Comprehensive risk assessment method and system for distribution transformer based on RF-GCMs

Abstract

The invention discloses a distribution transformer comprehensive risk assessment method and system based on RF-GCMs, and the method comprises the steps: generating local hour-level meteorological prediction data according to local historical meteorological data, local meteorological measurement data and GCMs typical climate daily average meteorological prediction data; determining an insulation aging acceleration factor, and determining a dynamic fault rate of the transformer based on the insulation aging acceleration factor; calculating the actual output of the photovoltaic output device according to the local hour-level weather prediction data and the rated output of the photovoltaic output device; calculating load shedding loss, determining service life loss of the distribution transformer according to the insulation aging acceleration factor, and calculating light abandoning loss based on actual output of the photovoltaic output device; and determining the comprehensive risk loss of the distribution transformer based on the dynamic fault rate of the transformer, the service life loss of the distribution transformer, the load shedding loss and the light abandoning loss.

Description

Technical Field

[0001] This invention relates to the field of transformer risk assessment technology, and more specifically, to a method and system for comprehensive risk assessment of distribution transformers based on RF-GCMs. Background Technology

[0002] To achieve sustainable energy consumption, improve the ecological environment, and mitigate global climate change, governments worldwide are vigorously developing renewable energy power generation technologies, such as wind and solar power, and actively promoting the green energy transition. Distributed photovoltaic (PV) power has been widely deployed in modern power distribution systems due to its advantages, including cleanliness, sustainability, flexible installation, and low transmission losses. Simultaneously, with the increasing penetration of renewable energy sources like wind and solar power in power distribution systems, the correlation between the risks of distribution transformers and meteorological factors has become increasingly close, given the strong correlation between wind and solar power and meteorological elements.

[0003] Distribution transformers play a crucial role in power distribution networks, serving as the final stage of voltage transformation. They are numerous and represent a significant asset of the distribution network. Therefore, scholars have proposed various methods to scientifically and comprehensively assess the health status and risks of distribution transformers, such as dissolved gas analysis in oil, frequency response analysis, online partial discharge testing, and vibration analysis. However, most of these methods require the installation of sensors or offline monitoring, resulting in high economic costs. Furthermore, the large number of distribution transformers and their relatively low individual cost make these methods unsuitable.

[0004] Lifetime loss calculation is also a method for risk assessment of distribution transformers. In most cases, the lifespan of a distribution transformer is equivalent to its insulation lifespan. Current lifetime loss research mostly relies on IEEE Std C57.91 to calculate the winding hot spot temperature based on transformer load rate and ambient temperature, further obtaining the insulation life loss, and then conducting risk assessment based on the remaining useful life (RUL). Some scholars calculate insulation life loss based on transformer temperature rise, and then use cluster analysis to quickly identify different load modes and risk levels, formulating effective risk management strategies. Other scholars use machine learning to calculate winding hot spot temperature rise based on ambient temperature and load rate, using Bayesian particle filtering to reduce the impact of measurement errors, considering the uncertainty of initial life and measurement error to finally obtain the probability density function of remaining life. Still other scholars have proposed a decision model to make trade-offs between improving transmission capacity and slowing down the transformer aging rate. Although this method cannot directly obtain the specific failure time of the transformer, it can reflect the insulation aging process of the distribution transformer to a certain extent, and the insulation aging condition of the transformer is also closely related to its failure rate. Summary of the Invention

[0005] Research on risk assessment of distribution transformers based on lifetime loss mainly faces the following problems: First, most studies rely solely on historical data or machine learning methods to predict the insulation lifetime loss of distribution transformers. However, with global warming, frequent extreme weather events, and increasing interannual climate differences, traditional risk assessment methods based on historical data are no longer sufficient to accurately reflect future risks. Machine learning methods, on the other hand, suffer from dependence on data quality, reduced generalization ability due to overfitting, and insufficient model interpretability. Second, few studies apply such models in distribution transformer risk assessment. Finally, most studies do not consider the correlation between load factor and meteorological factors when modeling lifetime loss. However, with the continuous integration of distributed photovoltaic systems, the connection between distribution transformer load factor and meteorological factors is becoming increasingly close.

[0006] Therefore, in order to effectively improve the efficiency and accuracy of risk assessment of distribution transformers after large-scale integration of new energy sources, and considering the influence of the correlation between weather, load, and photovoltaic output, this invention provides a comprehensive risk assessment method and system for distribution transformers based on RF-GCMs.

[0007] According to a first aspect of the present invention, a comprehensive risk assessment method for distribution transformers based on RF-GCMs is provided, comprising:

[0008] Based on local historical meteorological data, local meteorological measurement data, and typical daily average meteorological forecast data of GCMs, local hourly meteorological forecast data is generated.

[0009] Determine the insulation aging acceleration factor, and based on the insulation aging acceleration factor, determine the transformer dynamic failure rate;

[0010] Calculate the actual output of the photovoltaic power generation device based on local hourly meteorological forecast data and the rated output of the photovoltaic power generation device;

[0011] Calculate load shedding losses, determine the life loss of distribution transformers based on insulation aging acceleration factors, and calculate curtailment losses based on the actual output of photovoltaic power generation devices.

[0012] The comprehensive risk loss of distribution transformers is determined based on the dynamic failure rate of transformers, life loss of distribution transformers, load shedding loss, and curtailment loss.

[0013] Optionally, based on local historical meteorological data, local meteorological measurement data, and typical daily average meteorological forecast data of GCMs, local hourly meteorological forecast data are generated, including:

[0014] Based on local historical meteorological data, a random forest model (RF) was determined for the local daily average and hourly meteorological curves.

