A transformer area overload early warning system based on intelligent fusion terminal

Abstract

This invention relates to the field of edge sensing and intelligent early warning technology for new power systems, specifically to an overload early warning system for distribution transformer areas based on an intelligent fusion terminal. By collecting three-phase voltage, three-phase current, and neutral current in the distribution transformer area, the system performs sequence component decomposition on the three-phase electrical parameters to construct unbalanced excitation parameters. Based on the unbalanced excitation parameters and neutral current, it calculates the thermal offset migration of the internal heat source of the transformer relative to the winding center, generating a thermal offset migration sequence and a thermal offset migration intensity index. Then, it combines the thermal offset migration to generate zoned thermal load values, which are input into the dynamic thermal circuit model of the distribution transformer to deduce the local additional thermal state of the transformer and output overload early warning information. This invention has a high ability to identify local abnormal heat accumulation caused by zero-sequence return current under three-phase unbalanced conditions, requires no modification to the transformer itself, and is suitable for low-cost intelligent upgrades of existing distribution transformer areas.

Description

Technical Field

[0001] This invention relates to the field of new power system edge sensing and intelligent early warning technology, and more specifically to a transformer area overload early warning system based on an intelligent fusion terminal. Background Technology

[0002] Distribution network areas often face frequent switching of single-phase nonlinear loads, such as high-power single-phase charging facilities and single-phase photovoltaic access. To ensure the safe operation of distribution transformers, current engineering projects mostly use intelligent integrated terminals in distribution areas for overload early warning on the edge side. The more advanced early warning schemes mainly rely on the terminal to collect three-phase electrical parameters and apply a dynamic thermal circuit model based on the IEC 60076-7 standard. Its conventional logic is to extract the maximum value of the three-phase current or the equivalent total apparent power as input parameters, and combine the ambient temperature to calculate the transformer top oil temperature and winding hot spot temperature rise, triggering an alarm or disconnection command before the heat accumulation reaches the insulation aging threshold.

[0003] However, when the overall load rate of the transformer area does not reach the rated upper limit, but the three-phase imbalance is extremely high, the existing thermal circuit algorithm will determine that the transformer is in the safe range. In fact, a large amount of zero-sequence current flows back through the transformer neutral line, which will induce strong stray eddy currents in non-current-carrying structural components such as the tank wall and iron core clamps, resulting in significant asymmetric additional heating. Since the existing prediction models are mostly based on the assumption of three-phase symmetry, when the transformer neutral point bushing or local insulating oil has been severely overheated or even deteriorated and boiled, the early warning system may fail to report, causing serious accidents such as local burn-out of the transformer. This monitoring blind spot can easily lead to overheating of structural components and local insulation damage of the transformer without triggering the early warning. If we try to avoid this monitoring blind spot by installing fiber optic temperature measurement inside the transformer or infrared temperature sensor outside, we will face extremely high power outage modification costs and hardware investment, and it will not be able to meet the engineering requirements of low-cost, power outage-free intelligent upgrade of a large number of existing transformer areas. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention discloses a distribution transformer overload early warning system based on an intelligent fusion terminal. It aims to solve the problem that existing distribution transformer overload early warning schemes are insufficient in identifying localized additional heating caused by zero-sequence current return when the overall load rate of the distribution transformer has not reached the rated upper limit and the three-phase imbalance is high, leading to missed localized overheating warnings. Simultaneously, it considers the engineering application needs of existing distribution transformers that do not require power outage modifications, constructing an edge-side thermal risk perception and overload early warning mechanism for three-phase unbalanced operation conditions. This mechanism continuously analyzes and dynamically simulates the asymmetric heat accumulation process inside the transformer caused by single-phase load fluctuations, providing early warning of localized overheating risks in the neutral point vicinity and structurally sensitive areas without adding temperature measurement devices to the transformer itself.

[0005] To achieve the above-mentioned technical effects, the present invention adopts the following technical solution: A distribution transformer overload early warning system based on an intelligent fusion terminal includes an intelligent fusion terminal installed on the low-voltage side of the transformer in the distribution transformer area. The intelligent fusion terminal includes an operation access module, an electromagnetic thermal analysis module, and an early warning output module. The operation access module is used to collect data on the three-phase voltage, three-phase current, neutral current, and ambient temperature of the distribution transformer area. The electromagnetic thermal analysis module includes a sequence component decomposition unit, an unbalanced excitation unit, a thermal bias migration unit, and a thermal state deduction unit. The sequence component decomposition unit is used to perform symmetrical component decomposition on the three-phase voltage and three-phase current to obtain positive-sequence, negative-sequence, and zero-sequence components. The unbalanced excitation unit is used to accumulate and calculate the zero-sequence component within a preset time window, and construct unbalanced excitation parameters characterizing the asymmetrical load state of the transformer substation by combining the negative-sequence component. The thermal bias migration unit is used to calculate the thermal bias migration amount of the internal heat source of the transformer relative to the winding center based on the unbalanced excitation parameters and the neutral current, and construct a thermal bias migration sequence based on the continuous time-series thermal bias migration amount. The thermal state deduction unit is used to input the thermal bias migration amount as an additional loss constraint term into the dynamic thermal circuit model of the distribution transformer to generate local additional thermal state quantities corresponding to the transformer neutral point region, metal structural component region, and insulating medium region. The early warning output module is used to output overload early warning information or load control commands for the corresponding power distribution area based on the local additional heat status.

[0006] Based on the above technical solution, the positive and beneficial effects of the present invention are as follows: 1. This invention utilizes the existing sampling capabilities of the intelligent fusion terminal in the transformer substation to correlate and analyze three-phase electrical parameters and neutral line current. It extends the traditional overload judgment based solely on the total load rate or three-phase equivalent load to the identification of local asymmetric additional thermal states. It can identify local abnormal heat accumulation caused by zero-sequence return current in advance when the overall load rate of the transformer has not reached the rated upper limit. Compared with the existing early warning methods that rely on the overall thermal model, this invention has a higher perception capability for local overheating risks in areas such as the neutral point lead-out area and the area adjacent to structural components, reducing the operational risk of local insulation degradation or structural component burnout without triggering an early warning.

[0007] 2. This invention couples the current disturbance characteristics under three-phase unbalanced operation with the evolution of internal thermal distribution in the transformer. By continuously tracking the thermal bias migration process and extrapolating local additional thermal states, it expands the original static early warning based on average thermal state into a dynamic thermal risk early warning oriented to spatial location. It can not only provide overload alarm results for the transformer area, but also output the location indication information of the corresponding risk area in the circumferential direction of the transformer. Based on this, maintenance personnel can quickly locate key inspection parts and carry out targeted load adjustment, live temperature measurement or maintenance, shorten the fault diagnosis path, and improve the accuracy of transformer area operation and maintenance and the efficiency of on-site handling.

[0008] 3. The invention relies on the edge deployment of the intelligent converged terminal in the distribution area. It does not require the addition of fiber optic temperature measurement devices inside the distribution transformer or infrared temperature sensing devices outside the oil tank. It does not involve power outages or box opening modifications and is suitable for large-scale direct access in existing distribution areas. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, wherein: Figure 1 This is a schematic diagram of the framework structure of the present invention; Figure 2 This is a schematic diagram of the timing changes in the operating data of the transformer substation according to the present invention; Figure 3 This is a schematic diagram of thermal bias migration under top view of the present invention; Figure 4 This is a schematic diagram illustrating the time-series analysis and dwelling segment identification of the thermal bias migration sequence of the present invention; Figure 5 This is a schematic diagram of the circumferential projection and angular partitioning of the thermal bias migration amount in this invention; Figure 6 This is a block diagram of the dynamic thermal circuit model structure of the present invention; Figure 7 This is a schematic diagram of the internal model structure of the thermal state deduction unit of the present invention; Figure 8 This is a schematic diagram of the overload warning judgment logic for the transformer area in this invention; Figure 9 This is a comparison curve of the local temperature rise prediction results of the present invention. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0011] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0012] This invention relates to a specific application environment architecture or hardware architecture required for the execution of a distribution network overload early warning system based on an intelligent fusion terminal. First, the specific application environment architecture or hardware architecture of this invention is described. This invention is deployed in the low-voltage side operating environment of a distribution network area and is applicable to scenarios such as pole-mounted distribution transformers, public distribution network box-type transformers, and distribution terminals with low-voltage side feeder acquisition capabilities. The main control unit of the system is an intelligent fusion terminal installed on the low-voltage outgoing line side of the distribution network. The intelligent fusion terminal is electrically connected to the low-voltage side busbar of the distribution transformer and is connected to a three-phase voltage acquisition circuit and a three-phase current acquisition circuit, respectively. The system includes a neutral current acquisition circuit; the three-phase voltage acquisition circuit acquires the phase voltages of phases A, B, and C; the three-phase current acquisition circuit acquires the load currents of phases A, B, and C; and the neutral current acquisition circuit acquires the neutral return current of the distribution transformer. The intelligent fusion terminal integrates an edge computing processor, a sampling synchronization module, a local storage module, and a communication interface module. The edge computing processor executes the overload warning logic described in this invention. The local storage module caches the operating data and state sequences within a continuous sampling period. The communication interface module interacts with the distribution automation master station or the transformer area management platform. The above terminal acquisition structure and communication links can all adopt the conventional configuration of existing transformer area intelligent fusion terminals, which have been described in detail in existing technologies and will not be repeated here.

