Risk assessment and update optimization method and system for low-voltage distribution transformer

Abstract

The invention discloses a low-voltage distribution transformer risk assessment and update optimization method and system. Data such as operation, insulation, events and weather are collected and are equivalent to the same time scale, and equipment health scores and fault occurrence probabilities are given according to indexes such as loads, unbalance, harmonic waves, gas concentration and external force strength; and user loss, material and maintenance cost caused by power failure are converted into one consequence, and comparable risk values are obtained. In the updating stage, a net present value and a unit benefit ratio are calculated by surrounding common schemes such as same-capacity replacement, same-capacity high-efficiency-level replacement, capacity-increasing replacement, migration replacement and the like, replacement time and power failure duration are explicitly recorded, and a recommended scheme and an economical efficiency sorting list are output after sorting. According to the method, multi-source data consistency and interpretable evaluation are achieved, large-scale risk quantification is completed with a small number of parameters, economic benefits and user influences are brought into decision making, and investment benefits and power supply reliability are remarkably improved and updated.

Description

Technical Field

[0001] This invention relates to the field of power distribution network operation and maintenance and power asset management technology, and in particular to a method and system for risk assessment and optimization of low-voltage distribution transformers. Background Technology

[0002] Low-voltage distribution transformers are large in size and scattered in distribution. In many places, the management of equipment still relies on the service life of the transformers and emergency repairs after a failure. The actual failure risk is difficult to quantify in a timely manner. Replacement is not just a technical operation. It also involves a series of constraints such as budget, power outage window, construction team, and material availability. Existing solutions lack a closed-loop method that quantifies reliability, economic costs, and the impact of power outages on users in a unified manner.

[0003] There are already methods in the industry for using health indices and reliability assessments, but most of them remain at the level of models or medium and high voltage equipment, with insufficient implementation on the low-voltage side. The typical approach is to construct a health index based on condition characteristics and infer lifespan and failure rate, but this method has limited linkage with external impacts and operation and maintenance scenarios.

[0004] In recent years, grid-based operation and maintenance has been promoted on the distribution network side, which has begun to gather operational data, defect records, impact counts, and meteorological external forces. Combined with seasonal or special inspections, some people have also used equivalent service life to characterize the inflection point of insulation degradation and form a list-based arrangement, such as patent CN109713671A. However, this link usually stops at inspections and general maintenance, and has not yet been extended to scheduling constraints such as strategy optimization, combined outages, budget and non-concurrent operation on the same feeder, and has not yet formed a closed-loop learning.

[0005] Meanwhile, risk assessments oriented towards external forces, such as patent CN110390469A, combine thunderstorm environments with equipment lightning resistance levels to assess the lightning damage risk to distribution transformers, providing early warnings and classifications. These methods have some reference value, but they mostly focus on single external force scenarios. State parameters related to the aging of oil-paper insulation, such as DGA, moisture content, and dielectric loss, also need to be systematically classified. These factors are not assessed within a unified framework, along with the economics of replacement, user importance, and the impact of power outages.

[0006] In summary, existing technologies have two main shortcomings: inconsistent health assessment criteria and weak coupling with external forces; and a lack of quantitative indicators that unify risk reduction, cost, and power outage impact onto a single scale. Therefore, this invention proposes an integrated online assessment and optimization method and system that integrates health, risk, economics, and scheduling. Summary of the Invention

[0007] The purpose of this invention is to propose a method and system for risk assessment and optimization of low-voltage distribution transformers. This method enables online assessment in long-term operation scenarios on the distribution side, balancing real-time performance, non-intrusiveness, and interpretability. It establishes a quantitative correlation between risk and economic consequences through dual-track modeling using time-varying common factors of external forces, EOA (Equivalent Operating Age), and a health index HI. Under constraints such as budget and outage windows, it automatically generates optimal replacement plans, thereby meeting the quality and cost control requirements for ensuring power supply and reducing losses within a 20-30 year lifespan of distribution assets.

[0008] To achieve the above objectives, the present invention is implemented using the following technical solution: Firstly, a method for risk assessment and optimization of low-voltage distribution transformers is provided, including: acquiring multi-source data through a Supervisory Control and Data Acquisition (SCADA) and Advanced Metering Infrastructure (AMI) measurement module, an online dissolved gas analysis (DGA) oil chromatography online monitoring module, an event acquisition module, and a meteorological geographic information system (GIS) access module; The hot spot temperature θHS[n] is estimated based on the top oil temperature-winding hot spot thermal model, and the hourly aging factor FAA[n] is obtained according to the Arrhenius relation; Calculate the Equivalent Operating Life (EOA) and Health Index (HI) to conduct a dual-track health assessment. Failure probability is estimated based on health and external force characteristics to obtain the failure probability Pfail(Δ) for the target time window; The costs of materials, maintenance, and power outages are summarized into a single failure consequence C, which is then multiplied by the failure probability Pfail(Δ) within the window Δ to obtain the equipment risk R, thereby realizing the monetization and comparable ranking of risks. First, use NPVi(s) as the threshold to screen the schemes, select the best replacement scheme with the highest unit benefit ratio (RAER) for each device, then sort the entire network according to RAER from high to low and break up the tie according to Φi(s) rules to generate an economic priority list.