[0015] Based on local meteorological measurement data, the typical daily average meteorological forecast data of GCMs is corrected to obtain the local daily average meteorological forecast data of GCMs.

[0016] Local hourly weather forecast data are generated based on the random forest model (RF) and GCMs (Gross Meteorological Contexts) of local daily and hourly weather curves.

[0017] Optionally, based on local meteorological measurement data, the typical daily average meteorological forecast data of GCMs are corrected to obtain the local daily average meteorological forecast data of GCMs, including:

[0018] The accuracy of typical daily average meteorological forecast data of GCMs is measured by the root mean square error (RMSE), mean absolute error (MAE), and symmetric mean absolute percentage error (SMAPE). The specific calculation formulas are as follows:

[0019]

[0020]

[0021] Among them, y i This is the daily average actual meteorological data. represents the daily average weather forecast data, and n represents the sample size.

[0022] Optionally, an insulation aging acceleration factor is determined, and based on the insulation aging acceleration factor, the transformer dynamic failure rate is determined, including:

[0023] Insulation aging acceleration factor F is adopted AA The insulation aging rate of a transformer is described by (t), and the specific calculation formula is as follows:

[0024]

[0025] Where, θ H (t) represents the winding hot spot temperature, θ H The formula for calculating (t) is as follows:

[0026]

[0027] Where γ is the ratio of load loss at rated load to loss at zero load, and θ A i(t) represents the ambient temperature, and i(t) represents the transformer load at time t. r It is the rated load, Δθ H.R Δθ is the temperature rise of the winding hot spot under rated load. TO.R The oil temperature is at rated load. m and n are transformer parameters determined by looking up a table based on the transformer's cooling system. For most distribution oil-immersed transformers, m and n are both 0.8.

[0028] The insulation loss of the transformer is equivalent to the reduction in the remaining service life (RUL) under standard operating conditions by using an insulation aging acceleration factor. The specific recursive relationship is as follows:

[0029]

[0030] Based on the insulation aging acceleration factor, the transformer failure rate is modeled, and the specific formula is as follows:

[0031]

[0032] Among them, t e The equivalent operating time is obtained by subtracting the initial lifetime from the RUL, where B is a constant of 15000, β is 5.9, and C is 1.76 × 10⁻⁶. -12 θ0 is the rated operating temperature of the transformer, taken as 110℃.

[0033] Optionally, based on local hourly meteorological forecast data and the rated output of the photovoltaic power generation device, the actual output of the photovoltaic power generation device is calculated, including:

[0034] The influence of solar irradiance and temperature on photovoltaic output is considered through a photovoltaic output formula, as follows:

[0035]

[0036] In the formula: P PV The actual output of the photovoltaic power generation device; P N The rated output of the photovoltaic power generation device; G

[0037] G represents the actual solar irradiance. N The solar irradiance under standard test conditions is taken as 1000 W / m. 2 ;α P T represents the power temperature coefficient of the photovoltaic power generation device, taken as -0.35% / ℃; C T0 is the temperature of the photovoltaic power generation device; T0 is the ambient temperature; Tb is the temperature of the photovoltaic power generation device. STC The battery temperature under standard test conditions is taken as 25℃.

[0038] Optionally, calculate the load shedding loss, determine the life loss of the distribution transformer based on the insulation aging acceleration factor, and calculate the curtailment loss based on the actual output of the photovoltaic power generation unit, including:

[0039] According to the insulation aging acceleration factor F AA (t), determine the life loss S of the distribution transformer. I The specific formula is as follows:

[0040]

[0041] In the formula: L0 is the rated service life of the transformer, which is 180,000 hours according to IEEE C57.91; c trans

[0042] This refers to the purchase price of the transformer;

[0043] The formula for calculating load shedding loss is shown below:

[0044]

[0045] In the formula: P D (i) represents the load at time i, and c loss The load shedding penalty price is given, t is the time of transformer failure, and τ is the transformer failure repair time.

[0046] Based on the actual output of the photovoltaic power generation device, the curtailment loss is calculated using the following formula:

[0047]

[0048] In the formula: P PVG (i) represents the photovoltaic output at time i, and c PV The price is the penalty for abandoning light.

[0049] Optionally, based on the transformer dynamic failure rate, distribution transformer life loss, load shedding loss, and curtailment loss, the comprehensive risk loss of the distribution transformer is determined, including:

[0050] The specific formula for calculating the comprehensive risk loss of distribution transformers is as follows:

[0051] R isk =S I +p outage (t e |θ hst (S) LS +S PV (12).

[0052] According to another aspect of the present invention, a comprehensive risk assessment system for distribution transformers based on RF-GCMs is provided, comprising:

[0053] The weather forecast data generation module is used to generate local hourly weather forecast data based on local historical weather data, local weather measurement data, and typical daily average weather forecast data of GCMs.

[0054] The dynamic failure rate determination module is used to determine the insulation aging acceleration factor and, based on the insulation aging acceleration factor, to determine the transformer dynamic failure rate.

[0055] The module for calculating actual photovoltaic output is used to calculate the actual output of the photovoltaic power generation device based on local hourly meteorological forecast data and the rated output of the photovoltaic power generation device.

[0056] The transformer life loss determination module is used to calculate load shedding losses, determine the life loss of distribution transformers based on insulation aging acceleration factors, and calculate curtailment losses based on the actual output of photovoltaic power generation devices.

[0057] The comprehensive risk loss determination module is used to determine the comprehensive risk loss of distribution transformers based on the transformer dynamic failure rate, distribution transformer life loss, load shedding loss, and curtailment loss.

