Pole-mounted circuit breaker electrical variable monitoring method

Abstract

The invention relates to the technical field of electrical variable monitoring, and further relates to a pole-mounted circuit breaker electrical variable monitoring method. The method comprises the steps of 1, collecting a plurality of real-time electrical variables of the pole-mounted circuit breaker, performing normalization processing on each real-time electrical variable, and performing nonlinear fusion to form a single-dimensional electrical variable observation sequence; 2, acquiring historical sampling values of multiple electrical variable observation sequences within a fixed time delay interval so as to avoid embedding dimension explosion caused by vector operation and explicitly reflect a low-dimensional manifold hypothesis; and step 3, performing combined mapping on the intrinsic state quantity and the difference value of the insulation leakage current and the arc voltage, and predicting the fracture phase voltage in the future time. According to the invention, high-precision, high-robustness and prospective localized fault early warning and protection control can be realized, and the intelligent monitoring and safe operation capability of the pole-mounted circuit breaker in a complex power distribution environment can be obviously improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of electrical variable monitoring, and particularly relates to a method for monitoring electrical variables of a pole-mounted circuit breaker. Background Art

[0002] As an important protection switch device at the end of the distribution network, the pole-mounted circuit breaker is widely deployed in 10 - 35 kV medium-voltage distribution lines. Its operating state has a direct impact on the power supply reliability of the power grid and the fault handling efficiency. The pole-mounted circuit breaker operates in harsh outdoor environments such as high temperature, rain, dust, and corrosion for a long time, and is prone to various non-linear hidden faults such as contact deterioration, insulation aging, coil demagnetization, sticking of the closing mechanism, and arc re-ignition. The traditional method for judging the health state based on telemetry and telecommunication data relies on binary logic and manual experience, and it is difficult to reflect the internal dynamic deterioration process of the circuit breaker. Therefore, there are potential hazards such as "can trip but cannot close" and "seemingly normal but unable to disconnect the load". The current existing research on the state monitoring of pole-mounted circuit breakers mainly falls into the following categories: (1) Life estimation methods based on the number of operations or working years; (2) Feature extraction methods based on physical quantities such as contact resistance, current waveform, and voltage response; (3) Pattern recognition methods based on artificial intelligence; (4) Remote data fusion methods based on the Internet of Things platform. These technical routes have their own advantages, but there are still limitations to varying degrees in terms of engineering applicability, universality, real-time performance, and interpretability.

[0003] The first type of method uses the number of operating times or the service life of the equipment as the core index to estimate the remaining life of the circuit breaker. This type of method is simple and low-cost, but it ignores the influence of factors such as operating environment, current level, and operation shock on the equipment degradation rate, and it is difficult to reflect the true operating state. For example, in high-load and high-frequency operation scenarios, even if the circuit breaker has not reached the rated life times, problems such as sudden increase in contact resistance and demagnetization of the closing mechanism may have occurred; while in low-load and less-operated environments, the equipment may be in good condition for a long time but still be replaced in advance, resulting in waste of resources. The second type of method installs sensors on the circuit breaker body to collect variables such as current, voltage, temperature rise, contact resistance, and closing time, and extracts typical characteristic quantities for rule matching or threshold judgment. This type of method has a certain physical interpretability and is suitable for specific types of fault detection. However, it is highly dependent on feature extraction, and most of the parameter settings need to be adjusted manually, resulting in insufficient robustness. For example, the contact resistance is often affected by the load current fluctuation and generates instantaneous jitter. If the threshold value is too tight, it is easy to cause misjudgment, and if it is too loose, it may miss the detection of early contact deterioration. In addition, multiple faults (such as coil demagnetization and mechanism jamming) may present similar current waveforms, resulting in the lack of uniqueness of the features for discrimination. The third type of method takes algorithms such as artificial neural networks, support vector machines, and decision trees as the core, uses existing samples for supervised learning, and constructs a fault identification model. This type of method has a certain accuracy and adaptability when the samples are sufficient and the data quality is high. However, the artificial intelligence model is highly dependent on the scale of the sample library, the accuracy of data annotation, and the quality of model training, and there is generally a "black box" problem, making it difficult to explain the relationship between the prediction result and the physical mechanism of the equipment. In the power system scenario, engineering personnel have high requirements for the verifiability and physical consistency of the results, resulting in great resistance to the engineering implementation of this type of method. The fourth type of method emphasizes the system integration of the sensing end, the transmission end, and the platform end. Typically, it collects multi-dimensional sensor data through an edge computing gateway and uploads it to the cloud platform for centralized analysis and display. This type of method improves the multi-device collaborative monitoring ability and can achieve remote centralized management. However, its state determination still mostly depends on the platform rule library and template matching algorithm, and it fails to accurately model the true operating state of the equipment on the ontology side, making it difficult to meet the control requirements of "local intelligent judgment + fast closed-loop response". At the same time, restricted by communication delay and bandwidth, it is difficult to reduce the upload period of key state data to the millisecond level, and it is impossible to achieve real-time response to transient processes such as sudden arcs and rapid demagnetization. Summary of the Invention

[0004] The main purpose of the present invention is to provide a method for monitoring electrical variables of a pole-mounted circuit breaker. While improving the accuracy of state determination and reducing the misoperation rate, the present invention significantly shortens the operation and maintenance response time, extends the service life of the equipment, reduces the power outage risk, and provides interpretable and highly reliable online data support for lean asset management of the power grid.

[0005] To solve the above problems, the technical solution of the present invention is realized as follows: A method for monitoring electrical variables of a pole-mounted circuit breaker is provided, and the method includes: Step 1: Collect multiple real-time electrical variables of the pole-mounted circuit breaker, perform normalization processing on each real-time electrical variable, and then nonlinearly fuse them into a single-dimensional electrical variable observation sequence; Step 2: Within a fixed time delay interval, obtain historical sampling values of the electrical variable observation sequence multiple times, and perform two nonlinear processes on each historical sampling value respectively to obtain multiple scalar coordinates, specifically including: performing phase modulation with a harmonic function proportional to the sampling frequency to introduce time periodicity; using a gating function composed of the temperature of the excitation coil and the rated temperature rise to weaken the noise influence in the high-temperature section; then compressing all the scalar coordinates back into a single manifold internal state quantity in a geometric mean manner to avoid the explosion of the embedding dimension caused by vector operations and to explicitly reflect the low-dimensional manifold hypothesis; Step 3: Combine and map the internal state quantity with the difference between the insulation leakage current and the arc voltage to construct a real-time health index; according to the comparison between the real-time health index and the health threshold, if the real-time health index exceeds the set health threshold, an early warning is issued; continuously monitor the real-time health index, calculate the instantaneous gradient of the real-time health index with respect to time, and predict the break contact phase voltage at a future time.

[0006] Further, the real-time electrical variables include: the main circuit current of phase A of the pole-mounted circuit breaker, in A; the neutral line current, in A; the main circuit voltage of phase A of the pole-mounted circuit breaker, in V; the contact resistance of the contact, in Ω; the equivalent arc resistance, in Ω; the cumulative charge of the excitation coil, in C; the cumulative charge of the excitation coil at the previous historical time, in C; the sampling time delay step, in s, which is an integer multiple of the sampling period.

[0007] Further, Step 1 specifically includes: characterizing the total current intensity through the vector synthesis of the main circuit current of phase A of the pole-mounted circuit breaker and the neutral line current, reflecting the instantaneous power level in combination with the main circuit voltage of phase A of the pole-mounted circuit breaker, and using the series total resistance of the contact resistance of the contact and the equivalent arc resistance as the loss factor of the power transmission path to form a quantization index of the instantaneous power transmission efficiency as the power efficiency index; at the same time, introducing a logarithmic sensitivity detection mechanism for the change rate of the cumulative charge of the excitation coil, performing a logarithmic transformation on the absolute value of the increment of the cumulative charge of the excitation coil and the cumulative charge of the excitation coil at the previous historical time to realize the amplified detection of small magnetic field disturbances and obtain the magnetic field sensitivity index; finally, linearly superposing the power efficiency index and the magnetic field sensitivity index to form a single-dimensional electrical variable observation sequence.