[0013] In practice, the low-voltage side load of distribution transformers typically exhibits significant asymmetrical connection characteristics, especially in residential areas, areas with concentrated charging facilities, and areas with distributed photovoltaic power. High-power single-phase loads and single-phase power supply devices frequently switch on and off at different times. Under these operating conditions, the total three-phase load does not necessarily reach the upper limit of the distribution transformer's rated capacity. However, zero-sequence current continuously flows back through the neutral line, and this, combined with negative-sequence disturbances, affects the internal magnetic field distribution of the transformer, causing the leakage flux path to shift relative to the winding center. This results in asymmetrical additional heating at the neutral point, core clamping points, and areas near the tank. Correspondingly, the most common overload warning schemes in current engineering projects mainly rely on intelligent integrated terminals to collect three-phase data. After inputting the phase voltage and current, the maximum three-phase current, average load rate, or total apparent power is input into the dynamic thermal circuit model to estimate the top oil temperature and the temperature rise of winding hot spots. Based on the overall heat capacity level, an early warning judgment is made. This type of scheme has good applicability under conditions of increased overall load or continuous full load. However, the process of establishing its thermal model is generally based on the three-phase approximate equilibrium condition. The model input side focuses on the overall heat load level rather than the spatial distribution of local additional heat sources. Therefore, when the overall load rate of the transformer area is within the safe range but the degree of three-phase imbalance increases significantly, although the overall thermal state of the transformer does not show obvious abnormalities, additional stray losses have accumulated in local areas. This part of the heat has not been effectively entered into the early warning link of the traditional overall thermal model, resulting in obvious monitoring blind spots.

[0014] Specifically, the intelligent fusion terminal first performs sequence component decomposition on the three-phase current and three-phase voltage in the transformer area operation data, and extracts the three-phase unbalanced excitation information in combination with the neutral line current. Then, using the unbalanced excitation information as the driving quantity, it derives the offset migration process of the heat source inside the transformer relative to the winding center, transforming the originally unobservable local additional heat source position change into a continuously trackable thermal offset migration sequence. On this basis, the thermal offset migration result is then mapped to the circumferential structural region of the transformer winding to form the corresponding zoned thermal load, and coupled to the dynamic thermal circuit model of the distribution transformer for local thermal state deduction. Thus, the traditional dynamic thermal circuit model, which only performs thermal calculation on the overall load, is extended into a dual-channel thermal analysis path that simultaneously considers the overall load thermal state and the local unbalanced additional thermal state.

[0015] In other words, the key technical point of this invention is to utilize the existing sampling conditions of the intelligent fusion terminal in the distribution area to continuously extrapolate the migration of local heat sources inside the transformer that cannot be directly measured. This transforms the additional losses caused by zero-sequence current from an unobservable state quantity into a thermal load input that can participate in the thermal model calculation. In some more specific scenarios, when single-phase charging loads are frequently put into or taken out of operation in a short period of time, the overall load rate of the distribution area changes only slightly, but the neutral line return path will be rapidly reconfigured. At this time, this invention can still identify the thermal bias migration trend and lock the local thermal risk concentration area through continuous sampling data on the terminal side without increasing the modification of the distribution transformer itself.

[0016] It should be noted that the aforementioned thermal bias migration process, heat accumulation process, and local additional thermal state deduction process are not isolated in this invention, but rather are a collaborative analysis of the same transformer area's operating conditions in both time and space dimensions: namely, sequence component decomposition corresponds to the decomposition process of electrical side operation disturbances in the transformer area; thermal bias migration corresponds to the spatial offset process of additional heat sources inside the transformer; and local thermal state deduction corresponds to the heat accumulation process of additional heat inside the transformer structure. These three together constitute a closed-loop deduction link from current imbalance input to local thermal risk output. Based on this, this invention provides continuous sensing capability for local overheating risks that traditional overall thermal early warning cannot cover, without changing the structure of the distribution transformer or adding new built-in temperature measuring elements, while also taking into account the controllability of engineering investment and the requirements for early warning accuracy under the condition of batch deployment of existing equipment in the transformer area.

[0017] Having understood the overall system architecture described above, the specific implementation process of the transformer overload early warning system of this invention will be explained in detail below with reference to the accompanying drawings and specific embodiments. For details, please refer to [link / reference needed]. Figure 1As shown, an overload early warning system for distribution transformer substations based on an intelligent fusion terminal includes an intelligent fusion terminal installed on the low-voltage side of the transformer in the distribution transformer substation. The intelligent fusion terminal includes an operation access module, an electromagnetic thermal analysis module, and an early warning output module. In a specific system deployment embodiment, the operation access module is installed in the outgoing line area of ​​the low-voltage side of the transformer in the distribution transformer substation and integrated into the edge acquisition motherboard inside the intelligent fusion terminal. It serves as the data input front end for the subsequent electromagnetic thermal analysis module. The operation access module forms electrical connections with the A-phase, B-phase, and C-phase busbars and the neutral line on the low-voltage side to continuously acquire three-phase voltage, three-phase current, neutral line current, and ambient temperature data during the operation of the distribution transformer substation. The three-phase voltage signal is preferably acquired through an isolation voltage sampling unit, and the three-phase current and neutral current are preferably acquired through a through-core current transformer, Rogowski coil, or Hall current sensor. The ambient temperature data is acquired by a temperature detection unit located inside the housing of the intelligent fusion terminal or inside the distribution box, which is used to reflect the heat transfer boundary conditions of the air around the transformer. As a preferred industrial embodiment, the three-phase voltage and three-phase current adopt a synchronous sampling mechanism to ensure that the phase quantities maintain a unified phase reference during the subsequent sequence component decomposition process. However, in some optional embodiments with high requirements for terminal power consumption and edge computing power, a time-division polling sampling method can also be adopted, and the subsequent phasor reconstruction can be completed through a timestamp alignment mechanism.

[0018] Furthermore, the access module is equipped with a timing synchronization unit to perform unified time reference correction on the data from different sampling channels. This is because, under conditions of drastic load changes in the transformer area or frequent start-stop of single-phase impact loads, if there is time drift in the three-phase sampling, it will cause pseudo-unbalance in the subsequent calculation of negative sequence components and zero sequence components, which will lead to the erroneous amplification of the unbalance excitation parameters. Therefore, performing unified time-scale alignment on the data of each channel during the acquisition stage can effectively avoid the coupling effect of sampling phase deviation on the subsequent calculation of thermal offset migration.

[0019] In the specific data flow process, the operation access module maps the collected three-phase voltage data to the real-time power supply phase state of the low-voltage side of the distribution substation, maps the three-phase current data to the current distribution state of each phase load branch, maps the neutral line current to the asymmetric return current intensity state of the substation, and maps the ambient temperature data to the external heat dissipation boundary parameters in the dynamic thermal circuit model. That is, this implementation does not just perform ordinary sampling of the original electrical parameters, but converts the low-voltage side operating state into a structured state input set that can be called by the subsequent electromagnetic thermal analysis module. It should be noted that the neutral line current is not used as an ordinary line load parameter in this application, but as an important state quantity for identifying the degree of zero-sequence return current accumulation and participating in the subsequent thermal bias migration calculation. Since the connection of residential single-phase loads, distributed charging equipment and nonlinear inverter loads may form a continuous zero-sequence return current path in the substation, which will lead to local additional losses near the transformer neutral point. Therefore, by independently collecting the neutral line current, the underlying excitation basis can be provided for the subsequent spatial offset deduction of heat source.

[0020] In a preferred embodiment, the operation access module further includes a data preprocessing unit, which performs de-glitch filtering, abnormal mutation suppression, and power frequency synchronization interception on the acquired raw data. The de-glitch filtering is used to eliminate transient spike interference caused by power distribution switch operation, capacitor switching, and lightning strike induction. The abnormal mutation suppression is used to avoid misjudgment of the unbalanced excitation accumulation result caused by short-term surge current. The power frequency synchronization interception is used to ensure that the subsequent sequence component decomposition window is consistent with the power frequency period, thereby reducing the impact of spectrum leakage on the resolution accuracy of negative sequence components.

[0021] Please see Figure 2 The figure shows the time-series changes of each sampled quantity within a typical operating cycle of the transformer substation. The horizontal axis represents the operating time t, and the vertical axis represents the amplitude of the corresponding sampled quantity, including the A-phase current. Phase B current C-phase current and neutral current The change curve shows that, in the residential electricity consumption scenario simulated in this embodiment, as single-phase loads are gradually put into operation at night, the current of each phase shows an uneven growth trend.

[0022] In the first stage, the three-phase load is in a relatively balanced state, and the current curves of phases A, B, and C show basically the same trend. The neutral line current... Maintaining a low level, after entering the second stage, single-phase charging facilities are centrally connected to the A-phase side, and the A-phase current... The current rises significantly, with smaller changes in phases B and C. The neutral current rises synchronously. At this point, although the total load factor of the distribution area has not exceeded the rated operating capacity, the load distribution between phases has shifted. After entering the third stage, some residential photovoltaic systems are taken out of grid-connected operation, and the neutral current... Another fluctuation peak appeared.

[0023] As a preferred industrial embodiment, the data sampling period of the operation access module is set to 20ms to 200ms; however, in some alternative embodiments with different requirements for dynamic response speed, the sampling period can also be adjusted according to the load fluctuation characteristics of the distribution area. It should be understood that the specific sampling period defined here is only an example to adapt to the typical load fluctuation scenario of the low-voltage distribution area and does not constitute a limitation on the core electromagnetic thermal analysis logic of this application.

[0024] Furthermore, the electromagnetic thermal analysis module includes a sequence component decomposition unit, an unbalanced excitation unit, a thermal bias migration unit, and a thermal state deduction unit. First, it should be noted that in this embodiment, the continuous full-sample data is not analyzed at a fixed period. Instead, the three-phase voltage and three-phase current sampling sequences corresponding to the continuous variation range of the neutral current are used as the input for symmetrical component decomposition. The reason for this is that after a large number of single-phase residential loads, charging equipment, and inverter loads are randomly connected in the low-voltage distribution substation, the neutral return current state usually exhibits obvious phased changes. If a fixed-length time window is used to directly perform sequence component decomposition, it is easy to superimpose and average the electrical characteristics under different load conditions, resulting in a decrease in the ability of the negative-sequence and zero-sequence components to represent the true unbalanced state.

[0025] Therefore, in this embodiment, the sequence component decomposition unit first identifies the continuity of the neutral line current change trend. When the neutral line current changes in the same direction for multiple consecutive sampling periods, or remains within a preset fluctuation range, the corresponding time period is defined as the same neutral line current continuous change interval. When the neutral line current experiences a sudden reversal, step drift, or change of fluctuation direction, a new analytical interval is redefined.

