation threshold matrix from the associated power system operation procedure database and performs logical value comparison. The safety operation threshold matrix is pre-divided into normal operation range, over-limit warning range, and emergency load shedding range. In this embodiment, the value range of the normal operation range is 0-0.75, the value range of the over-limit warning range is 0.75-0.90, and the value range of the emergency load shedding range is 0.90-1.0. The early warning status assessment machine compares the real-time thermal load approximation rate parameters with the upper and lower boundary limits of the normal operation range, over-limit warning range, and emergency load shedding range. The early warning status assessment machine determines the current thermodynamic safety level of the target high-voltage cable.
[0060] S502, when the early warning status assessment machine determines that the current heat load approach rate parameter falls into the emergency load shedding range, the cloud-based diagnostic capacity expansion module 30 activates the internal active power redistribution algorithm. The cloud-based diagnostic capacity expansion module 30 generates a load transfer adjustment command based on the node power balance equation of the distribution network topology where the target high-voltage cable is located and the global maximum dynamic current carrying capacity. The command is calculated based on the difference between the current three-phase operating apparent power and the target maximum dynamic apparent power. The load transfer adjustment command carries the power reduction parameter. The cloud-based diagnostic capacity expansion module 30 sends the load transfer adjustment command to the distribution network data acquisition and monitoring control system. The distribution network data acquisition and monitoring control system drives the tie switch on the substation side to perform hardware load transfer action.
[0061] The formula for calculating the target power reduction in the active power redistribution algorithm is as follows: ; : The target power reduction amount that needs to be transferred or reduced from the current high-voltage cable power supply circuit; These are the fundamental algebraic constants for three-phase AC power systems; The rated line voltage parameters of the power grid system where the target high-voltage cable is located; This is a scalar representation of the current power frequency load current collected in real time at the edge side. The cloud-based diagnostic capacity expansion module outputs the global maximum dynamic current carrying capacity based on the thermoelectric coupling model; The steady-state operating load margin coefficient is specified in the power grid dispatching regulations. In this embodiment, the value range of the steady-state operating load margin coefficient is 0.85-0.95. This is the real-time physical power factor of the current power supply line.
[0062] S503, the cloud-based diagnostic capacity expansion module 30 extracts the partial discharge severity index reported by the edge computing module 20 and the currently calculated thermal load approximation rate parameter in parallel and performs two-dimensional spatial coordinate mapping. The cloud-based diagnostic capacity expansion module 30 calls the front-end visualization engine to convert the three-dimensional map partial discharge severity index and thermal load approximation rate parameter into two-dimensional scatter coordinates of the graphical operation interface. The cloud-based diagnostic capacity expansion module 30 generates a comprehensive health status map that maps the cable insulation aging degree and thermodynamic load pressure according to the position distribution of the scatter points in the two-dimensional coordinate system. The cloud-based diagnostic capacity expansion module 30 pushes the comprehensive health status map to the dispatch center display terminal through the hypertext transfer protocol. The dispatch center display terminal outputs the quantitative characteristics of the target equipment's deterioration evolution trend.
[0063] S504, after completing load scheduling and early warning push, the cloud-based diagnostic capacity expansion module 30 packages the full-dimensional perception data of the current scheduling cycle into the non-volatile historical operation database. The cloud-based diagnostic capacity expansion module 30 extracts the long-term operating condition data accumulated in the non-volatile historical operation database and the frequency records of historical over-limit early warning events of the equipment, and performs Pearson correlation statistical analysis. When the Pearson correlation coefficient calculation results show that there is a strong positive correlation between the operating time and the frequency of over-limit early warning, the cloud-based diagnostic capacity expansion module 30 adds a multiplicative aging amplification factor to the original insulation degradation penalty conversion coefficient for numerical correction. In this embodiment, the value range of the multiplicative aging amplification factor is 0.01-1.05.
[0064] For the communication interaction protocol of the distribution network data acquisition and monitoring control system and the graphics rendering call interface of the front-end visualization engine, those skilled in the art refer to the substation communication network standard documents published by the International Electrotechnical Commission and the manuals of mainstream graphics development libraries to write routine business code. The basic network topology node power balance equation and the data encapsulation and transmission format of the Hypertext Transfer Protocol are well-known technologies in the fields of power automation and computer software.
[0065] Application Examples: To aid in understanding the technical solution of this invention, the following provides application examples and test cases of the high-voltage cable dynamic capacity expansion method and system based on multi-parameter fusion and edge-cloud collaboration.