[0008] Furthermore, the scalar coordinates are obtained through the following process: Establish a time periodicity detection mechanism based on the harmonic sequence of the sampling frequency, and achieve the frequency domain separation of historical data through the embedding frequencies corresponding to different delay embedding dimensions, ensuring the independence of information at each historical moment in the frequency domain; Parallelly construct an adaptive noise suppression system based on the temperature of the excitation coil, establish a dynamic gating function using the ratio relationship between the temperature of the excitation coil and the rated temperature rise of the excitation coil, automatically reduce the signal weight under high temperature conditions to suppress thermal noise interference, and through the ratio operation of the frequency domain modulation signal and the temperature gating function, realize the conversion of the information of the historical electrical variable observation sequence to the current scalar coordinates, where the intrinsic dimension of the manifold determines the number of historical time sample points required to construct the low-dimensional state.

[0009] Furthermore, the intrinsic state quantity of the manifold is obtained through the following process: Use the equal-weight geometric mean algorithm to process the scalar coordinates of all delay embedding dimensions, perform a fractional power operation after unitizing and biasing each scalar coordinate, and set the power exponent to the reciprocal of the intrinsic dimension of the manifold to ensure the balanced contribution of information in each dimension. Finally, through the product operation, realize the dimensionality reduction mapping from multi-dimensional discrete information to a single continuous intrinsic state quantity of the manifold, and this state quantity can achieve dimensionality compression while maintaining the integrity of the original information.

[0010] Furthermore, the rated temperature rise of the excitation coil is obtained through the following process: Based on the heat dissipation capacity of the power loss of the excitation coil to the environment through the thermal resistance path of the excitation coil under steady-state conditions, establish a linear thermal resistance relationship between the power loss and the temperature rise, and introduce a safety factor to conservatively correct the theoretically calculated temperature rise. The numerical range of the safety factor ensures that the system has sufficient thermal safety margin under rated conditions, providing a reference value for the subsequent setting of the temperature gating function.

[0011] Furthermore, for the quantitative evaluation of the real-time health index of the equipment health status, a multi-coupling analysis method of the intrinsic state quantity of the manifold, the insulation leakage current, and the arc voltage characteristics is adopted, specifically including: Using the intrinsic state quantity of the manifold as the basic evaluation index of the equipment operation state, realizing the reverse correlation evaluation of the insulation performance through the reciprocal square root weighting of the insulation leakage current as the insulation weight, and introducing the normalized exponential function of the arc voltage time series change rate relative to the rated operating voltage of the circuit breaker as the arc amplification factor to amplify the exponential impact of the arc characteristics on the real-time health index. Through the triple product operation of the basic evaluation index, the insulation weight, and the arc amplification factor, construct a unified real-time health index that can comprehensively reflect the electrical health level of the equipment.

[0012] Furthermore, the health threshold for equipment health warning is set based on the safety margin determination principle under extreme operating conditions, specifically including: establishing an insulation safety margin according to the maximum allowable leakage current level that the equipment insulation system can withstand, realizing the inverse constraint relationship of the maximum allowable leakage current on the health threshold through the square root reciprocal transformation to obtain the insulation constraint; at the same time, establishing an arc safety margin based on the maximum allowable arc voltage transient amplitude that the arc system can withstand, and realizing the exponential constraint relationship of the arc transient on the health threshold through the normalization exponential transformation relative to the rated operating voltage of the circuit breaker to obtain the arc constraint; finally, determining the comprehensive health threshold through the product of the insulation constraint and the arc constraint.

[0013] Furthermore, the future evolution prediction of the breaker contact voltage is based on the superposition prediction method of the column-mounted circuit breaker phase A main circuit voltage reference and multi-physical field coupling correction: using the column-mounted circuit breaker phase A main circuit voltage as the initial reference value for prediction, reflecting the dynamic change trend of the equipment state through the real-time health index time evolution gradient, constructing a time-varying correction factor by combining the prediction time step, vacuum permeability, copper conductivity, and the characteristics of the DC resistance of the excitation coil and the contact resistance of the contact, and introducing the ratio of the circuit breaker heat capacity to the equivalent capacitor energy storage as the correction index of the energy conversion characteristic, calculating the correction amount of the voltage evolution through the multi-physical field synergy of electromagnetic coupling, impedance evolution, and energy conversion, and finally realizing the quantitative prediction of the breaker contact voltage at the future moment through the superposition of the initial reference value and the correction amount.

[0014] The on-pole circuit breaker electrical variable monitoring method of the present invention example has the following beneficial effects: By mapping the multi-source real-time electrical quantities of the on-pole circuit breaker into a one-dimensional observation sequence, and then using low-dimensional manifold reconstruction and geometric mean compression, a unified internal state quantity that can dynamically reflect conduction power, insulation deterioration, thermal degradation, and arc disturbance is constructed. Furthermore, a health index is generated at the millisecond level and combined with the time gradient to complete the forward prediction of the contact voltage, realizing a full-link local closed-loop from sampling, evaluation to decision-making. This method completely abandons artificial experience coefficients, and all constants are derived from the circuit breaker nameplate or international physical constants. Therefore, it can be directly transplanted across models and manufacturers, and numerical stability is maintained through normalization and logarithmic compression means, avoiding false alarms and missed alarms caused by traditional threshold tuning. The core of the algorithm only involves multiplication, division, logarithm, cosine, and exponential operations, and can run in real time on resource-constrained embedded controllers, without relying on external clouds or complex neural networks, and has extremely high deployment feasibility. By automatically suppressing high-temperature noise through a temperature gating function, strengthening power frequency characteristics through cosine modulation, and exponentially amplifying early anomalies through the health index, the present invention can issue early warnings at the budding stage of faults such as insulation leakage, arc reignition, and contact ablation, and trigger bypass voltage limit or tripping actions in advance based on trend prediction, upgrading the protection strategy from passive overlimit to active foresight. Compared with traditional methods that only rely on the number of operations or static characteristics, the present invention improves the state determination accuracy and reduces the misoperation rate, significantly shortens the operation and maintenance response time, extends the equipment service life, reduces the power outage risk, and provides interpretable and highly reliable online data support for lean power grid asset management. Description of the Drawings

[0015] Figure 1 It is a schematic flowchart of the method for monitoring electrical variables of an on-pole circuit breaker provided by an embodiment of the present invention. Detailed Embodiments

[0016] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0017] Reference Figure 1 : A method for monitoring electrical variables of an on-pole circuit breaker, the method comprising: Step 1: Collect multiple real-time electrical variables of the on-pole circuit breaker, perform normalization processing on each real-time electrical variable, and then non-linearly fuse them into a one-dimensional electrical variable observation sequence; First, it is required that the acquisition module simultaneously reads the line current, neutral line current, phase voltage, contact resistance of the contact, equivalent arc resistance, and cumulative charge of the excitation coil under millisecond-level clock synchronization, and strictly timestamps these data, so as to ensure that the physical states at the same logical instant are referenced in each subsequent fusion operation. Then, based on the reference values such as the rated current, rated voltage, rated on-resistance, and rated excitation charge given on the circuit breaker nameplate, the system dimensionlessizes the six types of original quantities. The method is to divide their respective instantaneous measurement values by the corresponding rated values and suppress the sampling noise through exponential smoothing filtering. This step not only solves the dominant deviation caused by unit differences but also compensates for the thermal drift generated by the equipment under day-night temperature differences and seasonal temperature differences through the sliding window tracking method, so that the normalized result remains amplitude-stable in extreme climate scenarios. After normalization, it enters the non-linear fusion stage. The system regards the line current and phase voltage as energy input channels, converts them into instantaneous energy flow indicators through the product relationship in terms of power, and then introduces the contact resistance and arc resistance as dynamic blocking factors of the energy flow to describe the real-time impact of the degradation of the conduction path on the energy flow in a fractional structure; at the same time, the neutral line current is most sensitive to early insulation leakage during three-phase imbalance, so it is separately extracted as an absolute value and then squared and amplified to enter the same fusion framework, so that the diagnostic weight of slight leakage can still be amplified under low load conditions.