[0026] Furthermore, this interval division method essentially uses the "zero-sequence return current stable stage" as the analytical boundary of the sequence component, ensuring that the corresponding load topology relationship within the same analytical window is basically consistent. That is, within the same neutral line current continuous change interval, the single-phase load distribution state, neutral point return current intensity, and three-phase coupling relationship within the transformer area have continuous evolution characteristics, thereby avoiding the discrete drift of the sequence component results caused by frequent switching of operating states.

[0027] In specific implementation, the sequence component decomposition unit can adopt a phasor analysis method based on discrete Fourier transform to extract synchronous phasors of three-phase voltage and three-phase current, and further obtain positive sequence component, negative sequence component and zero sequence component according to the symmetrical component transformation relationship. Among them, the positive sequence component is used to characterize the reference operating state under normal symmetrical power supply conditions; the negative sequence component is used to characterize the reverse rotating magnetomotive force distribution state caused by three-phase load imbalance; and the zero sequence component is used to characterize the neutral point return current accumulation state and the in-phase current distribution state.

[0028] As mentioned above regarding the underlying logic of the thermal bias migration formation mechanism, the extraction of negative sequence and zero sequence components in this application is not for conventional three-phase balance detection, but rather as the basic input for the subsequent construction of unbalanced excitation parameters. The negative sequence component corresponds to the asymmetric flux distribution state inside the transformer, which easily leads to the enhancement of local leakage flux paths; the zero sequence component corresponds to the zero sequence circulating current accumulation state near the neutral point, which easily causes the accumulation of additional losses in the neutral region. When the two work together, the heat source distribution inside the transformer will shift relative to the geometric center of the winding.

[0029] In a preferred embodiment, the sequence component decomposition unit further establishes corresponding time stamp information for each sequence component, enabling the subsequent unbalanced excitation unit to maintain a unified temporal state analysis based on a continuously changing interval. That is, the formation of the subsequent thermal bias migration amount is not based on a single instantaneous sampled value, but on the sequence component evolution process under continuous operation. As a preferred industrial embodiment, the sequence component decomposition unit adopts a power frequency synchronous sampling mechanism to reduce the impact of spectrum leakage on the accuracy of negative sequence analysis. However, in some optional embodiments with high requirements for edge real-time performance, a sliding recursive phasor algorithm or a fast iterative decomposition method can also be used to realize online sequence component calculation. It should be understood that the specific decomposition algorithm defined here is only an example adapted to typical low-voltage distribution substation operating conditions and does not constitute a limitation on the core sequence component analysis logic of this application. This method can effectively avoid the coupling interference caused by short-term impact loads, transient surge currents, and random start-stop disturbances on the analysis results of negative sequence components and zero sequence components, and improve the continuity of subsequent unbalanced excitation state identification.

[0030] After obtaining the negative-sequence and zero-sequence components of the corresponding neutral line current continuous variation range by the aforementioned sequence component decomposition unit, the unbalanced excitation unit further performs coupling analysis on the zero-sequence and negative-sequence components to construct unbalanced excitation parameters characterizing the asymmetrical load state of the transformer area. It should be noted that the "unbalanced excitation parameters" are not the three-phase unbalance index in the ordinary sense. They are not directly used to evaluate the power distribution quality, but are used to characterize the comprehensive excitation intensity of the low-voltage side asymmetrical load on the transformer's internal additional leakage flux loss, zero-sequence return current loss, and local heat source offset trend. Therefore, this parameter not only needs to reflect the continuous accumulation degree of zero-sequence return current, but also needs to reflect the influence of negative-sequence magnetomotive force disturbance on the stability of the internal flux distribution.

[0031] Based on this, the unbalanced excitation unit first performs cumulative calculations on the zero-sequence component within a preset time window. In specific implementation, the unbalanced excitation unit establishes a sliding time window according to the continuous sampling period and obtains the effective value of the zero-sequence current corresponding to each sampling moment within the sliding time window. The zero-sequence current sequence is formed according to the sampling order. The sliding time window preferably adopts a continuous recursive update method to ensure that the subsequent unbalanced excitation state can reflect the current operating trend of the transformer area. Furthermore, this embodiment does not directly perform overall integration on the zero-sequence current within the entire time window, but first calculates the change in the effective value of the zero-sequence current between adjacent sampling moments in the zero-sequence current sequence and extracts the local extreme points.

[0032] The reason is that, in the actual operation of residential transformer substations, electric heating equipment, charging equipment and intermittent single-phase loads will cause the zero-sequence return current to exhibit obvious pulsating characteristics. If the zero-sequence current of the entire window is directly integrated continuously, it is easy to treat short-term impacts and continuous off-center loads as equivalent, thereby amplifying the impact of transient disturbances on subsequent thermal state deduction.

[0033] Therefore, this implementation identifies the trend reversal point in the zero-sequence return current state by using local extreme points. That is, when the change in the effective value of the zero-sequence current changes from growth to decay, or from decay to growth, the corresponding sampling time is identified as a local extreme point. The sampling time corresponding to each local extreme point is used as a boundary to divide multiple cumulative segments. The essence of the above segment division is to further divide the continuous sliding time window into multiple "zero-sequence return current evolution stages". Each cumulative segment corresponds to a relatively stable zero-sequence current carrying process, thereby avoiding mutual cancellation or abnormal superposition between different operating states. Subsequently, the unbalanced excitation unit performs discrete integration on the effective value of the zero-sequence current in each cumulative segment and accumulates them sequentially according to the time order of each cumulative segment to obtain the unbalanced excitation base quantity of the corresponding sliding time window. The discrete integration is essentially used to characterize the degree of accumulation of zero-sequence return current energy in the time dimension. Since the zero-sequence return current corresponds to the continuous current carrying state near the neutral point, the cumulative integration result can better reflect the formation trend of the additional heat source inside the transformer compared to a single instantaneous current value.

[0034] In a preferred embodiment, the discrete integral adopts a sampling period weighted integral method to increase the proportion of long-term continuous off-center load state in the cumulative result; however, in some alternative embodiments with lower requirements for edge computing resources, an equal-interval discrete summation method can also be used to complete the zero-sequence cumulative calculation. It should be understood that the integral form limited here is only an example adapted to typical low-voltage distribution area fluctuation scenarios and does not constitute a limitation on the core zero-sequence excitation cumulative logic of this application.

[0035] After obtaining the unbalanced excitation base quantity, the unbalanced excitation unit further couples it with the effective value of the negative sequence current to generate unbalanced excitation parameters. Specifically, the unbalanced excitation unit first determines the zero-sequence current carrying capacity coefficient based on the ratio of the unbalanced excitation base quantity to the preset time window length. This coefficient characterizes the degree of sustained accumulation of zero-sequence return current per unit time, reflecting the cumulative intensity of the long-term asymmetric current carrying state in the neutral point region. Subsequently, the unbalanced excitation unit obtains the rate of change of the effective value of the negative sequence current within a continuous sampling period and calculates the negative sequence disturbance coefficient. in, This represents the effective value of the negative sequence current at the current sampling time. The effective value of the negative sequence current at the previous sampling time is given. It should be noted that the negative sequence disturbance coefficient introduced in this embodiment is not for simply amplifying the negative sequence component, but for characterizing the degree of dynamic disturbance during the change of negative sequence magnetomotive force. The reason is that the effects of a continuously stable unbalanced load and a frequently fluctuating unbalanced load on the leakage flux distribution inside the transformer are different. When the rate of change of the negative sequence current is small, it indicates that the current unbalanced state is relatively stable, and the spatial position of the internal additional heat source changes slowly. However, when the negative sequence current changes rapidly in a short period of time, it means that the reverse magnetomotive force distribution inside the winding is undergoing dynamic migration, which can easily induce local leakage flux path reconstruction and fluctuations in additional eddy current losses. Therefore, by dynamically modulating the zero sequence current intensity through the negative sequence disturbance coefficient, the final unbalanced excitation parameters can simultaneously possess the following characteristics: The characteristics of continuous accumulation of zero-order reflux and dynamic disturbance of negative-order magnetic field.

[0036] Finally, the unbalanced excitation unit uses the product of the zero-sequence current carrying capacity coefficient and the negative-sequence perturbation coefficient as the unbalanced excitation parameter and sends it to the subsequent thermal bias migration unit. This can effectively avoid the false triggering problem caused by judging the thermal state based solely on the instantaneous negative-sequence value or the instantaneous zero-sequence value, and improve the mapping stability of the subsequent thermal bias migration amount to the actual local additional heat source formation process.

[0037] The thermal bias migration unit further calculates the thermal bias migration of the internal heat source relative to the winding center based on the unbalanced excitation parameters and neutral current, and constructs a thermal bias migration sequence based on the continuous time-series thermal bias migration. It should be noted that the "thermal bias migration" is not a directly measured temperature parameter, nor is it the average temperature rise value in the ordinary thermal circuit model. Instead, it is a state quantity used to describe the degree of offset of the spatial distribution center of the additional heat source inside the transformer relative to the geometric center of the winding. In essence, it is a spatialized expression of the internal leakage flux path, zero-sequence return current path, and local additional loss concentration trend under asymmetrical load conditions. Since the traditional distribution transformer thermal model usually assumes that copper loss, iron loss, and additional loss are approximately uniformly distributed in space, it can only reflect the overall average temperature rise state. However, under conditions of continuous single-phase load imbalance, long-term zero-sequence return current, and dynamic disturbance of negative-sequence magnetomotive force, the internal heat source of the transformer is often not symmetrically distributed along the winding center, but rather spatially offsets towards the neutral point region, clamp region, or local leakage flux concentration region. Such heat source offset may form local hot spot accumulation even before the overall load rate reaches the alarm threshold.