[0066] A multi-parameter fusion and edge-cloud collaborative high-voltage cable dynamic capacity expansion system was deployed for the target 110kV cross-linked polyethylene high-voltage cable. The cable's rated current carrying capacity is 650A and the insulation material's factory-set limit physical temperature tolerance is 90℃. Under broadband connectivity, the data synchronization acquisition module 10 acquires the power frequency load current as 610A and the thermodynamic temperature as 65.2℃. The edge computing module 20 synchronously acquires the high-frequency pulse of partial discharge. The logic operation unit outputs a partial discharge severity index of 1.2 after dimensionality reduction calculation.
[0067] The cloud-based diagnostic capacity expansion module 30 receives multi-dimensional state parameters and sets the insulation degradation penalty conversion factor to 0.15. The thermodynamic mathematical model solver combines the ultimate tolerance physical temperature and the measured temperature, and after deducting the penalty calculation, obtains a dynamic allowable temperature rise margin of 24.62℃. The cloud-based diagnostic capacity expansion module 30 performs iterative calculations based on the AC equivalent physical resistance and the system's overall total thermal resistance, and outputs a global maximum dynamic current carrying capacity of 835A. The cloud-based diagnostic capacity expansion module 30 performs physical ratio calculations to generate a heat load approximation rate parameter of 0.73. The early warning status assessment machine determines that the current parameters fall within the preset normal operating range.
[0068] The main fiber optic link is cut off and the dual-network heterogeneous communication module 40 loses four heartbeat packets in a row, triggering the timeout packet loss threshold. The edge computing module 20 receives the interrupt signal and switches to offline operation. The edge computing module 20 calls the lightweight one-dimensional convolutional neural network model and inputs the temperature and current time sequence characteristics of the previous 128 time nodes and the partial discharge severity index recorded at the breakpoint. The lightweight one-dimensional convolutional neural network model forward inference outputs a predicted operating temperature of 91.5℃ after 30 minutes. The edge computing module 20 extracts the 82℃ offline physical protection temperature benchmark and performs difference calculation. The predicted operating temperature exceeds the limit, triggering the offline emergency trip control function to output a high-level Boolean instruction. The edge computing module 20 sends a circuit breaker disconnect execution signal via the local control bus.
[0069] Test example: The experimental verification test was performed on a cable thermal cycling physical platform. A 50-meter-long cross-linked polyethylene cable was placed inside a constant temperature and humidity chamber. Fiber optic temperature probes were arranged at three-meter intervals on the surface of the cable's outer sheath, and artificial insulation defect nodes were configured at the joints. The power supply access end was connected to a programmable high-current generator and a transient harmonic injection device.
[0070] During the experiment, a static rated current-carrying capacity model was loaded as a control group and a fixed output of 650 amperes was used as a reference boundary. The system of this invention was used as the experimental group and operated synchronously. The programmable high current generator output alternating load current and the main communication link was physically disconnected during the current load rise interval. The transient harmonic injection source input high-frequency pulse electrical signals to the artificial defect node. The infrared thermal imager and the precision power quality analyzer respectively acquired and recorded the skin physical temperature scalar and the control execution delay scalar.
[0071] The experimental control list was generated by combining the discrete sampling values of sensors from multiple time periods and dimensions with the recorded data of system state response parameters. The test results are shown in the table below.
[0072] Multidimensional parameter measurement table of control group and experimental group under multiple environmental and equipment operating conditions.
[0073] From the table above and appendix Figure 7 and attached Figure 8 It can be seen that: during the low load current period, the current carrying capacity value dynamically calculated and output by the experimental group system is generally greater than the fixed static physical safety threshold of the control group. During the parallel phase of high load current input and high frequency pulse injection, the thermal load approximation rate parameter generated by the experimental group system enters the over-limit warning range, and the calculated current carrying capacity value output by the corresponding system shows a cliff-like downward constraint characteristic. When the communication link is offline and enters the local computing state, the physical delay of the action control triggered by the edge device directly executing the hardware takeover operation is less than the end-to-end response delay scalar in the cloud network broadband connection state. The response delay is even reduced by -6.9ms. This strongly proves that in the state of communication obstruction or disconnection, the edge offline state function can shorten the physical action delay of emergency protection and ensure the safe operation of system equipment.