[0018] To avoid the huge pulses caused by arc instantaneous break or re-ignition from pulling the fusion sequence into spikes, the system also superimposes a self-suppression channel based on the logarithmic function between the power component and the resistance component. After mapping the charge difference of the excitation coil between adjacent sampling points into the magnetic potential perturbation energy, it is injected into the fusion formula in a logarithmic compression manner, which can not only make the actual operation action generate a recognizable step for the observation sequence but also prevent numerical overflow caused by short-term shocks. All normalized and non-linearly processed quantities are then written into the on-chip pipeline in a fixed order, and the composite operation is completed through the hardware multiplier-accumulator. The resulting one-dimensional observation sequence is a pure scalar data stream with equally spaced sampling, high-order coupling relationships, and significant noise filtered in terms of data structure; at the exit of the sequence generation, the firmware sets a sliding gating algorithm based on the cumulative thermal energy to continuously monitor the calculated temperature rise of the contact and the coil. As long as it is monitored that the overheating exceeds the design margin, the smoothing coefficient of the adjacent sampling step is automatically adjusted to reduce the chronic rise of the sequence baseline caused by long-term thermal drift and ensure numerical centralization. The entire step from the bottom layer drive to the top layer value output is encapsulated within the high-priority interrupt service of the real-time operating system, ensuring that all operations in a single cycle are completed within the sampling interval without omission, and writing the key intermediate data into the circular buffer through redundant check caching for subsequent manifold embedding algorithm calls. Step 2: Within a fixed time delay interval, obtain the historical sampling values of multiple electrical variable observation sequences, and perform two non-linear processes on each historical sampling value respectively to obtain multiple scalar coordinates, specifically including: performing phase modulation with a harmonic function proportional to the sampling frequency to introduce time periodicity; using a gating function composed of the temperature of the excitation coil and the rated temperature rise to weaken the noise influence in the high-temperature section; then compressing all the scalar coordinates back into a single manifold intrinsic state quantity in a geometric mean manner to avoid the explosion of the embedding dimension caused by vector operations and to explicitly reflect the low-dimensional manifold hypothesis; Step 2 uses the delay embedding idea in the theory of non-linear dynamic systems to re-characterize the multi-dimensional internal degrees of freedom hidden behind the time series in an explicit and extremely low-dimensional manner, thereby providing a physically interpretable state quantity for subsequent health assessment and trend prediction. Specifically, in the implementation process, first, at the firmware initialization stage, according to the breaker operation cycle, rated power frequency, and sampling rate provided by the manufacturer, combined with the embedded storage capacity, a fixed delay step and an embedding dimension sufficient to cover the control freedom are set to determine the depth of the historical snapshot buffer. After each new frame of the observation sequence is sampled, the system will push the current value into the circular buffer in chronological order and synchronously pop the oldest value to ensure that the buffer always stores several historical points at the same interval. Subsequently, the device enters the double non-linear transformation stage: First, apply a harmonic function modulation proportional to the sampling rate to each historical value in the buffer, which is equivalent to attaching a phase label synchronized with the power frequency to each historical point on the time axis. With the help of this phase information, the delay embedding coordinates can be made sensitive to periodic oscillations and harmonic disturbances, so that the internal dynamic structure of the system can still be fully revealed under conditions such as power fluctuations, excitation circuit oscillations, or low-frequency swaying.

[0019] Next, the system calls a gating function based on the upper limit of the insulation class temperature rise using the excitation coil temperature readings at the same historical moment. This function is almost transparent when the temperature is below the upper limit and compresses the coordinate amplitudes when the temperature approaches or exceeds the upper limit to suppress the deviation of the state estimation caused by sensor noise or the drift of the electrical parameters of the insulating medium in a high-temperature environment, thereby ensuring that thermal drift does not misjudge healthy equipment as abnormal. After completing harmonic modulation and temperature gating, the system has obtained scalar coordinates equal to the embedding dimension. To avoid the computational burden caused by processing vector space parallel operations in a resource-constrained field controller, it no longer retains these coordinates in vector form but uses a geometric mean strategy to compress all coordinates into a single intrinsic state scalar. The reason for choosing the geometric mean instead of the arithmetic mean or maximum-minimum reduction is that this operation can numerically balance the relative gain and attenuation of each coordinate amplitude and naturally maintain the linearly separable property of the multiplicative coupling relationship, enabling any subsequent health formula based on logarithmic or exponential mapping to be directly decomposed to the original coordinate level for physical interpretation.

[0020] When calculating the geometric mean, to avoid overflow of floating-point multiplication when the embedding dimension is large, the system uses an algorithm of segmented logarithmic summation and then exponential restoration, which significantly reduces the clock occupancy of the multiply-accumulate hardware while ensuring accuracy. After the intrinsic state scalar is generated, it is immediately written into an output register with only one-bit depth, and this register is exposed to the upper-layer health assessment task through memory mapping, enabling the real-time health index to read the latest streaming morphological quantity with extremely low latency. It should be emphasized that to allow the delay embedding to fully characterize the dynamics at both the high-frequency impact and low-frequency trend levels, the firmware also maintains an adaptive tuning mechanism in the background: when it detects that the variance of the observation sequence increases sharply or the autocorrelation coefficient decreases significantly within a short period of time, the system will automatically shorten the delay step to capture high-frequency anomalies with denser historical points; conversely, if the observation sequence shows long-period slow drift, the delay step will be extended to improve the resolution of slow-varying degradation. This tuning process is implemented inside the microcontroller through finite state machine logic and does not rely on external parameter issuance, thus ensuring that the manifold reconstruction always adaptively expands and contracts around the actual system degrees of freedom under different operating seasons, load levels, and operation frequencies.

[0021] Step 3: Combine and map the difference between the intrinsic state quantity, the insulation leakage current, and the arc voltage to construct a real-time health index; based on the comparison between the real-time health index and the health threshold, if the real-time health index exceeds the set health threshold, issue a warning; continuously monitor the real-time health index, calculate the instantaneous gradient of the real-time health index with respect to time, and predict the breaker phase voltage at future times.

[0022] Fuse the single manifold internal state variables obtained in the previous step with the key electrical variables that can directly reflect insulation and opening / closing transients to generate a real-time health index capable of immediately discriminating the equipment safety margin, and on this basis, achieve millisecond-level alarm triggering and future trend extrapolation. The implementation principle is as follows: First, according to the operating mechanism of the pole-mounted circuit breaker, regard the low-dimensional state variables after temperature gating as the concentrated expression of the coupling of the equipment's global energy and structure. Then, select the leakage current amplitude that is most sensitive to early insulation degradation and the arc voltage difference that is most intuitive for contact wear and arc reignition to construct a ternary coupling relationship. The system maintains a robust three-channel data path at the firmware level: the internal state variables are written into the shared register by the manifold module in each sampling period, the leakage current is synchronously collected after being connected to the isolation analog-to-digital converter by a high-resistance shunt, and the arc voltage difference is refreshed in real time by calculating the adjacent difference through a high-speed sampling buffer; the update timestamps of the three channels must be forced to align to avoid misjudgment caused by cross-delay. In the health index calculation logic, the engineering implementation selects a mapping mode of multiplication-division combination and logarithmic amplification, so that when any one of the three types of indicators deviates from the normal range, the index can show a monotonically increasing response curve, and when multiple indicators deteriorate simultaneously, the index increase will show an exponential superposition effect, so that minor anomalies and serious faults can be quickly distinguished in the numerical dimension.

[0023] To adapt to the natural drift of insulation current and arc voltage in different altitude, humidity, and temperature environments, the algorithm continuously collects the sliding average statistics within two weeks as the dynamic baseline during operation, and sets the health threshold as the dynamic baseline multiplied by a fixed safety factor. This factor comes from the factory inspection data of the same type of circuit breaker by the manufacturer and usually remains unchanged, thus ensuring cross-environment consistency without introducing human experience factors. Once the calculated real-time health index breaks through the threshold, the firmware immediately generates a high-priority alarm flag in the local interrupt and synchronously sends signals to the upper-level protection device and dispatching platform through relay contacts, IEC-61850 messages, or wireless LoRa frames; at the same time, since the index is continuously output instead of a binary flag, the upper computer can formulate disposal strategies in layers according to its specific value. For example, when it exceeds the threshold by 20%, a voice warning is sent first, when it exceeds 50%, the bypass switch is linked, and when it exceeds 80%, it trips directly to stop operation, so as to achieve hierarchical disposal of wear and out-of-control.