[0038] Please see Figure 3 The diagram shown is a top-view illustration of thermal bias migration in a transformer enclosure. The outer contour corresponds to the transformer enclosure boundary; the central area corresponds to the main winding area; the arrow direction indicates the direction of thermal center migration; and the arrow length indicates the amplitude of thermal bias migration. Figure 3In the initial state shown on the left, the three-phase load of the transformer substation is in a relatively balanced state. The heat inside the transformer is mainly concentrated in the winding area, and the heat center is located near the geometric center of the winding. As the single-phase load continues to operate under off-center load, the return current of the neutral line increases, the leakage flux distribution inside the transformer shifts, and the local additional heat begins to accumulate in the neutral point lead-out area and the area of ​​adjacent structural components. Correspondingly, the heat center in the figure shifts from the initial position to one side of the tank along the direction of the arrow. When the transformer substation returns to three-phase balance, the thermal offset migration gradually decreases, and the heat center returns to the winding center.

[0039] Therefore, the thermal bias migration unit in this invention does not directly judge the thermal state based on the total load, but first establishes a mapping relationship from "electrical imbalance state" to "internal heat source spatial offset state". Specifically, the thermal bias migration unit first obtains the phase offset relationship of the three-phase current on the low-voltage side of the transformer within the corresponding sampling period based on the unbalanced excitation parameters. The phase offset relationship here is used to reflect the spatial distribution direction of the current asymmetric magnetomotive force on the winding cross section. That is, when the load of a certain phase is too high for a long time or the zero-sequence return current is continuously concentrated, the phase coupling relationship between the currents of each phase will shift, which will cause the internal additional magnetic flux to no longer be uniformly distributed around the winding center.

[0040] Furthermore, the thermal bias migration unit combines the amplitude and direction of the neutral current to construct an additional magnetic flux action axis corresponding to the transformer winding cross-section. The amplitude of the neutral current is used to characterize the zero-sequence return current accumulation intensity, and the direction of the neutral current is used to characterize the additional magnetic flux offset trend. When the neutral return current continues to increase, it indicates that the zero-sequence current on the low-voltage side is forming a stable accumulation path. At this time, the leakage magnetic field distribution inside the transformer will gradually deviate from the normal symmetrical state and form an additional magnetic flux concentration area in the local winding. In the specific system deployment and implementation, the additional magnetic flux action axis can be understood as the equivalent projection axis of the main action direction of the additional leakage flux in the current sampling period on the winding cross-section. Its direction is determined by the phase offset relationship of the three-phase current, and its offset intensity is modulated by the unbalanced excitation parameters and the neutral current.

[0041] As a preferred embodiment, the thermal bias migration unit can pre-establish a two-dimensional coordinate model of the transformer winding cross-section, and use the winding geometric center as the reference point. The two-dimensional coordinate model is preferably constructed based on the transformer winding structural dimensions, phase arrangement position, and neutral point connection position.

[0042] Furthermore, the thermal bias migration unit determines the bias direction of the additional magnetic flux based on the spatial relationship between the axis of action of the additional magnetic flux and the reference point. It should be noted that the "bias direction" essentially reflects which local area the internal additional heat source gathers in. Since the path of the additional leakage flux is not fixed under different load bias conditions, the spatial offset direction of the heat source will also change. For example, when a single-phase load is concentrated in a certain phase for a long time, the additional magnetic flux usually shifts to the corresponding adjacent winding area; while when the zero-sequence return current is concentrated near the neutral point for a long time, the additional magnetic flux is more likely to gather in the neutral region.

[0043] After determining the direction of the additional magnetic flux bias, the unit calculates the radial offset distance between the axis of action of the additional magnetic flux and the reference point along the direction of the additional magnetic flux bias, and records this radial offset distance as the thermal offset migration amount of the current sampling period. Here, the radial offset distance represents the equivalent offset degree of the center of the additional heat source relative to the theoretical symmetrical heat center. That is, the larger the offset distance, the more uneven the spatial distribution of the internal additional loss, and the higher the risk of continuous heat accumulation in the local area. Furthermore, the introduction of thermal offset migration amount in this embodiment makes the subsequent thermal state deduction no longer rely solely on the total load rate, but can identify the early state in which continuous additional heat accumulation has already formed in the local area before the overall load is overloaded.

[0044] After obtaining the thermal offset migration amount corresponding to the current sampling period, the thermal offset migration unit further obtains multiple thermal offset migration amounts corresponding to consecutive sampling periods and arranges them in chronological order to form a thermal offset migration sequence. It should be noted that what is actually used for subsequent thermal risk identification in this application is not a single instantaneous thermal offset migration amount, but the evolution trend of the thermal offset migration amount in the continuous time dimension. This is because the formation of local hotspots is usually not caused by a single instantaneous offset, but by the long-term continuous residence of an additional heat source in the same area. Therefore, the thermal offset migration unit further performs time-series analysis on the thermal offset migration sequence.

[0045] Specifically, the thermal bias migration unit acquires the bias azimuth angle corresponding to each thermal bias migration amount within a continuous sampling period, and calculates the angle difference between the bias azimuth angles corresponding to adjacent sampling periods. The bias azimuth angle is used to characterize the spatial orientation change of the additional heat source on the winding cross section; the angle difference is used to reflect whether the additional heat source is in a continuous and stable residence state.

[0046] Please see Figure 4 The diagram shown illustrates the timing analysis and residence segment identification of the thermal bias migration sequence, as follows: Figure 4As shown in the upper part, the horizontal axis represents the continuous sampling time, and the vertical axis represents the thermal bias migration amount L. The thermal bias migration amount is used to characterize the equivalent offset degree of the center of the additional heat source inside the transformer relative to the winding center within the current sampling period. With the changes in the load structure of the transformer area, the zero-sequence return current intensity, and the degree of negative-sequence magnetomotive force disturbance, the thermal bias migration amount will continuously evolve over time, thereby forming a corresponding thermal bias migration sequence. It should be noted that in this invention, the thermal bias migration amount corresponding to a single sampling moment can only reflect the spatial offset state of the internal heat source at the current moment, and it is difficult to directly determine whether local heat has formed a continuous accumulation. Therefore, this embodiment further utilizes the thermal bias evolution trajectory in the continuous time dimension to analyze the spatial residence characteristics of the heat source.

[0047] like Figure 4 In the middle section, after obtaining the thermal offset migration amount corresponding to each sampling period, the thermal offset migration unit further determines the offset azimuth angle θ corresponding to each thermal offset migration amount. The offset azimuth angle is used to represent the spatial orientation of the thermal offset migration direction relative to the winding cross-section reference coordinate system. In specific implementation, the winding cross-section coordinate model can be used as a unified reference coordinate system, and the geometric center of the winding can be used as the origin of the polar coordinates. For any sampling period, the corresponding offset azimuth angle is determined according to the angle between the axis of action of the additional magnetic flux and the reference coordinate axis. Subsequently, the thermal offset migration unit calculates the angle difference Δθ between the offset azimuth angles of adjacent sampling periods and forms the corresponding angle difference sequence. Since the thermal offset migration amount may be affected by load fluctuations and thus change in amplitude, it is difficult to accurately determine whether the heat source is continuously acting on the same area based solely on the magnitude of the thermal offset migration amount. However, the offset azimuth angle can directly reflect the spatial position change of the heat source. Therefore, this invention uses the continuous stability of the offset azimuth angle as the basis for thermal residence identification.

[0048] like Figure 4 As shown in the lower part, when the offset azimuth angles corresponding to multiple consecutive sampling periods remain in similar directions, the angle difference Δθ between adjacent sampling periods will remain small. Conversely, when the spatial position of the heat source shifts significantly, the angle difference Δθ will increase significantly. Therefore, in this embodiment, when the angle difference between adjacent offset azimuth angles within 5 consecutive sampling periods is no greater than 15°, the corresponding time period is determined to be the same thermal offset residence segment. The determination condition of 5 consecutive sampling periods is used to eliminate the accidental stability phenomenon caused by instantaneous disturbances; the 15° angle threshold is used to limit the allowable fluctuation range of the spatial position of the heat source. When the thermal offset direction remains within this range, it can be considered that the additional heat source continuously acts on the same local area, such as... Figure 4 Sections 1, 2, and 4 all meet the above dwell criteria and are therefore identified as hot offset dwelling sections; while the offset azimuth angle changes more frequently in section 3 and the difference between adjacent angles exceeds the preset threshold, so it is determined to be a non-dwelling section.

[0049] Furthermore, for each identified thermal bias residence segment, the thermal bias migration unit records the corresponding start sampling time, end sampling time, and duration. The start sampling time corresponds to the sampling point where the residence condition is first met; the end sampling time corresponds to the sampling point where the residence state ends; and the duration is the cumulative duration between the two. The recorded residence duration will serve as an important input for subsequent calculations of the thermal bias migration intensity index. This is because the formation of local hotspots is related not only to the magnitude of the heat source offset but also to the duration of the heat source's continuous action in the same area. When an additional heat source resides in the same local area for a long time, the heat in that area is difficult to diffuse to the surrounding area in a timely manner, easily leading to a continuous accumulation of local temperature rise.

[0050] Specifically, when the angular difference between adjacent offset azimuth angles within five consecutive sampling periods is no greater than 15°, the corresponding sampling period is determined as the same thermal offset dwell segment. The reason for setting a continuous angle stability criterion here is: If the direction of the additional heat source continuously changes, it indicates that the internal additional losses are in a dynamic, shifting state, making it easier for heat to diffuse between different areas. Conversely, when the additional heat source remains near the same spatial direction for a long period, it means that the additional losses are continuously acting on a fixed local area, making it easier to form local heat accumulation and insulation aging. Therefore, this implementation does not only determine whether thermal bias exists, but further identifies whether thermal bias continuously resides in a fixed area. The thermal bias migration unit records the start sampling time, end sampling time, and duration of the corresponding thermal bias residence segment for subsequent heat accumulation level determination. Furthermore, the thermal bias migration unit calculates the thermal bias migration intensity index based on the thermal bias residence segment. Where M is the thermal bias migration intensity index; This represents the thermal bias migration amount corresponding to the i-th sampling period; The time interval between adjacent sampling periods is n; n is the total number of samples within the statistical period. The cumulative residence time of the thermal bias residence segment within the statistical period is denoted by , and T is the total duration of the statistical period. It should be noted that the above thermal bias migration intensity index does not simply represent the magnitude of the thermal bias, but rather couples the thermal bias change rate and the thermal bias residence persistence. Among them, the thermal bias change rate is used to characterize the degree of activity of the additional heat source migration, while the residence time proportion is used to characterize the stability of the continuous effect of the additional heat source in the local area.