[0058] According to another aspect of the invention, a computer-readable storage medium is also provided, on which a computer program is stored, characterized in that the program, when executed by a processor, implements the steps of any of the methods described herein.

[0059] According to another aspect of the present invention, an electronic device is also provided, comprising:

[0060] The computer-readable storage medium; and one or more processors for executing a program in the computer-readable storage medium.

[0061] Therefore, the Random Forest (RF) model was used to refine the data granularity of Global Climate Models (GCMs) to obtain hourly meteorological forecast data. Considering the correlation between meteorology, load, and photovoltaic output, the load rate of distribution transformers was predicted. The typical day method was used to comprehensively assess the risks of accelerated insulation aging, load loss, and curtailment, and finally, the comprehensive risk assessment results of distribution transformers were obtained. Attached Figure Description

[0062] Exemplary embodiments of the present invention can be more fully understood by referring to the following figures:

[0063] Figure 1 This is a flowchart illustrating a comprehensive risk assessment method for distribution transformers based on RF-GCMs as described in this embodiment.

[0064] Figure 2 This is a schematic diagram of the hourly meteorological forecast data generation framework described in this embodiment;

[0065] Figure 3 This is a schematic diagram of the load forecasting framework described in this embodiment;

[0066] Figure 4 This is a schematic diagram of the risk assessment process based on generated predictive data as described in this embodiment;

[0067] Figure 5 This is a schematic diagram of the calculation example described in this embodiment;

[0068] Figure 6 This is a schematic diagram of the hourly curves of historical temperature and actual load in a certain area of ​​Shanghai in 2022, as described in this embodiment.

[0069] Figure 7 This is a schematic diagram of the hourly predicted temperature and predicted load curves for a certain area in Shanghai in 2023, as described in this embodiment.

[0070] Figure 8 is a schematic diagram of the predicted results of the typical daily operation scenario of the distribution transformer in summer 2023 as described in this embodiment. Specifically: (a) temperature curve of typical summer day, (b) photovoltaic curve of typical summer day, (c) load curve of typical summer day;

[0071] Figure 9 This is a schematic diagram of the RUL evaluation results of the distribution transformer using different data as described in this embodiment;

[0072] Figure 10 This is a schematic diagram illustrating the impact of different photovoltaic installation ratios on the risk of distribution transformers as described in this embodiment.

[0073] Figure 11 This is a schematic diagram of a comprehensive risk assessment system for distribution transformers based on RF-GCMs as described in this embodiment. Detailed Implementation

[0074] Exemplary embodiments of the invention will now be described with reference to the accompanying drawings. However, the invention may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided to fully and completely disclose the invention and to fully convey its scope to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the drawings is not intended to limit the invention. In the drawings, the same units / elements are referred to by the same reference numerals.

[0075] Unless otherwise stated, the terms used herein (including technical terms) have their common meaning as understood by one of ordinary skill in the art. Furthermore, it is understood that terms defined in commonly used dictionaries should be understood to have a meaning consistent with the context of their relevant field, and not to be interpreted as having an idealized or overly formal meaning.

[0076] According to a first aspect of the present invention, a comprehensive risk assessment method 100 for distribution transformers based on RF-GCMs is provided, with reference to... Figure 1 As shown, the method 100 includes:

[0077] S101: Generate local hourly meteorological forecast data based on local historical meteorological data, local meteorological measurement data, and typical daily average meteorological forecast data of GCMs.

[0078] S102: Determine the insulation aging acceleration factor, and based on the insulation aging acceleration factor, determine the transformer dynamic failure rate;

[0079] S103: Calculate the actual output of the photovoltaic power generation device based on local hourly meteorological forecast data and the rated output of the photovoltaic power generation device;

[0080] S104: Calculate load shedding losses, determine the life loss of distribution transformers based on insulation aging acceleration factors, and calculate curtailment losses based on the actual output of photovoltaic power generation devices.

[0081] S105: Determine the comprehensive risk loss of distribution transformers based on transformer dynamic failure rate, distribution transformer life loss, load shedding loss, and curtailment loss.

[0082] Specifically, a refined forecasting method for meteorological data based on RF-GCMs.

[0083] Global climate models (GCMs) construct a global climate prediction framework by integrating the different climatic characteristics of various atmospheric and oceanic regions on the Earth's surface into a complete Earth system model. These models, through detailed modeling of multiple factors such as radiative forcing, ocean circulation, and snow and ice feedback, can obtain relatively accurate meteorological forecast data. The NASA Climate Adaptation Science Investigators (CASI) project, based on data from six global climate models, has conducted climate predictions for 13 NASA centers and their regions, obtaining daily average meteorological forecast data. This paper, based on the daily average meteorological forecast results from the CASI project, further extends them to hourly average meteorological forecasts for other similar climate regions using a random forest algorithm.

[0084] Although Global Meteorological Conditions (GCMs) have good accuracy, considering the differences in meteorological conditions across regions, it is necessary to derive the difference in average temperature between two locations based on the most recent year's measurement data and then correct the daily average meteorological curves obtained from GCMs. To further improve forecast accuracy, this paper comprehensively evaluates and compares the accuracy of daily average meteorological data predicted by GCMs and selects the daily average meteorological forecast curve with the highest forecast accuracy for subsequent processing. Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Symmetric Mean Absolute Percentage Error (SMAPE) are three common indicators for measuring forecast error. This paper uses these three indicators to measure the accuracy of meteorological forecast data, and the specific calculation formulas are as follows:

[0085]

[0086] Among them, y i This is the daily average actual meteorological data. represents the daily average weather forecast data, and n represents the sample size.