[0009] As an optional technical solution of the present invention, multi-source data is acquired through a Supervisory Control and Data Acquisition (SCADA) and Advanced Metrology Infrastructure (AMI) measurement module, an online dissolved gas analysis (DGA) oil chromatography online monitoring module, an event acquisition module, and a meteorological geographic information system (GIS) access module, including: The three-phase current, voltage, active power R, reactive power R, and harmonic spectrum are obtained through SCADA and AMI measurement modules and converted to the same time scale. The formulas for their interval mean and interval extreme values ​​are as follows:

[0010] In the formula: x[n] refers to the nth time window The average time value within the window, Wn is the length of the time window, Kn is the number of samples falling into Wn, and xmax[n] refers to the maximum value within the same window Wn; The dissolved gas contents, water content, and dielectric loss of H2, CH4, C2H6, C2H4, and C2H2 are obtained through the DGA online module and are represented as follows:

[0011] In the formula: gj is the continuous concentration sequence obtained by piecewise linear interpolation, and j belongs to the content of dissolved gases such as H2, CH4, C2H6, C2H4, and C2H2, as well as water content and medium loss; The event acquisition module records the counts of short circuits, closing events, lightning strikes, and tripping events. The formula for calculating the interval event count is as follows:

[0012] In the formula: te is the event timestamp (such as short circuit / closing / lightning strike), and Nevent[n] is the number of events within Wn; External force sequences such as thunderstorm density, icing thickness, extreme high-temperature days, and wind speed are obtained through the meteorological GIS access module. The formula for aggregating these external force sequences is as follows:

[0013] In the formula: Iq(t) is the value of the qth external force factor over time (e.g., thunderstorm density, ice thickness, extreme high temperature index, wind speed), and Iq[n] is the average external force value of Wn; As an optional technical solution of the present invention, the hot spot temperature θHS[n] is estimated based on the top oil temperature-winding hot spot thermal model, and the hourly aging factor FAA[n] is obtained according to the Arrhenius relation, including: The top oil temperature θTO[n] is calculated using a first-order thermal inertia model, and then the winding hot spot temperature θHS[n] is obtained by superimposing the increase in winding hot spot temperature. The calculation formula is as follows:

[0014] In the formula: θTO[n] is the top oil temperature in the nth time window, θHS[n] is the winding hot spot temperature in the nth time window, θamb[n] is the ambient temperature (from the meteorological module aligned to Wn), ΔθTO[n] is the top oil temperature rise caused by the load in the nth time window, ΔθHS[n] is the winding hot spot rise caused by the load in the nth time window, ΔθTO,R is the top oil temperature rise under rated load, ΔθHS,R is the winding hot spot rise under rated load, nto and nhs are the load temperature rise power exponents, and τTO is the first-order thermal time constant of the top oil temperature; Using the Arrhenius relative aging rate formula, the window-by-window aging acceleration factor FAA[n] is obtained, and the thermal equivalent aging time EAAT is accumulated over time, as shown in the following formula:

[0015] In the formula: FAA[n] is the aging acceleration factor of the nth window, THSn=θHSn+273.15 is the absolute temperature of the nth window. Tref is the reference absolute temperature, Ea / R is the ratio of the equivalent activation energy to the gas constant, EAAT is the thermal equivalent aging time within the selected statistical interval, and Δth is the value of the time step in hours; As an optional technical solution of the present invention, the calculation of the equivalent service life (EOA) and health index (HI) includes: summarizing operational, environmental, and chemical information into three categories of cumulative stresses: overload severity, damp heat effect, and shock count, as the basis for service life conversion; converting the equivalent thermal aging hours (EAAT) and the three categories of stresses into service life increments, and adding them to the actual service life to obtain the equivalent service life (EOA), the calculation formula of which is shown in the following formula: ; Where: EOA is the equivalent service life, Age is the actual service life of the equipment, EAAT is the thermal equivalent aging time, αT is the thermal aging conversion factor, αE is the service life conversion factor for overload severity, αM is the service life conversion factor for damp heat influence, and αS is the service life conversion factor for impact count. State features are extracted and uniformly monotonicized to 0-1 to ensure that larger values ​​indicate worse health. A non-negative weighted linear interpretable scoring method is used to output a health index of 0-100. A constrained learning form for weight calculation is also provided to facilitate automatic weight calibration in labeled scenarios. The EOA from the age perspective and the HI from the state perspective are fused into a joint score S, and a quantile method is used to generate grading thresholds, outputting levels H0 to H4 (H4 being the most severe). The calculation formula is as follows:

[0016] In the formula: τS(k) is the k-th threshold used for grading, τEOAhi is the "high age" threshold for EOA, and τHIlo is the "low health" threshold for HI. Output the grade label for each device. ; As an optional technical solution of the present invention, the calculation of the equivalent operating years (EOA) and health index (HI) for dual-track health assessment firstly normalizes the EOA and HI, and then performs a weighted summation of the external forces over a recent period using a 0-1 normalization method, adding them according to their weights to obtain a comprehensive risk factor; a basic annual failure rate (λ0) table is set according to the health level (Level), and λ0 is enlarged and converted into a window failure probability, the failure probability within the window is shown in the following formula:

[0017] In the formula: Δ is the target window length, and λΔ is the equivalent failure rate within the window.