[0074] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A method for dynamic capacity expansion of high-voltage cables based on multi-parameter fusion and edge-cloud collaboration, characterized in that, Includes the following steps: Simultaneously collect multi-dimensional operating status data of the target high-voltage cable. The multi-dimensional operating status data includes at least the cable surface thermodynamic temperature, power frequency load current, and high-frequency partial discharge signal collected by the underlying high-frequency pulse current sensor. Data preprocessing is performed on the multidimensional operating status data, and the heat load approximation rate parameter issued by the data monitoring network is extracted to determine the system computing power scheduling mode. Based on this, the physical sampling frequency of the underlying high-frequency pulse current sensor is adjusted, and the insulation degradation parameter is extracted based on the high-frequency partial discharge signal after the sampling frequency is adjusted. When the data monitoring network is connected, the insulation degradation parameter is introduced as a negative penalty term into the thermodynamic temperature rise equation, the global maximum dynamic current carrying capacity is output, and a new thermal load approximation rate parameter is generated synchronously based on the power frequency load current and returned to the edge side. When the data monitoring network is disconnected, the continuous power frequency load current sequence and the cable surface thermodynamic temperature sequence are extracted as input features. A lightweight one-dimensional convolutional neural network model is called to output the predicted operating temperature. A preset state machine penalty coefficient is introduced to solve the local dynamic current carrying capacity. The global maximum dynamic load capacity or the local dynamic load capacity is used as the safe threshold boundary for system operation. Combined with the new heat load approximation rate parameter, the output power of the underlying power generation equipment is adjusted and the tie switch is driven to perform load transfer action.
2. The high-voltage cable dynamic capacity expansion method based on multi-parameter fusion and edge-cloud collaboration according to claim 1, characterized in that, The steps of extracting the heat load approximation rate parameter issued by the data monitoring network to determine the system computing power scheduling mode and adjusting the physical sampling frequency according to the system computing power scheduling mode include: The extracted heat load approximation rate parameter is compared with the preset safety judgment threshold in the local non-volatile memory to calculate the approximation deviation. When the approximation deviation is greater than zero, a high-risk collaborative working mode is entered, and a hardware register configuration instruction based on the serial communication protocol is sent to increase the physical sampling frequency of the underlying high-frequency pulse current sensor from the reference frequency to a preset high-frequency threshold, and the priority of computing power resource allocation of the internal logic operation unit is reconstructed.
3. The high-voltage cable dynamic capacity expansion method based on multi-parameter fusion and edge-cloud collaboration according to claim 2, characterized in that, The step of extracting insulation degradation parameters based on the high-frequency partial discharge signal after adjusting the sampling frequency includes: In the high-risk collaborative working mode, the underlying high-frequency pulse current sensor, after increasing the physical sampling frequency, acquires a partial discharge dense sampling sequence, and performs partial discharge phase distribution feature extraction calculation on the partial discharge dense sampling sequence. The extracted high-frequency pulse amplitude and phase power frequency period coordinates are mapped and accumulated in a two-dimensional matrix to output a statistical probability distribution density matrix. Numerical dimensionality reduction is performed based on the statistical probability distribution density matrix to calculate the partial discharge severity index, which is then used as the insulation degradation parameter.
4. The high-voltage cable dynamic capacity expansion method based on multi-parameter fusion and edge-cloud collaboration according to claim 3, characterized in that, The step of introducing the insulation degradation parameter as a negative penalty term into the thermodynamic temperature rise equation and outputting the global maximum dynamic current carrying capacity includes: The partial discharge severity index, which is the insulation degradation parameter, is introduced as a linear negative penalty term into the dynamic allowable temperature rise margin equation of the cable insulation material. The dynamic allowable temperature rise margin is calculated and output by combining the ultimate physical temperature withstand capability of the target high-voltage cable with the current real-time thermodynamic temperature. The dynamic allowable temperature rise margin is substituted into the comprehensive thermodynamic constraint equation, which includes AC equivalent physical resistance and system comprehensive equivalent physical thermal resistance, and the square root iterative solution is performed to output the global maximum dynamic current carrying capacity.
5. The high-voltage cable dynamic capacity expansion method based on multi-parameter fusion and edge-cloud collaboration according to claim 1, characterized in that, The step of synchronously calculating and generating a new heat load approximation rate parameter based on the power frequency load current and driving the tie switch to perform load transfer operation includes: Extract the power frequency load current and combine it with the global maximum dynamic current carrying capacity to perform ratio mathematical calculations to generate the new heat load approximation rate parameter; When it is determined that the new heat load approximation rate parameter falls into the emergency load shedding range, the internal active power redistribution algorithm is activated. Based on the node power balance equation of the distribution network topology where the target high-voltage cable is located and the global maximum dynamic current carrying capacity, a load transfer adjustment command carrying the target power reduction amount is generated, driving the tie switch on the substation side to perform hardware load transfer action.
6. The high-voltage cable dynamic capacity expansion method based on multi-parameter fusion and edge-cloud collaboration according to claim 1, characterized in that, The determination steps when the data monitoring network is disconnected include: Send communication heartbeat packets to the cloud physical address of the data monitoring network, and count the frequency of continuous loss of heartbeat packets according to the preset timeout packet loss threshold; When the packet loss rate exceeds the preset timeout packet loss threshold, an interrupt signal is sent to the edge side and the backup wireless cellular network baseband chip is activated to perform network re-registration. When both primary and backup physical links are determined to be disconnected, the broadcast communication is marked as offline (Boolean), all state machine decision logic that depends on cloud parameters is forcibly suspended and the system switches to offline operation.