[0087] Random forest is a classic machine learning method widely used in regression and classification tasks, and it has high accuracy and robustness.

[0088] This paper uses Random Forest (RF) to mine the relationship between daily average meteorological curves and hourly meteorological curves. First, a large amount of historical meteorological data is cleaned and standardized. Then, annual, monthly, daily, and daily average meteorological data are extracted as feature variables, and hourly meteorological data is used as the target variable. Finally, the data is divided into training and test sets to train a Random Forest model. This model can further generate more refined hourly meteorological forecast data based on the daily average meteorological data predicted by the corrected GCMs (Geometrical Weather Modulations), laying the foundation for subsequent transformer insulation aging risk assessment. (References) Figure 2 As shown, Figure 2 This paper demonstrates a framework for generating hourly meteorological forecast data using RF based on daily average forecast data of GCMs.

[0089] The IEEE Std C57.91 standard, based on the Arrhenius model, uses the remaining service life of a transformer to measure its insulation condition. Specifically, it represents the remaining operational time of a transformer under standard operating conditions (ambient temperature 30°C, rated load) until its insulation performance deteriorates to the point where it can no longer meet insulation requirements. An insulation aging acceleration factor F is used. AA The insulation aging rate of a transformer is described by (t), and the specific calculation formula is as follows:

[0090]

[0091] Where, θ H (t) represents the hot spot temperature of the winding.

[0092] In most cases, the operating state of a transformer can be approximated as steady state, therefore the dynamic process of heat transfer can be ignored, θ H The formula for calculating (t) is as follows:

[0093]

[0094] Where γ is the ratio of load loss at rated load to loss at zero load, and θ A i(t) represents the ambient temperature, and i(t) represents the transformer load at time t. r It is the rated load, Δθ H.R Δθ is the temperature rise of the winding hot spot under rated load. TO.R The oil temperature is at rated load. m and n are transformer parameters determined by looking up a table based on the transformer's cooling system. For most distribution oil-immersed transformers, m and n are both 0.8.

[0095] The insulation loss of the transformer is equivalent to the reduction in RUL under standard operating conditions by using an insulation aging acceleration factor. The specific recursive relationship is as follows:

[0096]

[0097] Insulation aging in transformers not only affects the transformer's rated uptime (RUL) but is also closely related to its failure risk; transformers with more severe insulation aging are more prone to failure. To quantify the impact of insulation aging on transformer failure rate, the Arrhenius-Weibull model is used to model the transformer failure rate. The specific formula is as follows:

[0098]

[0099] Among them, t e The equivalent operating time is obtained by subtracting the initial lifetime from the RUL, where B is a constant of 15000, β is 5.9, and C is 1.76 × 10⁻⁶. -12 θ0 is the rated operating temperature of the transformer, taken as 110℃.

[0100] It is worth noting that this model not only uses t e and F AA (t) directly quantifies the impact of transformer insulation aging on the failure rate, and also indirectly quantifies the impact of ambient temperature and load rate on the transformer failure rate.

[0101] With the rapid development of the energy industry in various countries, the installed capacity of distributed photovoltaic power generation is increasing year by year, and the load factor of distribution transformers is becoming increasingly affected by meteorological conditions. Therefore, when predicting the load factor of distribution transformers, it is necessary to establish both a load model that takes into account the influence of meteorological factors and a photovoltaic output model.

[0102] Empirical Mode Decomposition (EMD) is an adaptive time-frequency analysis method. In load modeling, EMD decomposes the load time series into two parts: the daily basic load component and a random component related to meteorological factors. Artificial Neural Networks (ANNs) can effectively learn the relationship between complex inputs and outputs. Combining historical meteorological and load data, this paper trains an ANN to deeply explore the mapping relationship between meteorological data and the random load component. Finally, based on meteorological forecast data, the trained ANN is used to predict the random load component and synthesize it with the daily basic load component to obtain the final load forecast curve. The EMD-ANN-based load forecasting method can effectively account for the influence of meteorological factors in the load modeling process. Figure 3 As shown, Figure 3 A specific load forecasting framework was demonstrated.

[0103] Photovoltaic power output is mainly affected by solar irradiance and temperature. This paper considers the influence of solar irradiance and temperature on photovoltaic power output through a photovoltaic power output formula, as follows:

[0104]

[0105] In the formula: P PV The actual output of the photovoltaic power generation device; P N G represents the rated output of the photovoltaic power generation device; G is the actual solar irradiance; G N The solar irradiance under standard test conditions is taken as 1000 W / m²; α P T represents the power temperature coefficient of the photovoltaic power generation device, taken as -0.35% / ℃; C T0 is the temperature of the photovoltaic power generation device; T0 is the ambient temperature; Tb is the temperature of the photovoltaic power generation device. STC The battery temperature under standard test conditions is taken as 25℃.

[0106] In summary, this paper considers the impact of meteorological factors on load and photovoltaic output using EMD-ANN and photovoltaic output formulas respectively. Hourly load and photovoltaic output forecast data are obtained by combining the generated hourly meteorological forecast data, and then synthesized to obtain the load rate forecast data of the distribution transformer, which is used for subsequent insulation aging risk assessment.

[0107] This invention primarily considers insulation life loss, underload loss, and curtailment loss when conducting a comprehensive risk assessment of distribution transformers. The life loss of the distribution transformer, S... I The following formula is used to measure the risk of insulation aging in distribution transformers:

[0108]

[0109] In the formula: L0 is the rated service life of the transformer, which is 180,000 hours according to IEEE C57.91; c trans This is the purchase price of the transformer.