[0051] When the migration rate is high but the residence time is short, it indicates that the additional heat source is in a dynamic drifting state and its local heat accumulation effect is limited. However, when the migration direction is stable for a long time and the residence is continuous, it indicates that the additional heat source is continuously superimposed in a fixed area, which is more likely to form a local thermal aging risk. Therefore, by introducing both the rate of change and the residence time, we can avoid the misjudgment of short-period disturbances by the traditional instantaneous hot spot judgment method.

[0052] In a preferred embodiment, the thermal bias migration unit classifies the thermal accumulation level of the corresponding thermal bias residence section according to the thermal bias migration intensity index and outputs it to the thermal state inference unit. The thermal accumulation level can be divided into low accumulation state, medium accumulation state and high accumulation state, which correspond to different degrees of local additional heat accumulation trend. As a preferred industrial embodiment, the thermal accumulation level threshold can be adaptively adjusted according to the transformer capacity level, winding structure and historical operating conditions. However, in some optional embodiments with lower requirements for edge deployment complexity, a preset fixed level threshold can also be used for classification. It should be understood that the thermal accumulation level classification method defined here is only an example adapted to the local hot spot identification scenario of typical low-voltage distribution area and does not constitute a limitation on the core thermal bias migration parsing logic of this application.

[0053] In this way, the spatial offset state of traditionally unobservable internal additional heat sources can be transformed into a continuously analyzable thermal offset migration sequence and thermal accumulation level, providing spatialized constraint input for the deduction of local additional thermal states in subsequent dynamic thermal circuit models.

[0054] Furthermore, the thermal bias migration is input as an additional loss constraint term into the dynamic thermal circuit model of the distribution transformer through the thermal state deduction unit to generate local additional thermal state quantities corresponding to the transformer neutral point region, metal structural component region, and insulating medium region. It should be noted that the "local additional thermal state quantity" in this embodiment is not the overall oil temperature value or the average temperature rise value of the winding in the ordinary sense, but is used to characterize the additional heat accumulation state formed in the local area of ​​the transformer under the asymmetric thermal excitation. In essence, it reflects the premature thermal instability trend formed in the local area due to additional leakage flux loss, zero-sequence return current loss, and structural component eddy current loss before the overall load has reached the traditional overload threshold.

[0055] Traditional distribution transformer thermal models typically estimate overall temperature rise based solely on load factor, top oil temperature, and average winding current. These models assume an approximately uniform distribution of internal heat sources, thus only providing an overall average thermal state. However, under long-term three-phase unbalanced operation conditions in low-voltage distribution areas, additional heat sources often form spatial offsets along the winding circumference and continuously accumulate in localized areas. Therefore, the thermal state deduction unit does not directly employ the traditional lumped parameter thermal model. Instead, it introduces spatial additional thermal constraints corresponding to "thermal offset migration" into the dynamic thermal circuit model. Specifically, the thermal state deduction unit first uses the geometric center of the transformer winding cross-section as a reference point and projects the thermal offset migration corresponding to each sampling period along the corresponding offset azimuth angle onto the outer circumference of the winding. The winding cross-section is preferably represented by a two-dimensional circular structure model, with the winding geometric center serving as a unified thermal reference center. The thermal offset migration corresponds to the degree of offset of the additional heat source relative to the winding center, while the offset azimuth angle characterizes the spatial direction of the additional heat source along the winding circumference.

[0056] In other words, this embodiment essentially maps the "heat source spatial offset state" obtained by the aforementioned thermal bias migration unit to a "local thermal effect region" in the circumferential direction of the winding. Furthermore, the thermal state deduction unit obtains the intersection point of the thermal bias migration amount and the outer circumference of the winding, and determines the angular partition where this intersection point is located; please refer to [link to previous document]. Figure 5 The diagram showing the circumferential projection and angular partitioning of the thermal bias migration amount illustrates that after the aforementioned thermal bias migration unit obtains the thermal bias migration amount and the corresponding bias azimuth angle, the thermal state deduction unit first establishes a two-dimensional circumferential model of the transformer winding cross-section. This two-dimensional circumferential model uses the geometric center of the winding cross-section as the reference point and the outer circumference of the winding as the boundary of the thermal state space mapping. In this embodiment, the outer circumference of the winding is divided into multiple angular partitions according to the circumferential angle. Figure 5 The diagram shows angle partitions 1 to 6. It should be understood that the number of angle partitions in the diagram is for illustrative purposes only. In practical applications, the number of circumferential partitions can be set to different numbers depending on the transformer capacity, winding structure, and calculation accuracy requirements. For any sampling period corresponding to the thermal bias migration amount L, the thermal state deduction unit establishes a thermal bias migration vector from the winding center along the bias direction based on the corresponding bias azimuth angle θ, and extends this migration vector to the outer circumference of the winding. The thermal bias migration vector intersects with the outer circumference of the winding, and the thermal state deduction unit determines the target angle partition corresponding to the current thermal bias migration amount based on the angle region where this intersection point is located.

[0057] For example Figure 5The thermal bias migration amount L shown is projected along the bias azimuth angle θ, and its intersection point falls into angle partition 2. Then, the additional thermal excitation corresponding to the current sampling period is mapped to angle partition 2. It should be noted that the intersection point here does not represent the actual physical hot spot location inside the transformer, but is used to characterize the main direction of action of the current additional heat source in the circumferential direction of the winding. In the above way, the aforementioned unbalanced excitation state and thermal bias migration state can be converted into a circumferential thermal distribution state with a clear spatial location meaning, providing spatial constraints for subsequent local thermal state calculations.

[0058] As a preferred industrial embodiment, when the thermal bias migration extends to the outer circumference of the winding along the corresponding bias direction, the intersection point with the circumferential boundary is regarded as the main action area of ​​the current additional heat source. Subsequently, according to the angular position of the intersection point, the additional thermal excitation corresponding to the current sampling period is mapped to the corresponding angular partition.

[0059] It should be noted that the thermal offset migration is not simply treated as a single-point heat source. Instead, the spatial distribution relationship of the heat source is established by circumferential angular partitioning. The reason is that the additional heat sources inside the transformer are usually not absolutely concentrated at a certain point, but rather form a certain range of heat diffusion along the local winding area, clamping area and adjacent core structure. Therefore, establishing the thermal space mapping by angular partitioning can improve the ability of the subsequent thermal circuit model to express the local heat diffusion path.

[0060] After determining the corresponding angular partition of the thermal offset, the thermal state simulation unit allocates the thermal offset migration amount to the corresponding angular partition according to the intersection position, and generates the single-cycle additional thermal load value of the angular partition by combining the thermal accumulation level of the corresponding sampling period. The thermal offset migration amount is used to characterize the degree of additional heat source offset, and the thermal accumulation level is used to characterize the intensity of the continuous residence of the additional heat source in the region. That is, the single-cycle additional thermal load value is not only determined by the offset distance, but is also affected by the heat source offset amplitude and the heat source continuous residence stability.

[0061] For example, when the thermal offset migration is large but the residence time is short, it indicates that although the additional heat source has shifted significantly, it has not yet formed a continuous heat accumulation in a fixed area; while when the thermal offset migration continues to act stably on the same angular partition, it means that a continuous additional heat superposition has begun to form in the local area.

[0062] Therefore, by dynamically modulating the additional thermal load through the thermal accumulation level, the risk of transient thermal disturbances being misidentified as persistent hot spots can be effectively avoided.

[0063] In a preferred embodiment, the single-cycle additional thermal load value can be generated by weighting the thermal bias migration amount, the thermal accumulation level, the unbalanced excitation parameters, and the duration of the corresponding time window. However, in some alternative embodiments with lower requirements for edge computing complexity, a simplified mapping can be performed based solely on the thermal bias migration amount and the thermal accumulation level. It should be understood that the additional thermal load generation method here is only an example adapted to typical low-voltage power distribution scenarios and does not constitute a limitation on the core local thermal mapping mechanism of this application.

[0064] Subsequently, the thermal state simulation unit accumulates the additional thermal load values ​​of multiple single-cycle samples falling into the same angular partition within a continuous sampling period to form the partition thermal load value corresponding to that angular partition.

[0065] The accumulation process here is essentially used to simulate the continuous accumulation of additional heat in a local area over time. Since the formation of local hotspots is not caused by a single instantaneous thermal disturbance, but by the inability of additional heat to diffuse sufficiently over a long period of time, the accumulation of continuous periodic thermal loads can more closely approximate the actual local thermal accumulation state.

[0066] Further, please see Figure 6 The diagram shows the structure of the dynamic thermal circuit model for the distribution transformer. The dynamic thermal circuit model includes an oil temperature calculation branch, a winding temperature rise branch, and a circumferential additional heat branch. The oil temperature calculation branch is used to calculate the top-layer oil temperature of the transformer based on the real-time load status of the distribution area and the ambient temperature. It should be noted that the oil temperature calculation branch is mainly used to establish the overall thermal environment boundary of the transformer. Since the formation of local hot spots inside the transformer is still affected by the overall heat dissipation conditions, it is necessary to first determine the top oil temperature state; the winding temperature rise branch is used to determine the top oil temperature based on the top oil temperature. and load current calculation winding reference temperature rise Among them, the winding reference temperature rise represents the theoretical temperature rise state that the winding should reach under the condition of uniform heat source distribution, which reflects the overall thermal state of the transformer under normal load conditions.