[0024] In addition to the one-time threshold comparison, the algorithm calculates the time difference of the health index in each cycle to obtain gradient information. This gradient is embedded in a small first-order extrapolator and used together with the nameplate constants such as the contact resistance, coil resistance, and equivalent heat capacity of the device to estimate the possible value of the broken-phase voltage after several future sampling cycles. This predicted value can be displayed in the background visualization curve for maintenance engineers to analyze the deterioration process of the contact, and can also be used as a trigger condition for early action in the protection logic. To prevent false alarms caused by exponential spikes due to power grid transients or harmonic interference, the system designs a dual-window criterion: the short window has a length of five sampling points and is specifically used to capture sudden out-of-control; the long window has a length of hundreds of sampling points and mainly tracks slow-varying degradation, and the alarm is confirmed to be effective only when both windows show an upward trend, thus avoiding false actions caused by short-term spikes. The entire process runs completely in the edge controller of the circuit breaker body and does not rely on the cloud, ensuring that the status assessment and protection tripping can be completed locally even in case of communication interruption. In addition, the algorithm implementation fully considers the needs of equipment aging and firmware upgrade. All thresholds and constants are encapsulated in an independent read-only storage block, and new parameters can be written in batches through an external handheld terminal during upgrade without modifying the code, which greatly facilitates batch maintenance.

[0025] Further, the real-time electrical variables include: the main circuit current of phase A of the pole-mounted circuit breaker , with the unit of A; the neutral line current , with the unit of A; the main circuit voltage of phase A of the pole-mounted circuit breaker , with the unit of V; the contact resistance of the contact , with the unit of Ω; the equivalent resistance of the arc , with the unit of Ω; the accumulated charge of the excitation coil , with the unit of C; the accumulated charge of the excitation coil at the previous historical time , with the unit of C; the sampling time delay step , with the unit of s, which is an integer multiple of the sampling period; where \(t\) is the time.

[0026] In the method for monitoring electrical variables of the pole-mounted circuit breaker proposed in the present invention, all real-time collected data are established based on a set of basic quantities with clear physical meanings. Among them, the main circuit current of phase A of the pole-mounted circuit breaker is denoted by the symbol , with the unit of ampere, which directly reflects the instantaneous current intensity when the load is transmitted through the contact and the busbar; the supporting neutral line current is represented by , also with the unit of ampere, and is used to capture the change of the zero-sequence component caused by three-phase imbalance or insulation leakage; on the voltage side, the instantaneous phase voltage of phase A of the main circuit is denoted as , with the unit of volt, which is the direct benchmark for constructing the instantaneous power and judging the energy flow state of the system; in order to explicitly reflect the conduction quality in the energy flow link, the contact resistance of the contact and the arc equivalent resistance , both with the unit of ohm. The former reflects the metal contact state of static closed conduction, and the latter depicts the dynamic blocking characteristics of the arc channel to the current during breaking or reignition; the accumulated charge of the excitation coil on the driving mechanism side is in coulombs and is respectively represented by and . They record the magnetic potential energy storage inside the coil at the current sampling moment and the previous historical delay moment. The influence of a single make-and-break operation on the system magnetic field distribution can be quantified through the difference between them; the delay step itself is in seconds and must be an integer multiple of the sampling period. This design ensures that the intervals between each historical point in the buffer are uniform during subsequent delay embedding, while the time variable only exists as a unified index.

[0027] Through the above clearly defined physical quantities, the system can first normalize and map currents, voltages, resistances, and charges with different dimensions to the same dimensionless scale in step one, and then use the multiplication, division, logarithm, and difference relationships between them to construct a one-dimensional observation sequence; during the delay embedding process in step two, each historical sampling point carries the entire set of quantified physical quantity characteristics. After harmonic modulation and temperature gating, they jointly form scalar coordinates and are geometrically averaged and compressed into internal state quantities; in step three, the internal state quantities, together with the differential of the leakage current and arc voltage, enter the health index mapping model, so that any anomaly caused by or will be immediately amplified and presented in the index. At the same time, the change in the coil charge provides the driving force for the time gradient of the index, thereby supporting the forward prediction of the future contact voltage. Further, step 1 specifically includes: characterizing the total current intensity through the vector synthesis of the phase A main circuit current and the neutral line current of the pole-mounted circuit breaker, combining the phase A main circuit voltage of the pole-mounted circuit breaker to reflect the instantaneous power level, and using the series total resistance of the contact resistance and the arc equivalent resistance as the loss factor of the power transmission path to form a quantization index of the instantaneous power transmission efficiency; at the same time, introducing a logarithmic sensitivity detection mechanism for the change rate of the accumulated charge of the excitation coil, and performing a logarithmic transformation on the absolute value of the increment of the accumulated charge of the excitation coil and the accumulated charge of the excitation coil at the previous historical time to achieve the amplified detection of small magnetic field disturbances. Finally, linearly superimposing the power efficiency index and the magnetic field sensitivity index to form a one-dimensional electrical variable observation sequence. The electrical variable observation sequence over time is: .

[0028] In the whole structure, the numerator part is an estimate of the instantaneous electric energy flow intensity of the actual system. This item first calculates the main circuit current and the neutral line current Taking the square root after summing the squares is equivalent to calculating the magnitude of the instantaneous current vector in the main circuit of phase A and the neutral line, thus taking into account both the normal conduction current and the neutral line feedback caused by three-phase imbalance. By further multiplying with the phase voltage of phase A we can obtain an approximate expression of the instantaneous power. Although this power does not consider the power factor, in the analysis of the distribution side, the phase deviation is small, and such a simplification can approximately capture the power transmission state with a low computational cost.

[0029] This energy flow product is placed in a fractional structure with as the denominator, aiming to perform a coupling ratio mapping between the instantaneous energy flow intensity and the equivalent impedance of the conduction path. The contact resistance usually increases with the wear, ablation or oxidation of the contact surface, so it is an important indicator to characterize long-term contact deterioration; while the arc equivalent resistance reflects the non-linear blocking characteristics of the arc channel on the conduction path during short-term make or break processes. The sum of these two resistances constitutes the instantaneous total impedance of the entire current path, which can accurately reflect numerically whether the current can be effectively transmitted to the load end. When the contact condition is good and there is no arc, the impedance is low and the value of the whole fraction is high; while when an arc occurs or the contact deteriorates severely, the impedance increases, resulting in a significant decrease in this term. The overall fractional term thus formed represents the power transfer efficiency under unit impedance, that is, the electrical energy output ability under unit conduction resistance.

[0030] In addition to the power conduction path, the operating behavior of the circuit breaker is also clearly reflected in the formula, that is, by introducing a second term through a logarithmic mapping of the charge change term of the excitation coil to form the right half of the formula. The excitation coil is the driving source for the circuit breaker to complete the closing or opening operation, and the magnetic potential inside it is formed through charge accumulation. If there is a significant change in the coil charge amount at two time points before and after a certain sampling moment, it indicates that the circuit breaker has just completed an operating event. This charge change is wrapped in a logarithmic function, making the contribution value small when the change amplitude is small, and significantly amplifying when there is a jump. At the same time, the compression characteristic of the logarithmic function suppresses the spike interference of the charge value jitter on the entire sequence. Adding a constant ensures that even if the charge difference is zero, it will not cause illegal logarithmic calculation. The design of this term has two meanings: one is to provide a detectable path for the operating behavior of the circuit breaker, enabling the operating event to be characterized in the manifold; the other is to avoid the interference of errors caused by driving disturbances on the overall state judgment, and it has strong robustness under high-frequency operating conditions.