[0110] After a transformer in a power distribution system is shut down, it cannot be put back into operation until emergency repairs or maintenance are completed. Without considering load transfer, this will result in load shedding losses. The specific calculation formula is as follows:

[0111]

[0112] In the formula: P D (i) represents the load at time i, and c loss The load shedding penalty price is t, where t is the time of transformer failure and τ is the transformer failure repair time.

[0113] With the accelerated construction of new power distribution systems and the continuous increase in installed capacity of new energy sources such as photovoltaics, it is necessary to consider curtailment losses when conducting risk assessments. The specific calculation formula is as follows:

[0114]

[0115] In the formula: P PVG (i) represents the photovoltaic output at time i, and c PV The price is the penalty for abandoning light.

[0116] The above three types of risks are directly related to the potential economic losses of distribution transformers. Based on the concept of risk, the specific formula for calculating the comprehensive risk loss of distribution transformers is as follows:

[0117] R isk =S I +p outage (t e |θ hst (S) LS +S PV (12)

[0118] The typical day method is a commonly used method for long-term risk assessment. It selects representative data to construct typical days and samples based on the probability of different typical days to generate the operating environment of future equipment.

[0119] This method significantly reduces the complexity of data processing and computation while still providing relatively accurate reliability assessment results. Furthermore, the typical day method selects typical days within a certain period, maintaining high assessment accuracy and demonstrating good robustness even if the original data contains some prediction errors.

[0120] Since both meteorological and electrical factors exhibit significant seasonality, this paper divides the predicted operating scenarios of distribution transformers generated in Section 2 into four seasons. For each season, K-means clustering is used to obtain typical predicted scenarios, and Latin hypercube sampling (LHS) is used to sample these typical scenarios. Risk indicators are then calculated based on the samples. The risk assessment process based on the generated hourly prediction curves of meteorological and electrical factors is as follows: Figure 4 As shown.

[0121] This paper conducts a risk assessment of distribution transformers in Shanghai using historical load data from 2022, historical meteorological data from 2006 to 2022, and meteorological forecast data for 2023 from the Washington CASI project, which has similar climate conditions to Shanghai. The meteorological data used in this paper are all from NASA Langley Research Center's Global Energy Forecasting Project.

[0122] like Figure 5 As shown, in addition to the load, distributed photovoltaic power is also connected to the distribution transformer.

[0123] The distribution transformer in the example is model S11-100 / 10±2×2.5%, with a capacity of 100kVA. The transformer parameters are shown in Table 1. The average fault repair time is taken as four hours, the rated life is taken as 180,000 hours, and the initial remaining insulation life is set at 87,600 hours, approximately half of the rated life. The curtailment penalty and load abandonment penalty are 708 yuan / MW·h and 4936 yuan / MW·h, respectively.

[0124] Table 1 Parameters of Distribution Transformers

[0125]

[0126] The historical load and temperature curves for 2022 are as follows: Figure 6 As shown, the average annual load factor is 0.62.

[0127] Taking temperature forecasting as an example, the CASI project provides forecasts for six GCMs and a six-model integrated Full Ensemble Average (FEA) model, along with daily average forecast data for a total of 21 scenarios across three carbon emission scenarios: high, medium, and low. Based on the MAE of the 2022 CASI project's Washington, D.C. area forecasts, the forecast scenario with the highest accuracy was selected. Table 2 shows a comparison of the MAE of the 2022 forecasts for different models and scenarios. It can be seen that the MAE of the FEA model, which integrates the forecasts of the six models, is lower than that of the other models, decreasing by approximately 32.4%. Therefore, the FEA forecast data under the medium emission scenario with the lowest MAE was selected to generate hourly meteorological data.

[0128] Table 2 Comparison of 2022 MAE Prediction Results for Different Models and Different Scenarios

[0129]

[0130] Since the annual average temperature difference between Shanghai and Washington in 2022 was 4.17℃, the predicted daily average temperature for Shanghai in 2023 was corrected by adding this difference to the predicted temperature for Washington in the medium-emission scenario under the FEA.

[0131] This paper uses historical meteorological data of Shanghai from 2006 to 2022, with 80% of the historical data used as the training set and 20% as the test set, to train a Radio Frequency (RF) algorithm with 100 trees to explore the relationship between daily average temperature and hourly average temperature. By inputting the predicted daily average temperature for Shanghai in 2023 into the trained RF algorithm, the predicted hourly average temperature for Shanghai in 2023 can be obtained.

[0132] To verify the accuracy of the hourly temperature forecast data for Shanghai in 2023 generated in this paper, the actual hourly temperature data for Shanghai in 2023 was used as the baseline. The MSE, RMSE, and MAE of the following data types were calculated: historical data from 2022, 2023 forecast data based on Long Short-Term Memory (LSTM), 2023 forecast data based on Time Series Decomposition (TSD), and 2023 forecast data based on RF-GCMs. The specific results are shown in Table 3. It can be seen that the LSTM forecast data has the lowest accuracy. This is because the temperature data contains complex nonlinear patterns and long-term trends. Directly using LSTM for forecasting does not fully model these components, resulting in its inability to effectively capture these complex patterns. Furthermore, LSTM heavily relies on high-quality data; otherwise, it is prone to overfitting. TSD deconstructs time series data into components such as trend, seasonality, and randomness, effectively capturing the potential relationships in the data. Its forecast MAE is 86% of that of historical data, indicating relatively high accuracy. The 2023 forecast data generated by RF-GCMs has the highest accuracy, with MAE only 74% of the historical data. This is because there are certain variations in factors such as greenhouse gas concentrations, ocean circulation (such as El Niño), and ice cover between adjacent years, and the hourly forecast data generated based on RF-GCMs takes into account the influence of these factors, resulting in higher accuracy.