[0067] Based on this, the present invention further establishes a circumferential additional heat branch, such as Figure 6 As shown, the thermal load of each angular partition corresponds to the partition. Each heat source is input to its corresponding heat node as an independent heat source. The heat nodes are connected by thermal resistance. A thermal coupling relationship is established to simulate the conduction and diffusion of additional heat in the circumferential direction of the winding. Each thermal node corresponds to a local region in the circumferential direction of the winding. An additional temperature rise occurs at each thermal node after receiving the corresponding zone of thermal load. .

[0068] Since localized heat is not confined to a single area during actual operation, heat conduction correlations are established between adjacent hot nodes to simulate the diffusion of local hot spots to surrounding areas. Finally, the additional temperature rise output by each hot node and the winding reference temperature rise are input together into the thermal state superposition module for calculation, thereby obtaining the local additional thermal state quantity corresponding to each region. It should be noted that, Figure 6 The hot nodes in the model can not only correspond to the winding region, but can also be further mapped to the neutral point region, the metal structural component region, and the insulating medium region, thereby enabling the deduction of the local thermal state of different heat-sensitive parts.

[0069] In one specific implementation, the oil temperature calculation branch can perform recursive calculations based on real-time load rate, ambient temperature, and historical thermal inertia state to obtain the top oil temperature corresponding to the current sampling period.

[0070] As a preferred industrial implementation, the oil temperature recursion method in the IEC thermal model can be used; however, in some scenarios with lower requirements for edge computing resources, a simplified thermal inertia model can also be used for approximate calculation. There is no limitation on this. The winding temperature rise branch calculates the winding reference temperature rise based on the top oil temperature and load current. This winding reference temperature rise is used to characterize the theoretical temperature rise state that the winding as a whole should reach under normal uniform heat source conditions; that is, it serves as the background temperature rise for subsequent superposition of local additional thermal states.

[0071] Based on this, this embodiment further includes a circumferential additional heat branch. This circumferential additional heat branch is used to input the zone heat load value corresponding to each angle zone as an independent heat source, and establishes a heat conduction correlation according to the circumferential position of each angle zone on the winding cross-section. Please refer to [link to relevant documentation]. Figure 7 The diagram shows the internal model structure of the thermal state simulation unit. The winding hot nodes, insulating oil hot nodes, and structural component hot nodes within the dynamic thermal circuit model are connected through a thermal resistance network. Each hot node is configured with corresponding heat capacity parameters. During operation, the ambient temperature is used as a boundary condition input to the thermal network; the real-time load rate corresponds to the winding current-carrying heat source input; the unbalanced excitation parameters are input to the additional heat source branch; and the thermal bias migration is used to correct the position of the additional heat source in the thermal network. Traditional thermal circuit models typically only establish the overall thermal conduction relationship between oil and winding, winding and environment, and winding and core. However, this implementation further establishes multiple local additional thermal nodes along the circumference of the winding, enabling thermal coupling conduction relationships to be formed between regions at different angles.

[0072] That is, the additional heat generated in a certain angle partition not only affects the current local area, but also spreads to the adjacent area along the circumference of the winding. In a preferred embodiment, a thermal resistance coupling relationship is established between adjacent angle partitions, and each angle partition corresponds to an independent thermal capacity node. However, in some simplified structural embodiments, an equivalent thermal diffusivity coefficient can also be used to establish a circumferential thermal correlation, and this is not limited.

[0073] Finally, the circumferential additional heat branch outputs the additional temperature rise value corresponding to each angular zone. The thermal state extrapolation unit superimposes each additional temperature rise value with the winding reference temperature rise to generate the local additional thermal state quantity corresponding to each angular zone. Among them, the winding reference temperature rise reflects the overall uniform thermal state; the additional temperature rise reflects the local asymmetric heat accumulation state; the superposition of the two forms the final local additional thermal state quantity. Furthermore, the local additional thermal state quantity corresponding to different angular zones can be further mapped to the neutral point region, the metal structural component region, and the insulating medium region. Among them, the neutral point region mainly corresponds to the zero-sequence return current accumulation heat effect; the metal structural component region mainly corresponds to the leakage magnetic flux induced eddy current additional loss; and the insulating medium region mainly corresponds to the insulation aging risk caused by long-term local hot spots. In this way, the traditionally unobservable internal non-uniform additional thermal state can be transformed into a local thermal state distribution result with spatial location significance, providing an early thermal risk basis for subsequent overload warning and load control.

[0074] Furthermore, after obtaining the local additional heat state quantities corresponding to each angle zone, the early warning output module further outputs overload early warning information or load control instructions for the corresponding transformer area based on the distribution results of the local additional heat state. It should be noted that the early warning output logic in this embodiment is based on the spatial consistency between the "local additional heat state" and the "thermal bias continuous residence state" to identify the risk of local thermal instability in advance.

[0075] In other words, this application does not simply focus on "whether the overall load is overloaded", but rather focuses on identifying whether a local area has formed a continuous accumulation of additional heat when the overall load has not yet reached the traditional overload threshold. Therefore, the warning output module does not simply compare the local additional heat state quantity with a threshold, but further determines whether the current local heat accumulation area is consistent with the heat residence direction formed by the aforementioned heat offset migration. Essentially, it is verifying whether the local thermal state originates from continuous asymmetric thermal excitation, rather than short-term random thermal disturbance.

[0076] In practice, the early warning output module first reads the local additional thermal state quantity corresponding to each angle partition and identifies the angle partition with the largest local additional thermal state quantity. The "angle partition with the largest local additional thermal state quantity" here actually corresponds to the local area where the additional heat accumulation inside the transformer is most significant. Since the aforementioned thermal state inference unit has mapped the thermal bias migration state to the circumferential thermal distribution of the winding, this angle partition can be further mapped to the neutral point adjacent area, local winding area, clamp area, or insulation medium concentrated area.

[0077] Subsequently, the early warning output module matches the angle region with the largest local additional thermal state quantity with the offset azimuth angle corresponding to the thermal offset residence section. It should be noted that the core purpose of this "position matching" is to confirm whether the current local high-heat area is formed by a continuous and stable thermal offset residence. This is because, in actual operation, transient load fluctuations, short-cycle surge currents, or environmental heat dissipation disturbances may occur inside the transformer, causing a short-term increase in local thermal state. However, if this local thermal state is not consistent with the thermal offset migration direction, it usually indicates a transient thermal disturbance rather than a continuous local heat accumulation. Therefore, in this embodiment, only the angle region with the largest local additional thermal state quantity is matched. Only when the offset azimuth angles corresponding to the partition and the thermal offset residence section are located in the same angular partition are they considered to have spatial consistency between the current local high-temperature state and the continuous asymmetric thermal excitation. Furthermore, this spatial consistency determination essentially establishes a closed-loop verification relationship between the "heat source formation direction" and the "local heat accumulation location." That is, the aforementioned thermal offset migration unit is used to identify the formation and residence direction of the additional heat source; the thermal state inference unit is used to identify the actual local heat accumulation location; and the early warning output module further verifies whether the heat source direction and the heat accumulation location continuously coincide. In this way, the false triggering problem caused by short-term temperature rise fluctuations in traditional hot spot early warning can be effectively avoided.

[0078] After completing the location matching, the early warning output module further determines the duration of the thermal bias dwelling section.

[0079] In one specific implementation, when the duration of the thermal bias dwelling section is not less than 8 minutes, the corresponding angle zone is determined as a local overheating risk area. The reason for setting the duration criterion here is that local thermal instability is usually not caused by instantaneous additional thermal excitation, but by the additional heat source staying in a fixed area for a long time, which prevents the local heat from being fully diffused.

[0080] If the thermal bias direction lasts for a short period of time, it indicates that the additional heat source is still in a dynamic state of movement. At this time, the local heat can usually be gradually released through oil flow circulation and winding heat diffusion. However, when the thermal bias remains stably in the same area for a preset time, it means that the area has entered a state of continuous heat accumulation.

[0081] Therefore, in this embodiment, alarms are not simply based on instantaneous additional thermal state quantities, but a thermal bias dwell duration constraint is further introduced. It should be noted that the 8-minute dwell time duration here is mainly used to adapt to the typical thermal inertia characteristics of low-voltage distribution transformer areas. As a preferred industrial embodiment, the dwell time threshold can be adjusted according to the transformer capacity level, oil circulation method, and winding structure. However, in some embodiments with lower deployment complexity requirements, a fixed dwell time threshold can also be used. This is not limited.

[0082] When the following conditions are met simultaneously: the maximum angle of the local additional thermal state quantity is partitioned, the thermal bias residence direction is consistent, and the thermal bias residence time reaches a preset duration, the early warning output module determines the angle partition as a local overheating risk area. Furthermore, the early warning output module outputs the corresponding transformer overload early warning information based on the circumferential position of the local overheating risk area on the transformer winding cross section. The early warning information includes not only whether there is an overheating risk, but also the corresponding circumferential position of the risk area, the duration of the risk, the heat accumulation level, the local additional temperature rise trend, and the corresponding thermal bias migration direction.

[0083] In other words, the output of this implementation is not a traditional single "overload alarm", but local thermal risk information with spatial location significance.

[0084] Please see Figure 8 The diagram shown illustrates the logic for determining the overload warning of a transformer substation. Figure 8 The inputs include local additional thermal state quantities, winding hot spot temperature rise, and real-time load rate, derived from the output results of the thermal state deduction unit. The local additional thermal state quantities are represented by... Figure 7 The output local temperature rise prediction value; the winding hot spot temperature rise represents the winding hot spot temperature rise output by the thermal state simulation unit; the real-time load rate represents the real-time load level of the transformer area calculated by the operation access module.

[0085] For example, when the local overheating risk area is located near the neutral point for a long time, it can be determined that the risk of zero-sequence backflow accumulation is continuously increasing; when the local overheating risk area is located near the clamping component, it can be determined that the additional loss of leakage magnetic eddy current is continuously accumulating; when the local overheating risk area is located near the insulating medium, it indicates that the risk of local insulation thermal aging is increasing.

[0086] In a preferred embodiment, the early warning output module may further output load control commands, which may include: single-phase load transfer commands; flexible peak shaving commands; distributed charging equipment current limiting commands; or transformer area load redistribution commands.