[0018] As an optional technical solution of the present invention, the failure probability estimation of health and external force characteristics includes: summarizing the material and maintenance costs and the impact of power outage into a single failure consequence C, and then multiplying it by the failure probability Pfail(Δ) within the window Δ to obtain the equipment risk R.

[0019] As an optional technical solution of the present invention, four feasible replacement schemes s are provided; for each scheme s, annualized or windowed benefit items are calculated, including risk avoidance benefits, operation and maintenance savings, and energy efficiency benefits; using net present value as a threshold, feasible schemes are first screened, and then, combining two types of indicators—unit benefit ratio and power outage friendliness—an economic ranking list is obtained, including: The formula for calculating net present value is as follows:

[0020] In the formula, Cone,i(s) is the one-time total cost of option s, and Y is the number of years for economic evaluation; when NPVi(s)>0, it means that the option is economically feasible and proceeds to the next sorting step; when NPVi(s)≤0, it means that the option will not be changed for the time being.

[0021] Then sort the entire network by RAER from high to low and break up the tie using the Φi(s) rule to generate an economic priority list.

[0022] Secondly, a method and system for risk assessment and optimization of low-voltage distribution transformers are provided, characterized by including: a multi-source data acquisition module, used to acquire three-phase current, voltage, active power, reactive power, harmonics and event records through SCADA and event acquisition modules respectively; to acquire dissolved gas and water content and dielectric loss indicators in oil through DGA online monitoring; and to acquire external force information such as thunderstorm density, ice thickness, extreme high temperature index and wind speed through meteorological GIS access. The data preprocessing and alignment module is used to extract features such as load rate, imbalance, total harmonic distortion, TDCG and gas ratio, event density and external force intensity, and to perform 0-1 monotonicization according to preset upper and lower boundaries. The Health Assessment and Risk Monetization module is used to calculate EOA and HI and generate a health level Level based on them. The module then looks up the base annual failure rate based on the Level and calculates the target window failure probability by combining it with the comprehensive factor R0. The module combines material maintenance costs and power outage value loss into a primary consequence C and calculates the risk R. The strategy optimization module is used to filter solutions. It first filters based on NPVi(s) as the threshold, selects the solution with the highest RAER as the optimal replacement solution, and then sorts the entire network from high to low RAER and breaks up the tie according to the Φi(s) rule to generate an economic priority list. Compared with the prior art, the beneficial effects achieved by the present invention are as follows: 1. This invention equates SCADA and AMI measurements, DGA, event and meteorological GIS multi-source data to the same time scale, adopts an interpretable dual-track assessment combining equivalent operating years EOA and health index, and obtains a low-parameter, robust window failure probability using the grade baseline annual failure rate and comprehensive factor R0.

[0023] 2. The method monetizes factors such as power outage losses and material maintenance to form a comparable risk R, and automatically selects the best option among four options based on the risk reduction per unit cost (RAER), taking into account budget, important users, and power outage duration. The method is easy to deploy, easy to audit, versionable for governance, and can be scaled for scheduling and parameter calibration, significantly improving the marginal return on upgrade investment and power supply reliability. Attached Figure Description

[0024] Figure 1 This is a flowchart of the low-voltage distribution transformer risk assessment and update optimization method and system provided in the embodiments of the present invention. Detailed Implementation

[0025] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments and specific features in the embodiments are detailed descriptions of the technical solution of the present application, rather than limitations thereof. In the absence of conflict, the embodiments and technical features in the embodiments can be combined with each other.

[0026] Example 1: Figure 1 This is a flowchart of a low-voltage distribution transformer risk assessment and update optimization method and system according to Embodiment 1 of the present invention. This flowchart only illustrates the logical sequence of the methods described in this embodiment. Provided there are no conflicts, different methods may be used in other possible embodiments of the present invention. Figure 1 Perform the steps shown or described in the order indicated. See also Figure 1 The method implemented in this way specifically includes the following steps: Step 1: Acquire multi-source data through the Supervisory Control and Data Acquisition (SCADA) and Advanced Metrology Infrastructure (AMI) measurement module, the online dissolved gas analysis (DGA) oil chromatography online monitoring module, the event acquisition module, and the meteorological geographic information system (GIS) access module. Perform time registration on the multi-source data to make them equivalent to the same time scale.

[0027] Step 2: Estimate the hot spot temperature θHS[n] based on the top oil temperature-winding hot spot thermal model, and obtain the hourly aging factor FAA[n] according to the Arrhenius relation.

[0028] Step 3: Calculate the Equivalent Operating Life (EOA) and Health Index (HI) to conduct a dual-track health assessment.

[0029] Step 4: Estimate the failure probability based on health and external force characteristics to obtain the failure probability Pfail(Δ) for the target time window.