[0133] Table 3 Comparison of Hourly Historical Data and Forecast Data

[0134]

[0135] Based on the hourly forecast meteorological data for 2023 generated above, the EMD-ANN and photovoltaic output formula in Section 2.2 are used to consider the correlation between meteorological and electrical elements in the operation scenario of distribution transformers, and to predict the load and photovoltaic output.

[0136] Meteorological elements such as temperature and light intensity exhibit distinct seasonal characteristics, as do loads and photovoltaic power, which are also significantly affected by meteorological elements. (Reference) Figure 7 As shown, Figure 7 The image shows hourly curves of predicted temperature and predicted load for a certain location in Shanghai in 2023.

[0137] Therefore, when using k-means clustering for scenario reduction, the year is divided into four seasons, and four typical daily operating scenarios for distribution transformers are generated in each season. These 16 typical days are used as the basis for transformer fault risk assessment. Figure 8 shows the prediction results of typical daily operating scenarios for distribution transformers in a certain area of ​​Shanghai in the summer of 2023, mainly including predicted data on temperature, load, and photovoltaic output that are directly related to the risk of distribution transformers. The rated output of the photovoltaic unit is taken as 10kW.

[0138] Among them, scenarios 1-4 represent the four typical daily prediction scenarios for summer, with probabilities of 26.1%, 29.4%, 22.8%, and 21.7%, respectively.

[0139] To verify the effectiveness of the risk assessment method proposed in this paper, the risk assessment results based on the 2023 forecast data are compared with the actual risk assessment results of the distribution transformers obtained from the historical meteorological data of Shanghai in 2022. The actual risk assessment results of the distribution transformers obtained from the actual meteorological data of Shanghai in 2023 are used as the true values ​​to verify its effectiveness.

[0140] Taking the insulation life loss of a distribution transformer as an example, Figure 9 The presentation showcases the Residual Insulation Life (RUL) assessment results for distribution transformers without photovoltaic (PV) grid connection, using actual data from 2023, historical data from 2022, and predicted data for 2023. The overall trend shows a rapid decrease in insulation life throughout the year, followed by a slower decrease at the beginning and end. Figure 8 illustrates that the middle of the year coincides with summer, a period of peak temperatures and loads, leading to significant insulation aging and loss in transformers. However, the overall assessment results for the next year show that the actual data estimates an annual insulation life loss of 1220 hours, while the historical data estimates 1777 hours, a difference of 45.6%. The predicted data, however, estimates 1284 hours, a difference of only 5.2%, demonstrating a significant improvement in accuracy. This is because 2022 saw frequent extreme high temperatures, reaching a maximum of 39.4°C, while 2023's maximum temperature is only 35.1°C. The high temperatures also impacted the load, resulting in very similar insulation life loss patterns in both cases. Using historical data directly for risk assessment can be affected by year-to-year climate differences, resulting in significant errors. Using forecast data for risk assessment can significantly reduce this impact and achieve higher accuracy, but it is still affected by forecast errors. Therefore, the typical day method is needed to reduce the impact of raw data errors.

[0141] Table 4 shows the annual insulation life loss and annual failure rate calculated using historical data, historical typical days, predicted data, predicted typical days, and actual data. The insulation life loss and annual failure rate calculated using the typical day method are both within a 95% confidence interval.

[0142] Table 4 Comparison of Historical Data and Forecast Data

[0143]

[0144] As shown in Table 4, the 95% confidence interval of the predicted typical day assessment results can well contain the assessment results of the actual data. This is because the typical day method simulates the possible operating scenarios of the distribution transformer in the coming year by sampling typical days, and considers the situation more comprehensively. In addition, since the typical day method selects typical days over a period of time, it reduces the impact of extreme weather on the annual risk assessment results in certain periods. Combined with equations (1)-(4), it can be seen that the insulation aging condition has an exponential relationship with temperature. Therefore, using the typical day method will result in less average insulation life loss and average annual failure rate than the results obtained by directly using the original data.

[0145] Table 5. Risk Assessment Results for Predicting Typical Daily Data

[0146]

[0147] Table 5 presents the annual risk assessment results obtained by using predicted typical day data for distribution transformers not connected to photovoltaic systems, and compares them with the risk assessment results obtained using actual data. The table shows that the risk assessment results obtained using predicted typical day data still have high accuracy. Furthermore, annual insulation loss accounts for 13.9% of the annual comprehensive risk loss; therefore, it is necessary to consider the risk of insulation aging when conducting risk assessments of distribution transformers.

[0148] With the increasing penetration rate of photovoltaic (PV) power, it is necessary to consider the impact of PV integration when conducting risk assessments of distribution transformers. This paper sets up scenarios with different proportions of PV installed capacity to transformer capacity and uses the proposed risk assessment method based on predicted typical days to assess the risk of distribution transformers. The risk assessment results are as follows. Figure 10 As shown.