[0087] It should be noted that the load control in this embodiment is not a traditional uniform load limit based on the overall load rate, but a directional adjustment based on the direction of local thermal risk formation; for example, when the thermal bias is shifted toward a certain adjacent area for a long time, the load access ratio of the corresponding phase can be reduced first, thereby weakening the accumulation of zero-sequence return current in the corresponding direction.

[0088] As a preferred industrial embodiment, the early warning output module can send corresponding control commands to the distribution automation master station, the transformer edge controller, or the low-voltage flexible control equipment through the intelligent fusion terminal; however, in some embodiments that only have local early warning capabilities, it can also output only early warning status information, and there is no limitation on this. In this way, this embodiment can identify the trend of local heat accumulation in advance before the overall load reaches the traditional overload threshold, and output corresponding risk warnings based on the spatial residence characteristics of the heat source, thereby reducing the risk of winding insulation aging, structural component overheating, and thermal instability in the neutral point area caused by the continuous development of local hot spots.

[0089] Please see Figure 9 The graph shows a comparison of the predicted local temperature rise results. The horizontal axis represents the operating time t, and the vertical axis represents the local temperature rise value T. The first curve is the prediction result curve of this invention, the second curve is the prediction curve of the existing method, and the third curve is the actual measured temperature rise curve. The prediction curve of the existing method is the local temperature rise calculation result obtained only based on the dynamic thermal circuit model of the distribution transformer. Its input mainly includes the real-time load rate of the transformer area and the ambient temperature, without considering the additional loss caused by the three-phase unbalanced operation.

[0090] The predicted curve of this invention is a local temperature rise prediction result obtained by using unbalanced excitation parameters and thermal bias migration as additional inputs in the thermal state deduction based on a dynamic thermal circuit model. The actual measured temperature rise curve is the temperature rise change result of the corresponding local monitoring point under the simulation condition, used to characterize the actual local thermal state of the transformer. Figure 9 It is known that, under the condition that the total load rate of the transformer area does not reach the upper limit of the rated capacity and the three-phase imbalance continues to exist, the present invention can continuously track and predict the local additional temperature rise; compared with the existing methods based only on the dynamic thermal circuit model, it has a higher fitting ability for the local thermal state of the transformer neutral point area and structural component area, and can be used as the basis for the overload warning judgment of the transformer area.

[0091] To verify the ability of the intelligent fusion terminal-based distribution transformer overload early warning system described in this application to identify local thermal instability in advance, the applicant conducted a comparative test in a residential low-voltage distribution transformer area. The test object was a distribution transformer with a rated capacity of 400kVA. The low-voltage side was connected to residential lighting loads, split air conditioner loads, electric water heater loads, and some nighttime charging equipment for a long time. During the test, the ambient temperature fluctuated between 26.3℃ and 31.8℃. In order to avoid excessive influence of extreme operating conditions on the test results, no fault conditions were artificially constructed during the test, and only the actual operating load conditions were used for continuous observation.

[0092] The comparative example uses a traditional overload monitoring method based on load rate and top oil temperature threshold; the embodiment uses the early warning method described in this application based on unbalanced excitation parameters, thermal offset migration, and local additional thermal state. During the test, 72 hours of continuous operating data were collected, and typical operating segments with significant single-phase off-center load fluctuations were selected as the comparative analysis objects. Table 1 below shows the detection results of the traditional scheme and the scheme of this application under different operating segments: Table 1 Comparison of thermal risk identification results under different operating periods As shown in Table 1, during the 20:10-20:42 operating period, although the maximum phase load rate of the transformer area was only 83%, which did not reach the traditional overload alarm threshold, the neutral line current continued to rise due to the concentrated connection of single-phase loads from residents. The thermal bias migration direction remained in the same angular zone for more than 11 minutes. The system of this application has identified the local high heat accumulation state and output a local overheat risk warning in advance. In contrast, the traditional solution only judges based on the overall load rate and top oil temperature during this period and does not trigger an abnormal alarm.

[0093] Furthermore, during the shutdown inspection of the transformer after the aforementioned operating period ended, it was found that there were local temperature rise marks in the area near the neutral point of the corresponding winding that were significantly higher than those in the surrounding area. The temperature was re-measured on site using infrared thermometers, and the temperature in the local area was about 6.8°C higher than that in the adjacent area, which was basically consistent with the thermal bias residence direction identified in this application.

[0094] It should be noted that, due to the random fluctuations in residential load during on-site operation, the direction of thermal offset migration is not completely fixed at different times. Therefore, the local additional thermal state quantity fluctuates to some extent. Selecting some continuous sampling periods during the test period, the correspondence between the thermal offset migration quantity and the local additional thermal state quantity in this application was statistically analyzed. The results are shown in Table 2: Table 2 Correspondence between thermal offset migration state and local additional heat state As shown in Table 2, when the thermal offset migration direction remains stable within the continuous sampling period, the corresponding local additional thermal state quantity is concentrated in the same angular partition. However, when the offset direction changes significantly, the local thermal state distribution also shifts accordingly. The results indicate that there is a continuous correspondence between the "thermal offset migration - circumferential thermal load - local additional thermal state" established in this application. In addition, it was observed during the test that although the overall load rate was high in some operating periods, no continuous thermal residence was formed due to the frequent changes in the thermal offset migration direction. Therefore, this application did not output a local overheating warning. However, in some periods when the overall load rate was not high, significant local heat accumulation was still formed due to the long-term concentration of zero-sequence backflow in a fixed direction. This indicates that this application is not concerned with the overall load level in the traditional sense, but rather with whether the local additional heat source continuously forms heat accumulation in a fixed area.

[0095] Based on the above test results, it can be seen that the technical solution described in this application can identify the local heat accumulation state caused by long-term asymmetrical load before the traditional overall overload criterion is triggered. By establishing a spatial correlation between thermal bias residence behavior and local additional heat state, it can provide early warning of local thermal instability risk. Compared with the traditional monitoring method based on overall load rate or average oil temperature, this application is more suitable for identifying local thermal risks in long-term single-phase bias load and zero-sequence return scenarios in residential low-voltage distribution substations.

[0096] Finally, it should be noted that the parameters involved in this application, such as "unbalanced excitation parameters," "thermal bias migration," "thermal bias migration sequence," "heat accumulation level," and "local additional thermal state quantity," are all state characterization quantities established for the non-uniform additional thermal behavior inside the distribution transformer. They are mainly used to describe the evolution trend of the spatial distribution of internal heat sources under asymmetrical load conditions on the low-voltage side, and are not equivalent to actual physical temperature measurements, traditional load rate parameters, or single electrical protection settings. Among them, "thermal bias migration" is used to characterize the degree of spatial offset of the center of the additional heat source relative to the geometric center of the winding; "heat accumulation level" is used to characterize the intensity of the continuous residence of the additional heat source in a local area; and "local additional thermal state quantity" is used to characterize the local heat accumulation state formed in the corresponding area under the action of additional heat. The above state quantities have a continuous progressive relationship, and their underlying data flow logic is as follows: The access module is used to obtain real-time operating electrical parameters. Sequence component decomposition unit extracts asymmetric electrical features; Unbalanced excitation units construct asymmetric thermal excitation intensity; The thermal offset migration unit establishes the spatial offset state of the heat source; The thermal state simulation unit produces results of localized thermal accumulation. The early warning output module completes the assessment of local thermal instability risk.

[0097] Therefore, the overall technical solution of this application is neither a traditional single-point temperature monitoring system nor a common load prediction system. Instead, it establishes a continuous analytical system that maps the electrical imbalance state to the internal heat source space migration state layer by layer, and finally forms the result of early identification of local thermal risks.

[0098] Furthermore, the "thermal bias migration" in this application is a spatial equivalent modeling method for internal heat sources. Its purpose is to solve the problem that in existing low-voltage distribution transformer areas, due to long-term single-phase off-center loading, zero-sequence return current and negative-sequence magnetomotive force disturbance, local hot spots are formed in advance, but traditional average thermal models are difficult to identify in advance.

[0099] Therefore, this application focuses on the "spatial persistence" and "heat source retention" during the formation of local thermal instability, rather than the instantaneous temperature rise at a single moment.

[0100] In other words, even if the overall load has not reached the traditional overload state, as long as the local additional heat source continues to reside in a fixed area and forms heat accumulation, this application can still output a local heat risk warning in advance.

[0101] It should be further clarified that the descriptions of winding cross-section, angular partitioning, additional magnetic flux axis, circumferential additional thermal branch, thermal resistance coupling relationship, and local thermal nodes in this application are all equivalent thermal model structures established to realize the spatialization of local thermal state deduction, and do not limit the actual physical structure to be completely consistent with the model; for example, in some embodiments, a two-dimensional winding cross-section model can be used to establish the circumferential thermal mapping relationship; while in other embodiments, a three-dimensional finite node thermal network, a non-uniform thermal capacity model, or a multi-layer thermal coupling model can also be used to realize the corresponding local thermal state deduction.

[0102] Similarly, the parameters in this application, such as the number of angle partitions, the threshold for heat accumulation level, the threshold for dwell time, the length of the sliding time window, and the sampling period, can all be adjusted according to the transformer capacity level, cooling method, load type, transformer substation topology, and edge terminal computing power conditions. It should be understood that the above parameter settings are only preferred embodiments adapted to typical low-voltage distribution substation operating conditions and do not constitute a limitation on the core technical ideas of this application.

[0103] Furthermore, the contents of negative sequence component, zero sequence component, unbalanced excitation basic quantity and negative sequence disturbance coefficient in this application are mainly used to characterize the influence of low-pressure side asymmetric load state on the formation of internal additional heat source, and are not limited to adopting a specific mathematical expression form; for example, in some optional embodiments, harmonic distortion factor, zero sequence harmonic content, load phase deviation rate or neutral point potential drift can also be used as additional correction inputs to enhance the characterization ability of internal thermal disturbance state under complex load scenarios.