[0030] Step 5: Summarize the material and maintenance costs and the impact of power outages into a single failure consequence C, and then multiply it by the failure probability Pfail(Δ) within the window Δ to obtain the equipment risk R, thereby realizing the monetization and comparable ranking of risks.

[0031] Step 6: First, use NPVi(s) as the threshold to filter the schemes, select the best replacement scheme with the highest unit benefit ratio (RAER) for each device, then sort the entire network according to RAER from high to low and break up the tie according to the Φi(s) rule to generate an economic priority list.

[0032] The core idea of ​​this invention is to unify the multi-source operating data of low-voltage distribution transformers into a comparable index system, construct a failure probability-risk model, and perform economic evaluation to generate an economic priority list. The specific implementation process is as follows: In the data acquisition and processing stage, multi-source data are acquired through the Supervisory Control and Data Acquisition (SCADA) and Advanced Metrology Infrastructure (AMI) measurement module, the Online Dissolved Gas Analysis (DGA) oil chromatography online monitoring module, the event acquisition module, and the meteorological geographic information system (GIS) access module. The multi-source data are then time-registered to be equivalent to the same time scale.

[0033] Three-phase current, voltage, active power R, reactive power R, and harmonic spectrum are read using SCADA and AMI measurement modules and converted to the same time scale. Assuming a unified time step of Δt and a unified starting point t0, the aggregation formulas for the interval mean and interval extreme values ​​are as follows:

[0034] In the formula: x[n] refers to the nth time window The average time value within the window is given by Wn, where Wn is the length of the time window, Kn is the number of samples falling into Wn, and xmax[n] refers to the maximum value within the same window Wn.

[0035] Then calculate the key quantity: apparent power. Load factor K, current imbalance Uunb, and total harmonic distortion (THDI) are used as inputs for subsequent hot spot temperature rise and aging factors.

[0036] In the formula: K[n] is the load factor, in units of 1; S[n] is the interval average apparent power; Srated is the rated capacity of the transformer; Uunb[n] is the current imbalance, in units of 1; Ia, Ib, and Ic are the interval averages of the three-phase currents; Iavg=(Ia+Ib+Ic) / 3; THDI[n] is the total harmonic distortion of the current; Ih[n] is the root mean square of the h-th harmonic current; I1[n] is the current at the fundamental frequency (h=1); and h is the highest harmonic order calculated.

[0037] The dissolved gas contents (H2, CH4, C2H6, C2H4, C2H2, etc.), water content, and dielectric loss are read using the DGA online module and then converted to the same time scale, as follows:

[0038] In the formula: gj is the continuous concentration sequence obtained by piecewise linear interpolation, j belongs to the content of dissolved gases such as H2, CH4, C2H6, C2H4, C2H2, as well as water and medium loss, gj(ti) is the measured concentration of the j-th gas at time ti, and gj[n] is the average concentration of the j-th gas in Wn.

[0039] Then, the ratio of total dissolved gas TDCG to typical values ​​is calculated, as shown in the following formula:

[0040] In the formula: TDCG[n] is the total dissolved gas concentration, rjk[n] is the concentration ratio of the two gases, and ϵ is a very small positive number to avoid the denominator being zero.

[0041] The event acquisition module records the counts of short circuits, closing events, lightning strikes, and tripping events. The formula for calculating the interval event count is as follows:

[0042] In the formula: te is the event timestamp (such as short circuit / closing / lightning strike), and Nevent[n] is the number of events within Wn.

[0043] External force sequences such as thunderstorm density, icing thickness, extreme high-temperature days, and wind speed are obtained through the meteorological GIS access module. The formula for aggregating these external force sequences is as follows:

[0044] In the formula: Iq(t) is the value of the qth external force factor over time (such as thunderstorm density, ice thickness, extreme high temperature index, wind speed, etc.), and Iq[n] is the average external force value of Wn.

[0045] Align the datasets to obtain the dataset:

[0046] In estimating the hot spot temperature θHS[n] and aging factor FAA[n], the load rate K[n] and ambient temperature θamb at the same time scale Wn obtained in step 1 are used as inputs. The top oil temperature θTO[n] is calculated using a first-order thermal inertia model, and then the winding hot spot temperature θHS[n] is obtained by superimposing the increase in winding hot spot temperature, as shown below:

[0047] In the formula: θTO[n] is the top oil temperature in the nth time window, θHS[n] is the winding hot spot temperature in the nth time window, θamb[n] is the ambient temperature (from the meteorological module aligned to Wn), ΔθTO[n] is the increase in top oil temperature caused by load in the nth time window, ΔθHS[n] is the increase in winding hot spot temperature caused by load in the nth time window, ΔθTO,R is the increase in top oil temperature under rated load, ΔθHS,R is the increase in winding hot spot temperature under rated load, nto and nhs are the power exponents of the load temperature rise, and τTO is the first-order thermal time constant of the top oil temperature.

[0048] The θHS[n] and ΔθTO[n] sequences at the same time scale are obtained, providing temperature input for subsequent aging rate calculations.