[0149] As the proportion of photovoltaic (PV) capacity continues to increase, the overall risk of distribution transformers is decreasing. This is because, on the one hand, PV systems bear a portion of the load, reducing the amount of electricity that needs to be transmitted through distribution transformers, thus reducing the risk of insulation aging in distribution transformers. On the other hand, the decrease in the load rate of distribution transformers also leads to a decrease in their failure rate, further reducing the risk of load shedding. Although the risk of curtailment increases slightly, PV output is limited by sunlight intensity, resulting in a relatively low annual PV power generation, and the penalty for curtailment is lower than the penalty for load shedding. Therefore, the annual curtailment penalty accounts for a small proportion of the overall risk and has almost no impact on the reduction of the overall risk proportion. However, at the same time, the rate of reduction in overall risk, load shedding risk, insulation aging risk, and failure rate is slowing down. This is because PV units do not operate at night, so the overall risk of distribution transformers at night is not affected.

[0150] Therefore, the risk assessment method proposed in this paper can adapt well to the current situation of continuously increasing photovoltaic capacity, and shows that increasing the photovoltaic capacity connected to distribution transformers can slow down the insulation aging of distribution transformers and reduce their overall risk. However, when the proportion of photovoltaic capacity is high, the improvement of insulation aging and overall risk of distribution transformers by increasing photovoltaic installations will no longer be significant.

[0151] Optionally, based on local historical meteorological data, local meteorological measurement data, and typical daily average meteorological forecast data of GCMs, local hourly meteorological forecast data are generated, including:

[0152] Based on local historical meteorological data, a random forest model (RF) was determined for the local daily average and hourly meteorological curves.

[0153] Based on local meteorological measurement data, the typical daily average meteorological forecast data of GCMs is corrected to obtain the local daily average meteorological forecast data of GCMs.

[0154] Local hourly weather forecast data are generated based on the random forest model (RF) and GCMs (Gross Meteorological Contexts) of local daily and hourly weather curves.

[0155] Optionally, based on local meteorological measurement data, the typical daily average meteorological forecast data of GCMs are corrected to obtain the local daily average meteorological forecast data of GCMs, including:

[0156] The accuracy of typical daily average meteorological forecast data of GCMs is measured by the root mean square error (RMSE), mean absolute error (MAE), and symmetric mean absolute percentage error (SMAPE). The specific calculation formulas are as follows:

[0157]

[0158] Among them, y iThis is the daily average actual meteorological data. represents the daily average weather forecast data, and n represents the sample size.

[0159] Optionally, an insulation aging acceleration factor is determined, and based on the insulation aging acceleration factor, the transformer dynamic failure rate is determined, including:

[0160] Insulation aging acceleration factor F is adopted AA The insulation aging rate of a transformer is described by (t), and the specific calculation formula is as follows:

[0161]

[0162] Where, θ H (t) represents the winding hot spot temperature, θ H The formula for calculating (t) is as follows:

[0163]

[0164] Where γ is the ratio of load loss at rated load to loss at zero load, and θ A i(t) represents the ambient temperature, and i(t) represents the transformer load at time t. r It is the rated load, Δθ H.R Δθ is the temperature rise of the winding hot spot under rated load. TO.R The oil temperature is at rated load. m and n are transformer parameters determined by looking up a table based on the transformer's cooling system. For most distribution oil-immersed transformers, m and n are both 0.8.

[0165] The insulation loss of the transformer is equivalent to the reduction in the remaining service life (RUL) under standard operating conditions by using an insulation aging acceleration factor. The specific recursive relationship is as follows:

[0166]

[0167] Based on the insulation aging acceleration factor, the transformer failure rate is modeled, and the specific formula is as follows:

[0168]

[0169] Among them, t e The equivalent operating time is obtained by subtracting the initial lifetime from the RUL, where B is a constant of 15000, β is 5.9, and C is 1.76 × 10⁻⁶. -12 θ0 is the rated operating temperature of the transformer, taken as 110℃.

[0170] Optionally, based on local hourly meteorological forecast data and the rated output of the photovoltaic power generation device, the actual output of the photovoltaic power generation device is calculated, including:

[0171] The influence of solar irradiance and temperature on photovoltaic output is considered through a photovoltaic output formula, as follows:

[0172]

[0173] In the formula: P PV The actual output of the photovoltaic power generation device; P N The rated output of the photovoltaic power generation device; G

[0174] G represents the actual solar irradiance. N The solar irradiance under standard test conditions is taken as 1000 W / m. 2 ;α P T represents the power temperature coefficient of the photovoltaic power generation device, taken as -0.35% / ℃; C T0 is the temperature of the photovoltaic power generation device; T0 is the ambient temperature; Tb is the temperature of the photovoltaic power generation device. STC The battery temperature under standard test conditions is taken as 25℃.

[0175] Optionally, calculate the load shedding loss, determine the life loss of the distribution transformer based on the insulation aging acceleration factor, and calculate the curtailment loss based on the actual output of the photovoltaic power generation unit, including:

[0176] According to the insulation aging acceleration factor F AA (t), determine the life loss S of the distribution transformer. I The specific formula is as follows:

[0177]

[0178] In the formula: L0 is the rated service life of the transformer, which is 180,000 hours according to IEEE C57.91; c trans

[0179] This refers to the purchase price of the transformer;

[0180] The formula for calculating load shedding loss is shown below:

[0181]

[0182] In the formula: P D (i) represents the load at time i, and c loss The load shedding penalty price is given, t is the time of transformer failure, and τ is the transformer failure repair time.

[0183] Based on the actual output of the photovoltaic power generation device, the curtailment loss is calculated using the following formula:

[0184]

[0185] In the formula: P PVG(i) represents the photovoltaic output at time i, and c PV The price is the penalty for abandoning light.