[0031] The whole The sequence structure unifies multiple objectives: it unifies different physical quantity units within the same expression; it establishes a non-linear mapping relationship through the combination of fractions and logarithms; it introduces explicit non-linear coupling through operations such as modulus length, multiplication, and difference; more importantly, the expression does not contain any artificial adjustment factors or empirical parameters, and all its variables are derived from the physical quantities that can be real-time collected by the device itself, which ensures that the model adaptability will not decrease due to individual differences, manufacturing deviations, or different operating environments in subsequent modeling. At the engineering implementation level, this formula only requires basic addition, multiplication, square root, and logarithm operations, and is suitable for online real-time operation of low-resource embedded chips.

[0032] Furthermore, the scalar coordinate is obtained through the following process: establish a time periodicity detection mechanism based on the harmonic sequence of the sampling frequency, and achieve the frequency domain separation of historical data through the embedding frequencies corresponding to different delay embedding dimensions to ensure the independence of information at each historical moment in the frequency domain; parallel construct an adaptive noise suppression system based on the temperature of the excitation coil, establish a dynamic gating function using the ratio relationship between the temperature of the excitation coil and the rated temperature rise of the excitation coil, automatically reduce the signal weight under high temperature conditions to suppress thermal noise interference, and realize the conversion of the historical electrical variable observation sequence information to the current scalar coordinate through the ratio operation of the frequency domain modulation signal and the temperature gating function, where the intrinsic dimension of the manifold determines the number of historical time sample points required to construct the low-dimensional state. The scalar coordinate is: ; where, represents the scalar coordinate corresponding to the th delay embedding dimension; is the intrinsic dimension of the manifold, indicating how many historical time sample points are used to construct the low-dimensional state; is the electrical variable observation sequence of the th delay embedding dimension; represents the temperature of the excitation coil of the th delay embedding dimension, with the unit of K; is the rated temperature rise of the excitation coil, with the unit of K; is the embedding frequency of the th delay embedding dimension, , is the sampling frequency.

[0033] By the given integer multiple sampling period slice and extract the observation sequence on the historical axis to form to in total The formula uses a uniformly spaced sample, each of which represents the overall electrical state of the circuit breaker at different instants in the past, to construct a history vector sufficient to expand the system's degrees of freedom. However, in order to avoid the storage and computational overhead caused by vector operations, this formula does not directly retain the vector, but immediately applies cosine modulation to each history segment. The embedded frequency of this modulation term is Proportional to the sampling frequency, making the difference The corresponding coordinates are distributed equidistantly in the frequency domain, so that the power frequency and its harmonic components are explicitly superimposed into the coordinate expression, ensuring that when the main circuit current or voltage has periodic oscillations caused by load fluctuations, the manifold coordinates can maintain a sensitive response with consistent phase. The denominator corresponding to the modulation term introduces the temperature gating function When the temperature of the excitation coil is much lower than the rated temperature rise, the function is approximately a constant of one, and has almost no effect on the coordinate amplitude; when the temperature approaches or exceeds the rated temperature rise, the exponential term increases significantly, causing the denominator to rapidly amplify, thereby suppressing the high-amplitude jitter caused by thermal noise.

[0034] In this way, the periodic modulation ensures the capture of harmonics and power frequency characteristics, and the temperature control gate suppresses the risk of noise amplification of the measured value in a high temperature environment. The two together ensure that each All contain the largest information entropy in the smallest scalar space. It is worth noting that Taken from The temperature sensor readings with the same historical index ensure the consistent time base for thermal control and power observation; It is directly given by the nameplate insulation level and does not change with the environment or operator experience, so this formula does not introduce any adjustment factors. In general, the structure completes four layers of coupling: the first is time coupling, which weaves different historical fragments into the coordinates with a uniform lag step; the second is frequency coupling, which incorporates the power frequency and its resonance into the state quantity under cosine modulation; the third is thermal coupling, which realizes dynamic suppression of thermal degradation through the temperature gating function; the fourth is amplitude coupling, which uses the nonlinear structure of the observation sequence itself to transmit the relationship between current, voltage and resistance.

[0035] When all Corresponding When the subsequent geometric mean is compressed into a single intrinsic state, these coupling relationships are not eliminated, but are retained in the combined weights of the exponent and the logarithm through the commutativity of multiplication and exponentiation operations. Therefore, the final low-dimensional flow morphology inherits the harmonic structure of the system, retains the suppression characteristics of thermal degradation, and maintains stable and differentiable smoothness in numerical terms, providing a reliable prerequisite for the continuous evolution and gradient calculation of the real-time health index. With only one read buffer, one multiplication, one cosine table lookup, and one exponential table lookup plus division, the generation of scalar coordinates can be completed within one clock cycle of an embedded microcontroller without vector storage, greatly reducing the requirements for on-chip memory and computing power. At the same time, the ratio of the frequency index to the intrinsic dimension allows developers to dynamically adjust the dimension size according to the actual degrees of freedom of the circuit breaker, and the structure in which the embedded frequency increases linearly

[0036] guarantees that the increase in dimension will not cause frequency aliasing. Through this design, even under high altitude, strong sunlight, or severe cold environments, the drastic changes in the temperature of the excitation coil can be effectively limited by the gating function, while the harmonic injection caused by mechanical shocks and electromagnetic interference common in industrial sites can be significantly revealed through cosine modulation. Thus, manifold analysis can not only accurately capture latent faults but also distinguish whether the electrical anomaly is caused by structural thermal degradation or short-term harmonic interference. This method of completing multi-dimensional feature fusion at the scalar level enables subsequent steps to directly perform health assessment and future trend prediction on a single intrinsic state without the need for dimensionality reduction or feature selection, demonstrating the comprehensive advantages of the present invention in algorithm compactness, physical interpretability, and embedded deployability.

[0037] Furthermore, the rated temperature rise of the excitation coil is obtained through the following process: Based on the heat dissipation capacity of the power loss of the excitation coil to the environment through the thermal resistance path under steady-state conditions, a linear thermal resistance relationship between the power loss and the temperature rise is established, and a safety factor is introduced to conservatively correct the theoretically calculated temperature rise. The numerical range of the safety factor ensures that the system has sufficient thermal safety margin under rated conditions, providing a reference value for setting the subsequent temperature gating function. The manifold intrinsic state quantity of time is: The role of the add-one operation in The coordinates are involved as a whole through multiplication, which means that if the state of a certain historical delay dimension shows significant anomalies, such as a sharp increase in instantaneous impedance due to arc reignition or a compression of the amplitude in the high-temperature section due to the thermal suppression mechanism, it will be directly reflected as an overall increase or decrease in the multiplication, thereby generating an amplified global response, and this global amplification can more accurately capture the degree of danger when the multi-dimensional state synchronization is unbalanced compared with the traditional arithmetic mean. At the same time, the geometric averaging operation of taking the root of the product result also plays a role in normalizing the dimension and stabilizing the amplitude, so that no matter how the embedding dimension is adjusted, the magnitude is stable within an interval of the same order as a single coordinate, which not only avoids the numerical explosion caused by dimensional expansion, but also ensures that no additional normalization is required when constructing the real-time health index using the exponential mapping subsequently. Physically, this structure can be regarded as a kind of manifold volume measure of "equal contribution of each historical state quantity": when all are close to zero, is approximately equal to one, indicating that the system is in the nominal working condition; when any one-dimensional coordinate shows a positive or negative deviation, the geometric mean will deviate from one, indicating that the trajectory of the system at this time scale has deviated from the normal manifold. More importantly, since generally includes a cosine harmonic modulation term and a temperature control suppression term, the continuous change presented on the time axis actually simultaneously reflects the comprehensive results of three dynamics: power frequency power perturbation, harmonic sideband energy, and thermal degradation rate, which makes become a highly information-intensive and physically interpretable single index. When implementing in engineering, in order to derive with extremely low latency on the embedded controller, the firmware usually directly writes the pre-computation table into the look-up table, and then uses the algorithm of logarithmic summation and exponential reduction to convert the multiplication chain into a summation operation and eliminate the risk of floating-point multiplication overflow; combined with the means of loop unrolling, the entire dimensional geometric mean can be completed within one sampling period and will not pose a bottleneck to the refresh rate of the real-time health index. It is worth emphasizing that when the system dynamically adjusts the embedding dimension to adapt to different load or seasonal working conditions, the in the geometric mean index automatically reorganizes and normalizes the ratio, so there is no need to modify the downstream threshold, which enables the method to maintain the same false alarm rate and miss rate under on-site conditions of different circuit breaker models, different sampling rates, and even different communication rates.