[0104] Furthermore, the dynamic thermal path model in this application is preferably deployed on the local edge side of the intelligent fusion terminal to realize real-time online simulation of local thermal risks; however, in some computing power centralized implementations, some thermal state analysis tasks can also be deployed on the power distribution master station, edge gateway or cloud-side collaborative computing platform, and front-end data acquisition and state synchronization can be completed through the intelligent fusion terminal; therefore, this application does not limit each functional unit to exist in the form of a completely independent software module or physical hardware module.

[0105] In some implementations, the sequential component decomposition unit, unbalanced excitation unit, thermal bias migration unit, and thermal state inference unit can be implemented using a unified edge inference engine, or they can be implemented collaboratively using hardware logic units and software algorithm modules. Similarly, the early warning output module can output local alarm information, or it can be linked with the transformer area flexible control system, distributed load management system, or distribution automation system to achieve corresponding load regulation and control. Finally, it should be noted that the core protection focus of this application is not limited to a specific thermal model, a specific integration method, or a specific threshold strategy, but rather to identify the spatial offset of internal additional heat sources through the asymmetric electrical state of the low-voltage side, and further establish the correlation between the local heat accumulation state and the thermal bias residence behavior, thereby realizing the early inference and early warning of the risk of local thermal instability in the distribution transformer area.

[0106] Any technical solution that uses the above core idea to perform continuous time-series analysis of the internal heat source offset state and is used for local thermal risk identification should fall within the protection scope of this application.

[0107] The above provides a detailed description of an overload early warning system for distribution transformers based on an intelligent fusion terminal, as provided in this application. Those skilled in the art will recognize that the algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented using electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention. The above description of the embodiments is merely for the purpose of helping to understand the method and core ideas of this application. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A distribution transformer overload early warning system based on an intelligent fusion terminal; comprising an intelligent fusion terminal installed on the low-voltage side of the transformer in the distribution transformer area, characterized in that: The intelligent fusion terminal includes an operation access module, an electromagnetic thermal analysis module, and an early warning output module; The operation access module is used to collect data on three-phase voltage, three-phase current, neutral current, and ambient temperature of the distribution substation. The electromagnetic thermal analysis module includes a sequence component decomposition unit, an unbalanced excitation unit, a thermal bias migration unit, and a thermal state deduction unit. The sequence component decomposition unit is used to perform symmetrical component decomposition on the three-phase voltage and three-phase current to obtain positive sequence component, negative sequence component and zero sequence component. The unbalanced excitation unit is used to accumulate the zero-sequence component within a preset time window, and combine it with the negative-sequence component to construct unbalanced excitation parameters characterizing the asymmetric load state of the transformer area; the thermal bias migration unit is used to calculate the thermal bias migration amount of the heat source inside the transformer relative to the winding center based on the unbalanced excitation parameters and the neutral line current, and construct a thermal bias migration sequence based on the continuous time-series thermal bias migration amount. The thermal state deduction unit is used to input the thermal bias migration as an additional loss constraint term into the dynamic thermal circuit model of the distribution transformer to generate local additional thermal state quantities corresponding to the transformer neutral point region, metal structural component region and insulating medium region. The early warning output module is used to output overload early warning information or load control commands for the corresponding power distribution area based on the local additional heat status.

2. The overload early warning system for distribution transformers based on an intelligent fusion terminal according to claim 1, characterized in that: The sequence component decomposition unit uses the three-phase voltage and three-phase current sampling sequence corresponding to the continuous variation range of the neutral line current as the input quantity for symmetrical component decomposition, and outputs the positive sequence component, negative sequence component and zero sequence component corresponding to the continuous variation range of the neutral line current.

3. The overload early warning system for distribution transformers based on an intelligent fusion terminal according to claim 1, characterized in that: The cumulative calculation of the zero-sequence component by the unbalanced excitation unit includes: A sliding time window is established according to the continuous sampling period, and the effective value of the zero-sequence current corresponding to each sampling moment within the sliding time window is obtained, and a zero-sequence current sequence is formed according to the sampling order. Calculate the change in the effective value of the zero-sequence current between adjacent sampling times in the zero-sequence current sequence, and extract the local extreme points in the change in the effective value of the zero-sequence current. Multiple cumulative segments are divided with the sampling time corresponding to each local extreme point as the boundary; Discrete integration is performed on the effective value of the zero-sequence current in each accumulation segment, and the values ​​are accumulated sequentially according to the time order of each accumulation segment to obtain the unbalanced excitation base quantity corresponding to the sliding time window.

4. The overload early warning system for distribution transformers based on an intelligent fusion terminal according to claim 3, characterized in that: The unbalanced excitation unit couples the unbalanced excitation base quantity with the effective value of the negative sequence current to generate unbalanced excitation parameters. The coupling calculation process includes: The zero-sequence current carrying capacity coefficient is determined based on the ratio of the unbalanced excitation base quantity to the preset time window length. The rate of change of the effective value of the negative sequence current within a continuous sampling period is obtained, and the negative sequence disturbance coefficient is calculated. The expression is as follows: in, This represents the effective value of the negative sequence current at the current sampling time. This is the effective value of the negative sequence current at the previous sampling time; The product of the zero-sequence current carrying capacity coefficient and the negative-sequence perturbation coefficient is used as the unbalanced excitation parameter.

5. A transformer overload early warning system based on an intelligent fusion terminal according to claim 1, characterized in that: The thermal bias migration unit calculates the thermal bias migration amount of the internal heat source of the transformer relative to the winding center based on the unbalanced excitation parameters and the neutral line current, and constructs a thermal bias migration sequence based on the continuous time-series thermal bias migration amount, including: The phase shift relationship of the three-phase current on the low-voltage side of the transformer within the corresponding sampling period is obtained based on the unbalanced excitation parameters, and the additional magnetic flux action axis corresponding to the transformer winding section is constructed by combining the amplitude and direction of the neutral line current. Using the geometric center of the transformer winding cross-section as a reference point, the bias direction of the additional magnetic flux is determined based on the spatial relationship between the axis of action of the additional magnetic flux and the reference point. Calculate the radial offset distance between the axis of action of the additional magnetic flux and the reference point along the direction of the additional magnetic flux bias, and record the radial offset distance as the thermal offset migration amount of the current sampling period; Several thermal bias migration values ​​corresponding to consecutive sampling periods are obtained and arranged in chronological order to form the thermal bias migration sequence.

6. The overload early warning system for distribution transformers based on an intelligent fusion terminal according to claim 5, characterized in that: The hot bias migration unit is also used for time-series parsing of the hot bias migration sequence, including: Obtain the offset azimuth angle corresponding to each thermal offset migration within a continuous sampling period; and calculate the angle difference between the corresponding offset azimuth angles of adjacent sampling periods; When the angle difference between adjacent offset azimuth angles within 5 consecutive sampling periods is no greater than 15°, the corresponding sampling period is determined as the same hot offset dwelling segment. Record the start sampling time, end sampling time, and duration corresponding to the thermal bias dwell section.

7. A transformer overload early warning system based on an intelligent fusion terminal according to claim 6, characterized in that: The thermal bias migration unit is also used to calculate the thermal bias migration intensity index based on the thermal bias residence section; The thermal bias migration intensity index is determined based on the rate of change of thermal bias migration within a continuous sampling period and the residence time of the thermal bias residence segment within the corresponding statistical period, and is expressed as: in, This is the thermal bias migration intensity index; For the first Thermal bias migration amount corresponding to each sampling period; The time interval between adjacent sampling periods; The total number of samples taken within the statistical period; The cumulative residence time of the thermal bias residence segment within the statistical period; The total duration of the statistical period; The thermal bias migration unit classifies the thermal accumulation level corresponding to the thermal bias residence section according to the thermal bias migration intensity index and outputs it to the thermal state inference unit.

8. A transformer overload early warning system based on an intelligent fusion terminal according to claim 1 or 7, characterized in that: The working process of the thermal state simulation unit includes: Using the geometric center of the transformer winding section as the reference point, the thermal bias migration corresponding to each sampling period is projected onto the outer circumference of the winding along the corresponding bias azimuth angle. Obtain the position of the intersection point between the thermal bias migration amount and the outer circumference of the winding, and determine the angular partition where the intersection point is located; The thermal offset migration is allocated to the corresponding angular partition according to the intersection position, and the single-cycle additional thermal load value of the angular partition is generated by combining the thermal accumulation level of the corresponding sampling period. The additional thermal load values ​​of multiple single-cycle samples falling into the same angular partition within a continuous sampling period are accumulated to form the partition thermal load value corresponding to the angular partition.

9. A transformer overload early warning system based on an intelligent fusion terminal according to claim 1, characterized in that: The dynamic thermal circuit model of the distribution transformer includes an oil temperature calculation branch, a winding temperature rise branch, and a circumferential additional heat branch; the oil temperature calculation branch is used to calculate the top oil temperature of the transformer based on the real-time load status of the distribution area and the ambient temperature. The winding temperature rise branch is used to calculate the winding reference temperature rise based on the top oil temperature and load current; the circumferential additional heat branch is used to input the zone heat load value corresponding to each angle zone as an independent heat source, and establish heat conduction correlation according to the circumferential position of each angle zone on the winding cross section. The circumferential additional heat branch outputs the additional temperature rise value corresponding to each angle partition. The thermal state deduction unit superimposes each additional temperature rise value with the winding reference temperature rise to generate the local additional thermal state quantity corresponding to each angle partition.

10. A transformer overload early warning system based on an intelligent fusion terminal according to claim 1, characterized in that: The operation process of the early warning output module includes: Read the local additional thermal state quantities corresponding to each angular partition, and match the angular partition with the largest local additional thermal state quantity with the offset azimuth angle corresponding to the thermal offset residence section. When the angle zone with the largest local additional thermal state quantity is located in the same angle zone as the offset azimuth of the thermal offset residence section, and the duration of the thermal offset residence section is not less than 8 minutes, the angle zone is determined as a local overheating risk area. Based on the circumferential position of the local overheating risk area on the transformer winding cross section, the corresponding transformer overload warning information is output.

Images