[0049] The hotspot temperatures obtained in step 2 are then converted to absolute temperatures. The Arrhenius relative aging rate formula is used to obtain the window-by-window aging acceleration factor FAA[n], and the thermally equivalent aging time EAAT is accumulated over time. The expression is as follows:

[0050] In the formula: FAA[n] is the aging acceleration factor of the nth window. It is the absolute temperature of the nth window. Tref is the reference absolute temperature, Ea / R is the ratio of the equivalent activation energy to the gas constant, EAAT is the thermal equivalent aging time within the selected statistical interval, and Δth is the time step value in hours.

[0051] In constructing a dual-track health assessment using the Equivalent Operating Age (EOA) and Health Index (HI), firstly, the operational, environmental, and chemical information obtained from steps 1 and 2 is summarized into three categories of cumulative stresses: overload severity, damp heat effect, and shock count, which serve as the basis for the lifespan conversion.

[0052] In the formula: Eover is the cumulative overload severity, K[k] is the load rate of the k-th window, Kthr is the overload threshold, p is the severity amplification index, and Δth is the time step converted to hours, as shown in the following formula:

[0053] In the formula: Imoist is the cumulative amount of the influence of damp heat, w¯[k] is the time window average of the water content in the oil in the k-th window, tanδ¯[k] is the average of the medium loss factor in the k-th window, and aw and atanδ are the weights that unify the water content with the dimensionless quantity, as shown in the following formula:

[0054] In the formula: Nshock is the total number of shock events during the assessment period, and Nevent[k] is the number of events in the k-th window.

[0055] Subsequently, the equivalent thermal aging hours (EAAT) in step 2 and the three types of stresses in step 3 are equivalently converted into service life increments and added to the actual service life to obtain the equivalent service life (EOA). The calculation formula is as follows:

[0056] In the formula: EOA is the equivalent operating years, Age is the actual service years of the equipment, EAAT is the thermal equivalent aging hours, αT is the thermal aging conversion factor, αE is the year conversion factor for overload severity, αM is the year conversion factor for damp heat influence, and αS is the year conversion factor for shock count. EOA will be used together with HI from step 3 for health classification and as one of the explanatory variables for hazard rate modeling in step 4.

[0057] Then, by extracting state features from steps 1 and 2, a uniform 0-1 monotonicity is applied to ensure that a larger value indicates a worse performance. The formula is as follows:

[0058] In the formula: xi[n] is the i-th original feature (from the Daligned dataset), li and ui are reasonable lower / upper bounds of the i-th feature, and si is the 0-1 feature after unifying the direction. The larger the value, the worse the health. The result is a feature set {si}i=1d with consistent direction.

[0059] A non-negative weighted linear interpretable scoring method is used to output a health index of 0-100. A constrained learning form for weight calculation is also provided to facilitate automatic weight calibration in labeled scenarios, as shown in the following equation:

[0060] In the formula: HI is the health index (0-100, the higher the value, the healthier), and wi is the weight of the i-th feature.

[0061] Finally, the EOA from the age perspective and the HI from the state perspective are merged into a joint score S, and a grading threshold is generated using the quantile method, outputting grades H0 to H4 (H4 being the most severe). The joint score calculation formula is as follows:

[0062] In the formula: S is the combined health deterioration score (dimension controllable from 0 to 2, the larger the value, the worse), λ1 and λ2 are the two-view weights λ1+λ2=1, λ.≥0, EOAmin and EOAmax are the minimum and maximum reference values ​​of EOA in history or within a group, and g1 and g2 are the monotonic mappings that normalize EOA and HI to 0-1.

[0063] The threshold is generated using the quantile method, and the mapping from the result to the level is calculated using the following formula:

[0064]

[0065]

[0066] In the formula: Qp(S) is the p-th quantile of the score S, τS(k) is the k-th threshold used for grading, τEOAhi is the "high age" threshold of EOA, and τHIlo is the "low health" threshold of HI. The output is the level label Level∈{H0,H1,H2,H3,H4} for each device. EOA, HI, and Level serve as key inputs for hazard rate modeling in step 4.

[0067] In estimating the failure probability Pfail(Δ) of the target time window based on the aforementioned health and external force characteristics, the equivalent operating years (EOA) and health index (HI) are first normalized. Then, the external forces (thunderstorms, icing, high temperatures, and wind) in the recent period are normalized to 0-1 and weighted and summed. Finally, the three factors are added together according to their weights to obtain R0. The calculation formula is as follows:

[0068] In the formula: EOAnorm is the 0-1 normalized result of EOA (the larger the value, the older the result), HInorm is the 0-1 reverse normalized result of HI (the larger the value, the worse the result), Iq¯(14d) represents the time average of the past 14 days, wq is the external force channel weight, Stress is the 0-1 comprehensive external force intensity (the larger the value, the more unfavorable the result), a1, a2, and a3 are the three-branch weights, and R0 is the 0-1 comprehensive risk factor.