[0038] The rated temperature rise of the excitation coil is obtained through the following process: Based on the heat dissipation capacity of the power loss of the excitation coil to the environment through the thermal resistance path under steady-state conditions, a linear thermal resistance relationship between the power loss and the temperature rise is established, and a safety factor is introduced to conservatively correct the theoretically calculated temperature rise. The numerical range of the safety factor ensures that the system has sufficient thermal safety margin under rated conditions, providing a reference value for setting the subsequent temperature gating function. The rated temperature rise of the excitation coil is: ; where is the power loss of the excitation coil, in watts; is the thermal resistance of the excitation coil, in K / W; is the safety factor, with a value range of 1.2 to 1.5.

[0039] In the overall thermal-electrical coupling modeling framework of the present invention, the rated temperature rise of the excitation coil is regarded as the only constant used for the temperature gating function. It directly determines whether the gating denominator suppresses the amplitude of each historical coordinate during the step two delay embedding process, thus playing a decisive role in the sensitivity of the low-dimensional manifold's morphology and health index. The derivation of this formula follows the thermal equilibrium principle of electrical products: Under steady-state rated conditions, the copper loss inside the excitation coil is transferred to the outer surface of the winding in the form of thermal power , and then dissipates to the environment through the equivalent thermal resistance . When a stable temperature difference is formed between the heat source and the environment, the system enters thermal equilibrium. At this time, the increase in the winding temperature relative to the ambient temperature is the rated temperature rise. To ensure that the maximum temperature under long-term operation conditions is still lower than the limit allowed by the insulation class, a safety factor is introduced into the formula. It is equivalent to setting a margin on the product of the thermal resistance and the loss, making the actually measured temperature rise under daily conditions far lower than the insulation aging threshold. Among them is determined based on the IEC motor insulation standard and the aging test results of circuit breaker manufacturers, generally selected between 1.2 and 1.5. The power loss is given by the product of the rated excitation current and the coil's DC resistance. The DC resistance can be obtained by four-wire measurement at 20°C and then converted to the actual operating temperature according to the temperature coefficient of copper; when the circuit breaker needs to work in high altitude or high ambient temperature areas for a long time, engineers will introduce the actual maximum ambient temperature and air density at the site into the heat dissipation model to recalculate , so as to obtain a more conservative temperature rise value that conforms to the local climate conditions. The thermal resistance It is determined by the thermal conductivity between the winding and the shell, ventilation conditions, insulation varnish thickness and installation position. It can be obtained through heat flow meter testing or finite element thermal field simulation in the prototype stage, while the statistical average value of type tests is often used in the mass production stage. After substituting these parameters into the formula, it can be locked at one time. , and write it into the constant table of the embedded firmware; Under normal circumstances, the excitation coil of Class B insulation is obtained at a room temperature of 40°C The F grade is between 90K and 115K. Throughout the entire life cycle of the temperature gating function, it affects the degree of compression of the coordinate amplitude during high temperatures in summer and determines the sensitivity of the gating function to amplification of slight anomalies during low temperatures in winter. Therefore, it is necessary to correctly evaluate and and for Leaving a reasonable margin is critical to ensure that the model neither over-suppresses anomalies nor alarms prematurely. When the coil temperature rises due to changes in copper loss under extreme load conditions, the temperature control gate denominator increases and automatically decreases. The amplitude of the gating function is set to avoid the accumulation of high temperature noise into false fault signals in the embedding dimension; and when the ambient temperature returns to normal, the gating function quickly returns to the linear region of approximation to unity, so that the manifold coordinates can maintain high resolution for the real electrical anomaly.

[0040] The real-time health index quantitative assessment of the health status of the equipment adopts a multivariate coupling analysis method of the manifold intrinsic state quantity, insulation leakage current and arc voltage characteristics, specifically including: using the manifold intrinsic state quantity as the basic evaluation indicator of the equipment operation status, realizing the reverse correlation evaluation of the insulation performance by weighting the inverse square root of the insulation leakage current, and introducing a normalized exponential function of the arc voltage time series change rate relative to the rated working voltage of the circuit breaker to realize the exponential amplification of the arc characteristics on the real-time health index, and constructing a unified real-time health index that can comprehensively reflect the electrical health level of the equipment through the ternary product operation of the basic state, insulation weight and arc amplification factor. Real-time health index of time for: ; in, for Insulation leakage current over time, in mA; for Arc voltage at time, in V; for Arc voltage at time, in V; It is the rated working voltage of the circuit breaker, in V.

[0041] As a manifold internal state quantity obtained by delaying the embedding of the geometric mean, it has already integrated multi-dimensional factors such as the power transmission efficiency of the main circuit, the deterioration amplitude of the contact impedance, the hot state of the excitation coil, and harmonic injection into a single scale; numerically it corresponds to the nominal operating trajectory, and being significantly higher or lower than 1 means that the system dynamics has deviated from the healthy manifold. Secondly, the insulation leakage current is sampled at the milliamp level and enters the denominator. It is first added to the constant 1 and then square-rooted. Its design purpose is to soften the influence of small-scale micro-leakage through the square-root relationship while ensuring that the denominator is not zero when there is no leakage current; when the insulation ages or gets damp, resulting in a significant increase in the leakage current, the denominator increases, causing the overall exponent to decrease, thereby forming an immediate amplification of the deterioration of the insulation state numerically. The third channel is the exponential term driven by the arc voltage difference The molecular difference directly depicts the instantaneous rise or decay of the arc channel voltage from the previous sampling period to the current sampling period. This difference is then divided by the rated operating voltage to achieve cross-model normalization and then amplified by the exponential function into an exponential gain; the introduction of the exponential form can immediately make show an order-of-magnitude jump when high-energy events such as contact ablation or short-circuit breaking cause a sharp change in the arc voltage, while maintaining an approximately linear response to minor jitters and not causing false alarms.

[0042] The three physical channels show non-linear mutual restraint after being coupled by multiplication: if the flow form quantity slightly deviates due to load fluctuations and rises slightly, but the insulation leakage current is still low and the arc voltage fluctuates smoothly, then the denominator and the exponential term will suppress this slight deviation to ensure stays within the safety range; conversely, if contact deterioration and insulation aging occur simultaneously, that is, is high and increases, then although the denominator increases, it cannot completely offset the multiplication amplification. The exponential term will make things worse when the arc disturbance increases, thereby driving to quickly cross the threshold and trigger an alarm. This structure also naturally has time consistency because the difference period of is exactly the same as the manifold embedding delay, enabling the health index to jointly evaluate the dynamics and transient behavior using the same time window within each sampling step; at the same time, because all variables are directly measured by physical quantities or calibrated by nameplate parameters, so the calculation of does not require any on-site parameter adjustment. In engineering, to ensure that extreme scenarios such as high temperature, strong harmonics, or lightning strikes do not cause exponential explosion, the firmware implementation uses logarithmic domain accumulation to calculate the exponent and then clips the results greater than the preset saturation value. This not only maintains the numerical continuity within the boundary but also prevents extreme values from impacting the downstream protection logic. Combining the foregoing derivation of the threshold of the present invention, it can be seen that on a typical 12 kV pole-mounted circuit breaker, when When it exceeds approximately 0.70, it enters the alarm area. During the real-time operation process, the monitoring unit calculates cyclically with a fixed sampling period , and if it detects that the index breaks through the threshold, a high-priority interrupt is generated on the local MCU to trigger the tripping of the hard relay or upload the remote signal; at the same time, the system also performs a five-point sliding gradient calculation on

[0043] . If the gradient shows a continuous positive value and the amplitude exceeds the threshold of 20%, it can be determined that the fault is accelerating and deteriorating. At this time, the protection logic will execute the trip with a shorter delay, thereby suppressing the fault in the early stage. In this way, the health index not only gives a "good or bad" judgment in terms of static numerical values, but also provides information on the fault development rate through the continuous-time evolution curve, laying a data foundation for predictive maintenance, life assessment, and hierarchical response. The health threshold for equipment health warning is set based on the safety boundary determination principle under extreme operating conditions, specifically including: establishing an insulation safety boundary according to the maximum allowable leakage current level that the equipment insulation system can withstand, and realizing the inverse constraint relationship between the maximum allowable leakage current and the health threshold through the square root reciprocal transformation; at the same time, based on the maximum allowable arc voltage transient amplitude that the arc system can withstand, establishing an arc safety boundary, and realizing the exponential constraint relationship of the arc transient on the health threshold through the normalization index transformation relative to the rated operating voltage of the circuit breaker. Finally, the comprehensive health threshold is determined by the product of the insulation constraint and the arc constraint. The health threshold ; where is the maximum allowable leakage current; is the maximum allowable arc voltage transient.