7. The high-voltage cable dynamic capacity expansion method based on multi-parameter fusion and edge-cloud collaboration according to claim 6, characterized in that, The step of extracting the continuous power frequency load current sequence and the cable surface thermodynamic temperature sequence as input features includes: In the offline operation state, the continuous power frequency load current sequence and the cable surface thermodynamic temperature sequence of a predetermined time step number are extracted from the multi-dimensional operation state data; The partial discharge severity index of the last successful interaction cached before the communication disconnection is used as the static compensation dimension. The continuous power frequency load current sequence, the cable surface thermodynamic temperature sequence, and the static compensation dimension are combined to form a three-channel input feature and spliced into a two-dimensional time-series feature input matrix. The internal fixed-point arithmetic unit is invoked to perform a maximum-minimum-extreme value normalization preprocessing operation on the two-dimensional time-series feature input matrix, constraining all elements of the input tensor matrix to a dimensionless numerical space.
8. The high-voltage cable dynamic capacity expansion method based on multi-parameter fusion and edge-cloud collaboration according to claim 7, characterized in that, The step of calling the output of the lightweight one-dimensional convolutional neural network model to predict the operating temperature includes: The preprocessed two-dimensional temporal feature input matrix is input into a cascaded one-dimensional causal convolutional layer, a max pooling layer, and a fully connected layer. Local temporal dependency features are extracted along the discrete time dimension through the one-dimensional causal convolutional layer, and data leakage at future time points is avoided through the causal padding mechanism. The max-pooling layer performs a local spatial dimension downsampling feature compression operation along the feature channel; The fully connected layer maps the high-dimensional temporal features output by pooling to single-node scalar values, and forward inference outputs the predicted operating temperature of the target high-voltage cable.
9. The high-voltage cable dynamic capacity expansion method based on multi-parameter fusion and edge-cloud collaboration according to claim 1, characterized in that, The step of solving the local dynamic load capacity by introducing a preset state machine penalty coefficient includes: Extract the cable's factory-set maximum withstand temperature and subtract the state machine penalty coefficient to derive the offline physical protection temperature reference. The predicted operating temperature and the offline physical protection temperature reference are combined to perform a sign difference calculation to generate a dynamic temperature rise margin. When the predicted operating temperature exceeds the limit, a high-level Boolean instruction is output through the offline emergency trip control function. The dynamic temperature rise margin is substituted as the target constraint into the locally solidified lightweight transient thermal circuit model to perform an inverse solution that deduces the load current from the upper temperature limit, and the local dynamic current carrying capacity is output.
10. A high-voltage cable dynamic capacity expansion system based on multi-parameter fusion and edge-cloud collaboration, characterized in that, The high-voltage cable dynamic capacity expansion method using the multi-parameter fusion and edge-cloud collaboration method according to any one of claims 1-9 includes: The data synchronization acquisition module is deployed at the target high-voltage cable node to synchronously acquire multi-dimensional operating status data of the target high-voltage cable. The multi-dimensional operating status data includes at least the cable surface thermodynamic temperature, power frequency load current, and high-frequency partial discharge signal acquired by the underlying high-frequency pulse current sensor. The edge computing module is used to perform data preprocessing on the multi-dimensional operating status data, extract the heat load approximation rate parameter sent by the data monitoring network to determine the system computing power scheduling mode, adjust the physical sampling frequency of the underlying high-frequency pulse current sensor accordingly, and extract insulation degradation parameters based on the high-frequency partial discharge signal after adjusting the sampling frequency; and extract the continuous power frequency load current sequence and the cable surface thermodynamic temperature sequence as input features when the data monitoring network is disconnected, call the lightweight one-dimensional convolutional neural network model to output the predicted operating temperature, and introduce a preset state machine penalty coefficient to solve the local dynamic current carrying capacity. The cloud-based diagnostic capacity expansion module is used to introduce the insulation degradation parameter as a negative penalty term into the thermodynamic temperature rise equation when the data monitoring network is connected, output the global maximum dynamic current carrying capacity and calculate and generate a new heat load approximation rate parameter and return it to the edge computing module. The dual-network heterogeneous communication module is used to connect the edge computing module and the cloud diagnostic capacity expansion module to build a data monitoring network, and broadcasts a communication offline Boolean flag to the local bus when a physical network disconnection is detected. The instruction takeover scheduling module is used to adjust the output power of the underlying power generation equipment and drive the tie switch to perform load transfer actions, using the global maximum dynamic load or the local dynamic load as the safety threshold boundary for system operation.