[0069] Then, a base annual failure rate λ0 table is set according to the health level. This can be obtained from historical statistics or empirical methods, and λ0 is then amplified and converted into a window failure probability, the formula of which is as follows:

[0070] In the formula: λ0(Level) is the base annual failure rate corresponding to this level. Δ is the target window length, λΔ is the equivalent failure rate within the window, and Pfail(Δ) is the failure probability within the window.

[0071] In the stage of monetizing and ranking risks, the costs of materials and maintenance, as well as user losses caused by power outages, are first combined to obtain the cost C of a single failure consequence:

[0072] In the formula: Cfix is ​​the material cost plus maintenance cost (including tools, transportation, etc.), VOLL is the value loss per hour of power outage per user, Nuser is the number of affected users, Tout is the expected power outage duration, and wimp is the weight of important users.

[0073] Then, the risk R is calculated to obtain the risk value for each device, as shown in the following formula:

[0074] In the formula: R is the baseline risk value of the device within the window Δ. Furthermore, the following four feasible replacement options are listed for device i: Replacement with the same capacity: New version with the same capacity.

[0075] Replacement with a high-efficiency model of the same capacity: Replace with a high-efficiency model of the same capacity (lower no-load / load loss).

[0076] Capacity increase and replacement: Increase capacity (e.g., 200 → 315 kVA), and upgrade switches and cable cross-sections if necessary.

[0077] Relocation and replacement: New site, complete civil engineering, platform, and secondary side relocation, and adjust the line if necessary.

[0078] For each option s, calculate the annualized or windowed benefits, including risk mitigation benefits, operational and maintenance savings, and energy efficiency benefits, using the following formulas:

[0079] In the formula: ΔRi(s) is the risk reduction brought about by strategy s, ΔOi(s) is the reduction in annual operation and maintenance costs after replacement, ΔEi(s) is the annual electricity cost saving, Ri,before and Rafter,i(s) are the risk values ​​before and after equipment replacement, Obee,i and Obee,i are the annual operation and maintenance costs before and after equipment replacement, and Ploss,oldt and Ploss,new(s)(t) are the power loss of the original / new equipment.

[0080] Then, using net present value (NPV) as a threshold, feasible options are first screened to obtain a feasible set of transformer replacement options. The NPV calculation formula is as follows:

[0081] In the formula: Cone,i(s) is the one-time total cost of option s, and Y is the number of years for economic evaluation. When NPVi(s)>0, it indicates economic feasibility, and the next step of ranking is performed; NPVi(s)≤0 indicates that no change is made for the time being.

[0082] For each feasible solution s, two types of indicators are given: unit benefit ratio and power outage friendliness, which are used for horizontal ranking and parallel elimination. The calculation formulas for these indicators are as follows:

[0083] In the formula: RAERep,i(s) is the unit benefit ratio of scheme s, and the larger the ratio, the higher the priority. Φi(s) is the outage friendliness of scheme s, and the closer it is to 1, the shorter the outage or the zero outage, and the more friendly it is.

[0084] In the set of feasible solutions for device i, the most cost-effective one is selected, as shown below:

[0085] In the formula, si⋆ represents the optimal replacement plan for device i. Extract the si⋆ of all devices and sort them from highest to lowest according to the primary key RAERrep,i(s). Φi(s) is used to break up ties. This yields a transparent, auditable, and reproducible list of economic rankings: first, select the best device within the group (NPVi(s) > 0 and RAERrep,i(s) is maximized), then sort between devices (primary key RAERrep,i(s) + Φi(s)), which directly serves as the economic priority input for the update plan.

[0086] For example, in a distribution area or feeder containing multiple low-voltage distribution transformers, the system first obtains the initial state variables (load rate, unbalance, harmonics, etc.) and their rate of change for each device through SCADA and AMI. Then, combined with information such as events, DGA oil chromatography, and external force intensity from meteorological GIS, the windowed prediction model will provide the forward value of the failure probability and risk for each device in the next assessment period, and will be updated on a rolling basis when new data arrives. At the same time, it completes the fields such as the health level, risk growth rate, geographical location, device ID, and estimated power outage duration and user impact of the device, forming asset semantic information that includes health level, location, economic consequences (primary consequence C), and time-series risk trajectory.

[0087] The asset semantic information is used in the subsequent economic assessment and update decision module: the state characteristics (HI, EOA), external force intensity, and user-side impact (VOLL × number of users × planned power outage duration) are weighted and fused on the same time scale, and the weights are adaptive according to the scenario (e.g., increase the weight of load and temperature during high-temperature periods, and increase the weight of lightning exposure before thunderstorms) to obtain the fused risk-benefit characteristics; on this basis, the NPV threshold and unit benefit ratio RAER are calculated for the four candidate schemes respectively, and the power outage duration voltage drop and user loss reduction brought by the parallel supply / mobile transformer are explicitly included, and the recommended results of whether to replace, how to replace, and when to replace are output.

[0088] Before the arrival of summer's high temperatures, the system identified several urban transformer substations with high loads and high VOLL. For these devices, the fusion characteristics showed that the main risks were load-driven, and the decision-making module tended to upgrade or replace them with high-efficiency replacements of the same capacity. If the user was of high importance, the planned power outage was reduced to a minute-level switching, and a priority implementation list and suggested time windows were provided based on RAER (Rapid Availability and Responsibility).