[0044] This threshold formula is completely determined by the maximum allowable leakage current and the maximum allowable arc voltage transient that can be found on the nameplate. Its structure adopts the same molecular-denominator and exponential coupling method as the real-time health index: the in the denominator reflects the leakage power that the insulation system can tolerate under extreme operating conditions. The square root operation ensures that small leakage will not unnecessarily reduce the threshold value. When the leakage current reaches the nameplate limit, the denominator value just suppresses the threshold value, so that the measured leakage slightly higher than this limit can drive the index to break through the warning; the in the exponential term normalizes and compares the arc voltage rise limit specified by the manufacturer with the rated voltage of the equipment, and amplifies it in exponential form. When the actual arc difference approaches the limit, the exponential term grows to the upper limit required by the threshold. In this way, as long as any component in the real-time health index approaches the limit specification, will necessarily approach or exceed , thus triggering an early warning. The derivation logic follows the strictest thermal - electrical - insulation collaborative safety principle: If the insulation leakage has reached , even if the arc voltage is stable, the system is still judged as critical; conversely, if the arc voltage transient reaches , even if the leakage current is still at a low value, the threshold is also crossed, ensuring that the out - of - control of a single link is sufficient to trigger an alarm. Numerically, taking a typical 12 kV pole - mounted circuit breaker as an example, when and , substituting the rated voltage gives , which value is written into the read - only register during system initialization; when replacing the circuit breaker of the same specification on - site, there is no need to modify it. If it is replaced with a higher level such as 24 kV or 35 kV, only the nameplate parameters need to be updated to automatically calculate the new threshold, while the downstream protection logic and interface presentation remain completely unchanged. On the other hand, since the denominator and the exponential term have opposite adjustment effects on different physical phenomena, in actual operation, the health index often needs to trigger both channels simultaneously to cross the threshold, which provides a natural protection barrier against false alarms on - site; and when one of the two channels deteriorates sharply, the super - linear response of the exponential term or the denominator term can ensure that the fault is amplified in milliseconds, preventing potential hazards from being covered up. The simple form of the threshold formula also brings advantages at the firmware implementation level: The MCU only needs to perform a logarithmic and square - root operation once according to the three nameplate parameters saved in the EEPROM at startup, and then cache the threshold into the register. The subsequent comparison process only involves floating - point subtraction, which is very suitable for long - term unattended operation of resource - constrained edge controllers.

[0045] The prediction of the future evolution of the break - point phase voltage is based on the superposition prediction method of the voltage reference of the main circuit of phase A of the pole - mounted circuit breaker and the multi - physical - field coupling correction: Taking the voltage of the main circuit of phase A of the pole - mounted circuit breaker as the initial reference value for prediction, reflecting the dynamic change trend of the equipment state through the time evolution gradient of the real - time health index, constructing a time - varying correction factor by combining the prediction time step, vacuum permeability, copper conductivity, and the characteristics of the DC resistance of the excitation coil and the contact resistance of the contact, and introducing the ratio of the thermal capacity of the circuit breaker to the equivalent capacitance energy storage as the correction index for the energy conversion characteristics, calculating the correction amount of the voltage evolution through the multi - physical - field collaborative action of electromagnetic coupling, impedance evolution, and energy conversion, and finally realizing the quantitative prediction of the break - point phase voltage at a future moment through the superposition of the reference voltage and the correction amount. The prediction result is: ; where is the prediction time step, ; is the vacuum permeability; is the copper conductivity; is the DC resistance of the excitation coil at time, with the unit of Ω; is the thermal capacity of the circuit breaker, with the unit of J·K⁻¹; is the equivalent capacitance energy storage, with the unit of J·V⁻²; is the future time possible value of the broken-phase voltage.

[0046] By integrating the time gradient information, the electrothermal coupling constant, and the reference phase voltage into a single analytical expression, the formula gives the phase voltage that the circuit breaker contact may withstand within the future time window, and thus provides a decision-making basis for the automatic protection logic. The first term of the formula directly uses the current line voltage of phase A as the reference potential, meaning that the prediction evaluates the increment of additional risk at the existing bus voltage level; the second term introduces the time lead and the magnetic and electrical conductivity properties of the material with a proportionality factor , which physically reflects the combined restriction of the magnetic field energy accumulation rate and the heat dissipation ability of the metal conductor on the future arc reignition voltage. Among them, the vacuum permeability is an electromagnetic constant, while the copper conductivity represents the dissipation ability of the conductor in the excitation circuit to transient current; the time gradient captures the rate at which the health index rises or falls over time. If the gradient is positive and large, it indicates that the system state is rapidly deteriorating along the unstable direction, and the predicted voltage gain also increases accordingly; the proportional term before the gradient thus maps the "deterioration rate" to the "future voltage rise", making the prediction result not only depend on the current health level but also pay more attention to its change trend. The remaining power exponent factor then takes the ratio of the sum of the equivalent DC resistances of the drive winding and the contact to the internal state quantity of the manifold, and then exponentiates it with the ratio of the heat capacity to the equivalent capacitance energy storage . This structure comes from the logarithmic linearization derivation of the distribution rate of electromagnetic transient energy between thermal and electrical energy storage media: when the contact or coil resistance increases, the numerator increases accordingly, resulting in an overall increase in the exponent term, which will further raise the predicted voltage, reflecting the amplification effect of conduction degradation on the opening potential; if the system has a large heat capacity and a small energy storage capacitor, the power exponent is larger, indicating that in a device with weak thermal inertia and limited capacitance energy storage, once the resistance rises, it is more likely to cause a high voltage at the contact, so protection measures must be taken earlier. Conversely, when deviates from 1 and tends to decrease (mostly reflecting the continuous existence of the arc channel or the decrease in conduction efficiency), the denominator increases, which will suppress the exponent term, indicating that the system has entered a low-power transmission state. At this time, even if the resistance increases slightly, it will not significantly raise the contact voltage, so the algorithm automatically reduces the predicted value to avoid false alarms. Through this composite exponential form, the prediction model couples the electromagnetic energy storage rate, the heat dissipation ability, the contact resistance degradation rate, and the overall dynamic health trend. Without any external empirical factors, it provides a prediction for the future Quantitative risk assessment. At the implementation level, in each sampling period, the embedded firmware first calculates using the differential method , and then calls the power lookup table to complete the exponential operation, with the floating-point overhead limited within the allowable range of a single cycle; if exceeds the safety upper limit set by IEC62271-100 or user-defined, the system immediately enters the early protection state, and the lead time is the sum of the communication and execution delays, which can reserve sufficient time for bypass switching, voltage limit module input, or fast tripping. Overall, the prediction formula upgrades the traditional passive protection based on static threshold tripping to trend-based proactive prevention by mapping the current "state snapshot" and "rate of change" of the system to the quantifiable future voltage threats.

[0047] The present invention has been implemented on a 12 kV, 630 A pole-mounted circuit breaker. The sampling frequency is 1 kHz, and the delay step sampling points, and the intrinsic dimension of the manifold .

[0048] Current moment : , , , , , , .

[0049] ; One step before : Obtained in the same way . Two steps before : Obtained .

[0050] The temperature of the excitation coil is 345 K, and the rated temperature rise , Modulation selection . Thus: ; .

[0051] Insulation leakage current (per unit 0.0012), arc voltage .

[0052] Thus .