[0089] For parks with high VOLL and significant user weight, the system prioritizes evaluating efficient or capacity-increasing solutions with the same capacity, and incorporates parallel power supply and mobile variable parameters into NPV and RAER. When the economics of the two solutions are similar, outage friendliness and risk reduction are used as parallel breakthroughs, and the implementation path of zero outage or minute-level outage is recommended.

[0090] Example 2: This embodiment of the invention also provides a method and system for risk assessment and optimization of low-voltage distribution transformers, characterized by comprising: The multi-source data acquisition module is used to acquire three-phase current, voltage, active power, reactive power, harmonics and event records through SCADA and event acquisition modules respectively, to acquire dissolved gas and water content and dielectric loss indicators in oil through DGA online monitoring, and to acquire external force information such as thunderstorm density, ice thickness, extreme high temperature index and wind speed through meteorological GIS access; The data preprocessing and alignment module is used to extract features such as load rate, imbalance, total harmonic distortion, TDCG and gas ratio, event density and external force intensity, and to perform 0-1 monotonicization according to preset upper and lower boundaries. The Health Assessment and Risk Monetization module is used to calculate EOA and HI and generate a health level Level based on them. The module then looks up the base annual failure rate based on the Level and calculates the target window failure probability by combining it with the comprehensive factor R0. The module combines material maintenance costs and power outage value loss into a primary consequence C and calculates the risk R. The strategy optimization module is used to filter solutions. It first filters based on NPVi(s) as the threshold, selects the solution with the highest RAER as the optimal replacement solution, and then sorts the entire network according to RAER from high to low and breaks up ties according to Φi(s) rules to generate an economic priority list.

[0091] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for risk assessment and optimization of low-voltage distribution transformers, characterized in that, include: Step 1: Acquire multi-source data through the Supervisory Control and Data Acquisition (SCADA) and Advanced Metrology Infrastructure (AMI) measurement module, the online dissolved gas analysis (DGA) oil chromatography online monitoring module, the event acquisition module, and the meteorological geographic information system (GIS) access module. Perform time registration on the multi-source data to make them equivalent to the same time scale. Step 2: Estimate hot spot temperature based on top oil temperature-winding hot spot thermal model The time-dependent aging factor was obtained based on the Arrhenius-Arrhenius relation. ; Step 3: Calculate the Equivalent Operating Years (EOA) and Health Index (HI) to conduct a dual-track health assessment; Step 4: Estimate the failure probability based on health and external force characteristics to obtain the failure probability within the target time window. ; Step 5: Summarize the material and repair costs, as well as the impact of the power outage, into a single failure consequence. Then with the window Failure probability within Multiplication yields equipment risk This enables the monetization and comparable ranking of risks. Step 6: with To screen for solutions, the optimal replacement solution with the highest RAER (Return on Efficiency) for each device is selected. Then, the entire network is sorted by RAER from highest to lowest. The rules are broken down into parallel categories to generate an economic priority list.

2. The method for risk assessment and optimization of low-voltage distribution transformers according to claim 1, characterized in that, Multi-source data is acquired through the Supervisory Control and Data Acquisition (SCADA) and Advanced Metrology Infrastructure (AMI) measurement module, the online dissolved gas analysis (DGA) oil chromatography online monitoring module, the event acquisition module, and the meteorological geographic information system (GIS) access module, including: Three-phase current, voltage, and active power are obtained through SCADA and AMI measurement modules. P No merit Q The harmonic spectrum, equivalent to the same time scale, has the following formulas for its interval mean and interval extreme values: In the formula: It refers to the first a time window Average time within, The length of the time window. To fall into The number of samples, Refers to the same window The maximum value within; The DGA online module is used to obtain the content of dissolved gases such as H2, CH4, C2H6, C2H4, and C2H2, as well as water content and dielectric loss. In the formula: The continuous concentration sequence is obtained by piecewise linear interpolation. It includes the content of dissolved gases such as H2, CH4, C2H6, C2H4, and C2H2, as well as water content and medium loss; The event acquisition module records the counts of short circuits, closing events, lightning strikes, and tripping events. The formula for calculating the interval event count is as follows: In the formula: For event timestamps (such as short circuit / closing / lightning strike). for Number of internal events; External force sequences such as thunderstorm density, icing thickness, extreme high-temperature days, and wind speed are obtained through the meteorological GIS access module. The formula for aggregating these external force sequences is as follows: In the formula: It is the first The values ​​of external force factors over time (e.g., thunderstorm density, icing thickness, extreme high temperature index, wind speed). yes Average external force value.

3. The method for risk assessment and optimization of low-voltage distribution transformers according to claim 1, characterized in that, Hot spot temperature estimated based on top oil temperature-winding hot spot thermal model According to the Arrhenius relation, the time-dependent aging factor was obtained. ,include: The top oil temperature was calculated using a first-order thermal inertia model. The temperature of the winding hot spot is then obtained by superimposing the increase in the winding hot spot temperature. The calculation formula is as follows: In the formula: It is the first Top oil temperature within a time window, It is the first Winding hotspot temperature within a time window It is the ambient temperature (from the meteorological module aligned to) ), It is the first The amount of top oil temperature rise caused by load within a time window It is the first The amount of increase in winding hotspot due to load within a time window. The top oil temperature rises under rated load. The hot spot of the winding increases under rated load. It is the power exponent of load-induced temperature rise. It is the first-order thermal time constant of the top oil temperature; The window-by-window aging acceleration factor was obtained using the Arrhenius relative aging rate formula. And the thermal equivalent aging time is accumulated over time. As shown in the following formula: In the formula: It is the first Window aging acceleration factor, It is the first The absolute temperature of each window, For reference absolute temperature, The ratio of the equivalent activation energy to the gas constant. To select the thermal equivalent aging time within the statistical interval, The value represents the time step in hours.

4. The method for risk assessment and optimization of low-voltage distribution transformers according to claim 1, characterized in that, Calculate the Equivalent Operating Age (EOA) and Health Index (HI) to conduct a dual-track health assessment, including: Operational, environmental, and chemical information is aggregated into three categories of accumulative stresses: overload severity, damp heat effect, and shock count, which serve as the basis for calculating service life; equivalent thermal aging hours are also included. The equivalent service life (EOA) is obtained by converting the three types of stress into an equivalent service life increment and adding it to the actual service life. The calculation formula is as follows: In the formula: For equivalent operating years, This refers to the actual service life of the equipment. This refers to the thermal equivalent aging time. This is the thermal aging conversion factor. The years of service are used to calculate the severity of the overload. This is the annual conversion factor for the influence of heat and humidity. The conversion factor for the number of years of impact counts; State features are extracted and uniformly monotonicized to 0-1 to ensure that larger values ​​indicate worse health. A non-negative weighted linear interpretable scoring method is used to output a health index of 0-100. A constrained learning form for weight calculation is also provided to facilitate automatic weight calibration in labeled scenarios. The EOA from the age perspective and the HI from the state perspective are fused into a joint score S, and a quantile method is used to generate grading thresholds, outputting levels H0 to H4. The calculation formula is as follows: In the formula: The first for grading A threshold, yes The "high age" threshold, yes The "low health" threshold; output the rating label for each device. .

5. The method for risk assessment and optimization of low-voltage distribution transformers according to claim 1, characterized in that, The equivalent operating years (EOA) and health index (HI) are normalized, and the external forces in the recent period are normalized to 0-1 and weighted and summed. The weighted sums are then used to obtain the comprehensive risk factor. A baseline annual failure rate is set based on the health level. Table, and put The failure probability within the window is then magnified and converted to a window value, as shown in the following formula: In the formula: It is the target window length. It is the equivalent failure rate within the window.

6. The method for risk assessment and optimization of low-voltage distribution transformers according to claim 1, characterized in that, The costs of materials and repairs, as well as the impact of the power outage, are summarized as the consequences of a single fault. Then with the window Failure probability within Multiplication yields equipment risk The formula is as follows: In the formula: It is the sum of material costs and maintenance costs (including tools, transportation, etc.). It represents the value loss per hour of power outage for each user. This refers to the number of affected users. Expected power outage duration: It is an important user weight. Is the device in the window? The baseline risk value within.

7. The method for risk assessment and optimization of low-voltage distribution transformers according to claim 1, characterized in that, Four feasible replacement options are provided. For each scheme Calculate annualized or windowed benefits, including risk mitigation benefits, operational savings, and energy efficiency benefits; use net present value (NPV) as a threshold to screen feasible solutions, with the NPV calculation formula as follows: In the formula It is a plan The total cost in one go It is the number of years for economic assessment; when If the situation is deemed economically feasible, proceed to the next step of sorting. This indicates that replacement will not be carried out for the time being; by combining two indicators, namely unit benefit ratio and power outage friendliness, an economic ranking list is obtained.

8. A risk assessment and optimization system for low-voltage distribution transformers, characterized in that, include: The multi-source data acquisition module is used to acquire three-phase current, voltage, active power, reactive power, harmonics and event records through SCADA and event acquisition modules respectively, to acquire dissolved gas and water content and dielectric loss indicators in oil through DGA online monitoring, and to acquire external force information such as thunderstorm density, ice thickness, extreme high temperature index and wind speed through meteorological GIS access; The data preprocessing and alignment module is used to extract features such as load rate, imbalance, total harmonic distortion, TDCG and gas ratio, event density and external force intensity, and to perform 0-1 monotonicization according to preset upper and lower boundaries. The health assessment and risk monetization module is used to calculate EOA and HI and generate a health level based on them. The baseline annual failure rate is then obtained by looking up the level and combining it with a comprehensive factor. Calculate the probability of target window failure; Combine material repair costs and power outage value loss into a single consequence C and calculate the risk R; The strategy optimization module is used to filter solutions, in order to First, a threshold is used for screening, and the option with the highest RAER is selected as the optimal replacement solution. Then, the entire network is sorted from highest to lowest RAER and... The rules are broken down into parallel categories to generate an economic priority list.

Images