[0053] The nameplate allows a maximum leakage of 2 mA and a maximum arc transient of 40 V, so .

[0054] Currently it is only 0.057 higher, within the "first-level warning" zone, and the system issues an alarm first without tripping immediately.

[0055] ; ; Gradient approximation ; (Here 0.744 is obtained by the same method.) Contact resistance . Substitute into .

[0056] The voltage rise is only 2 V, far lower than 1.5 times the rated upper limit. Based on this, the automation device maintains operation but continuously monitors at a high frequency.

[0057] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring electrical variables of a pole-mounted circuit breaker, characterized in that, The method includes: Step 1: Collect multiple real-time electrical variables of the pole-mounted circuit breaker, perform normalization processing on each real-time electrical variable, and then non-linearly fuse them into a single-dimensional electrical variable observation sequence; Step 2: Obtain historical sampling values of the electrical variable observation sequence multiple times within a fixed time delay interval, and perform two non-linear processes on each historical sampling value respectively to obtain multiple scalar coordinates. Specifically, use a harmonic function proportional to the sampling frequency for phase modulation to introduce time periodicity; use a gating function composed of the temperature of the excitation coil and the rated temperature rise to weaken the noise influence in the high-temperature section; then compress all the scalar coordinates back into a single manifold internal state quantity in a geometric mean manner to avoid the explosion of the embedding dimension caused by vector operations and make the low-dimensional manifold hypothesis be explicitly reflected; Step 3: Combine and map the internal state quantity with the difference between the insulation leakage current and the arc voltage to construct a real-time health index; according to the comparison between the real-time health index and the health threshold, if the real-time health index exceeds the set health threshold, issue a warning; continuously monitor the real-time health index, calculate the instantaneous gradient of the real-time health index with respect to time, and predict the contact voltage at the future time.

2. The on-pole circuit breaker electrical variable monitoring method according to claim 1, wherein The real-time electrical variables include: the main circuit current of phase A of the pole-mounted circuit breaker, unit: A; the neutral line current, unit: A; the main circuit voltage of phase A of the pole-mounted circuit breaker, unit: V; the contact resistance of the contact, unit: Ω; the equivalent arc resistance, unit: Ω; the cumulative charge of the excitation coil, unit: C; the cumulative charge of the excitation coil at the previous historical time, unit: C; the sampling time delay step, unit: s, which is an integer multiple of the sampling period.

3. The on-pole circuit breaker electrical variable monitoring method according to claim 2, characterized in that, Step 1 specifically includes: characterizing the total current intensity through the vector synthesis of the main circuit current of phase A of the pole-mounted circuit breaker and the neutral line current, reflecting the instantaneous power level in combination with the main circuit voltage of phase A of the pole-mounted circuit breaker, and using the total series resistance of the contact resistance of the contact and the equivalent arc resistance as the loss factor of the power transmission path to form a quantization index of the instantaneous power transmission efficiency as the power efficiency index; at the same time, introduce a logarithmic sensitivity detection mechanism for the change rate of the cumulative charge of the excitation coil, perform a logarithmic transformation on the absolute value of the increment of the cumulative charge of the excitation coil and the cumulative charge of the excitation coil at the previous historical time to achieve amplified detection of small magnetic field disturbances and obtain the magnetic field sensitivity index; finally, linearly superimpose the power efficiency index and the magnetic field sensitivity index to form a single-dimensional electrical variable observation sequence.

4. The on-pole circuit breaker electrical variable monitoring method according to claim 3, wherein, The scalar coordinates are obtained through the following process: Establish a time periodicity detection mechanism based on the harmonic sequence of the sampling frequency, and achieve the frequency domain separation of historical data through the embedding frequencies corresponding to different delay embedding dimensions to ensure the independence of information at each historical moment in the frequency domain; Parallelly construct an adaptive noise suppression system based on the temperature of the excitation coil, establish a dynamic gating function using the ratio relationship between the temperature of the excitation coil and the rated temperature rise of the excitation coil, automatically reduce the signal weight under high temperature conditions to suppress thermal noise interference, and realize the conversion of the information of the historical electrical variable observation sequence to the current scalar coordinates through the ratio operation of the frequency domain modulation signal and the temperature gating function, where the intrinsic dimension of the manifold determines the number of historical time sample points required to construct the low-dimensional state.

5. The on-pole circuit breaker electrical variable monitoring method according to claim 4, characterized in that, The intrinsic state quantity of the manifold is obtained through the following process: Use the equal-weight geometric mean algorithm to process the scalar coordinates of all delay embedding dimensions, perform a fractional power operation after unitizing and biasing each scalar coordinate, and set the power exponent to the reciprocal of the intrinsic dimension of the manifold to ensure the balanced contribution of information in each dimension. Finally, through the product operation, realize the dimensionality reduction mapping from multi-dimensional discrete information to a single continuous intrinsic state quantity of the manifold, and this state quantity can achieve dimensionality compression while maintaining the integrity of the original information.

6. The on-pole circuit breaker electrical variable monitoring method according to claim 5, characterized in that, The rated temperature rise of the excitation coil is obtained through the following process: Based on the heat dissipation ability of the power loss of the excitation coil to the environment through the thermal resistance path under steady-state conditions, establish a linear thermal resistance relationship between the power loss and the temperature rise, and introduce a safety factor to conservatively correct the theoretically calculated temperature rise. The numerical range of the safety factor ensures that the system has sufficient thermal safety margin under rated conditions, providing a reference value for the subsequent setting of the temperature gating function.

7. The method for monitoring electrical variables of a pole-mounted circuit breaker according to claim 6, wherein, The real-time health index quantitative assessment of the equipment health status adopts a multi-coupling analysis method of the intrinsic state quantity of the manifold, insulation leakage current and arc voltage characteristics, specifically including: Using the intrinsic state quantity of the manifold as the basic evaluation index of the equipment operation state, realizing the reverse correlation assessment of the insulation performance through the reciprocal square root weighting of the insulation leakage current as the insulation weight, and introducing the normalized exponential function of the arc voltage time series change rate relative to the rated operating voltage of the circuit breaker as the arc amplification factor to realize the exponential amplification of the influence of the arc characteristics on the real-time health index. Through the ternary product operation of the basic evaluation index, insulation weight and arc amplification factor, construct a unified real-time health index that can comprehensively reflect the electrical health level of the equipment.

8. The on-pole circuit breaker electrical variable monitoring method according to claim 7, wherein The health threshold setting for equipment health warning is based on the safety boundary determination principle under extreme conditions, specifically including: Establish an insulation safety boundary according to the maximum allowable leakage current level that the equipment insulation system can withstand, and realize the reverse constraint relationship between the maximum allowable leakage current and the health threshold through the reciprocal square root transformation to obtain the insulation constraint; At the same time, based on the maximum allowable arc voltage transient amplitude that the arc system can withstand, establish an arc safety boundary, and realize the exponential constraint relationship between the arc transient and the health threshold through the normalized exponential transformation relative to the rated operating voltage of the circuit breaker to obtain the arc constraint; Finally, determine the comprehensive health threshold through the product of the insulation constraint and the arc constraint.

9. The on-pole circuit breaker electrical variable monitoring method according to claim 8, characterized in that, The prediction of the future evolution of the disconnector phase voltage is based on the superposition prediction method corrected by the coupling of the A-phase main circuit voltage reference of the pole-mounted circuit breaker and multi-physical fields: taking the A-phase main circuit voltage of the pole-mounted circuit breaker as the initial reference value for prediction, reflecting the dynamic change trend of the equipment state through the time evolution gradient of the real-time health index, constructing a time-varying correction factor by combining the prediction time step, vacuum permeability, copper conductivity, and the characteristics of the DC resistance of the excitation coil and the contact resistance of the contact, and introducing the ratio of the thermal capacity of the circuit breaker to the equivalent capacitance energy storage as the correction index of the energy conversion characteristic, calculating the correction amount of the voltage evolution through the collaborative action of multi-physical fields such as electromagnetic coupling, impedance evolution, and energy conversion, and finally realizing the quantitative prediction of the disconnector phase voltage at future moments through the superposition of the initial reference value and the correction amount.

Images