[0186] Optionally, based on the transformer dynamic failure rate, distribution transformer life loss, load shedding loss, and curtailment loss, the comprehensive risk loss of the distribution transformer is determined, including:

[0187] The specific formula for calculating the comprehensive risk loss of distribution transformers is as follows:

[0188] R isk =S I +p outage (t e |θ hst (S) LS +S PV (12).

[0189] Therefore, the Random Forest (RF) model was used to refine the data granularity of Global Climate Models (GCMs) to obtain hourly meteorological forecast data. Considering the correlation between meteorology, load, and photovoltaic output, the load rate of distribution transformers was predicted. The typical day method was used to comprehensively assess the risks of accelerated insulation aging, load loss, and curtailment, and finally, the comprehensive risk assessment results of distribution transformers were obtained.

[0190] According to another aspect of the present invention, a comprehensive risk assessment system 1100 for distribution transformers based on RF-GCMs is also provided, with reference to... Figure 11 As shown, the system 1100 includes:

[0191] The weather forecast data generation module 1110 is used to generate local hourly weather forecast data based on local historical weather data, local weather measurement data, and typical daily average weather forecast data of GCMs.

[0192] The dynamic failure rate determination module 1120 is used to determine the insulation aging acceleration factor and, based on the insulation aging acceleration factor, determine the transformer dynamic failure rate.

[0193] The photovoltaic actual output calculation module 1130 is used to calculate the actual output of the photovoltaic power generation device based on local hourly meteorological forecast data and the rated output of the photovoltaic power generation device.

[0194] The transformer life loss determination module 1140 is used to calculate load shedding loss, determine the life loss of distribution transformers based on insulation aging acceleration factor, and calculate curtailment loss based on actual output of photovoltaic power generation device.

[0195] The Comprehensive Risk Loss Module 1150 is used to determine the comprehensive risk loss of distribution transformers based on the transformer dynamic failure rate, distribution transformer life loss, load shedding loss, and curtailment loss.

[0196] The integrated risk assessment system 1100 for distribution transformers based on RF-GCMs in one embodiment of the present invention corresponds to the integrated risk assessment method 100 for distribution transformers based on RF-GCMs in another embodiment of the present invention, and will not be described again here.

[0197] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0198] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0199] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0200] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0201] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0202] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for comprehensive risk assessment of distribution transformers based on RF-GCMs, characterized in that, The method comprises the following steps: generating local hourly weather prediction data according to local historical weather data, local weather measurement data and GCMs typical climate daily average weather prediction data; determining an insulation aging acceleration factor, and determining a transformer dynamic failure rate based on the insulation aging acceleration factor; calculating the actual output of the photovoltaic output device according to the local hourly weather prediction data and the rated output of the photovoltaic output device; calculating the load shedding loss, determining the power distribution transformer life loss according to the insulation aging acceleration factor, and calculating the light abandonment loss based on the actual output of the photovoltaic output device; determining the comprehensive risk loss of the power distribution transformer based on the transformer dynamic failure rate, the power distribution transformer life loss, the load shedding loss and the light abandonment loss.

2. The method of claim 1, wherein, The method comprises the following steps: determining a random forest model RF of local daily average and hourly weather curve according to local historical weather data; correcting the GCMs typical climate daily average weather prediction data to obtain GCMs local daily average weather prediction data according to local weather measurement data; generating local hourly weather prediction data according to the random forest model RF of local daily average and hourly weather curve and the GCMs local daily average weather prediction data.

3. The method of claim 2, wherein, The method comprises the following steps: measuring the accuracy of the GCMs typical climate daily average weather prediction data based on root mean square error RMSE, mean absolute error MAE and symmetric average absolute percentage error SMAPE, and the specific calculation formulas are as follows: wherein y i is the daily average meteorological actual data, is the daily average meteorological prediction data, and n is the sample number.

4. The method of claim 1, wherein, determining an insulation aging acceleration factor, and determining a transformer dynamic failure rate based on the insulation aging acceleration factor, which comprises the following steps: Insulation aging acceleration factor F is adopted AA The insulation aging rate of a transformer is described by (t), and the specific calculation formula is as follows: where θ H (t) is the winding hot-spot temperature, θ H (t) is calculated as follows: where γ is the ratio of the load loss at rated load to the loss at zero load, θ A (t) is the ambient temperature, i(t) is the transformer load at time t, i r is the rated load, Δθ H.R is the winding hot-spot temperature rise at rated load, Δθ TO.R is the oil temperature at rated load, and m and n are transformer parameters determined from a look-up table according to the cooling system of the transformer, and m and n are both 0.8 for most distribution oil-immersed transformers. equivalent the insulation loss of the transformer to the reduction of the remaining useful life RUL under standard working conditions through the insulation aging acceleration factor, and the specific recursive relationship is as follows: modeling the transformer failure rate based on the insulation aging acceleration factor, and the specific formula is as follows: Where, t e is the equivalent operating time, B is a constant of 15000, β is 5.9, C is 1.76x10 -12 , and θ0 is the rated operating temperature of the transformer, which is 110℃.

5. The method of claim 1, wherein, calculating the actual output of the photovoltaic output device according to the local hourly weather prediction data and the rated output of the photovoltaic output device, which comprises the following steps: considering the influence of solar irradiance and temperature on photovoltaic output through the photovoltaic output formula, and the specific formula is as follows: where: P PV is the actual power output of the photovoltaic power plant; P N is the rated power output of the photovoltaic power plant; G is the actual solar irradiance; G N is the solar irradiance under standard test conditions, taken as 1000 W / m 2 ; a P is the power temperature coefficient of the photovoltaic power plant, taken as -0.35% / °C; T C is the temperature of the photovoltaic power plant; T0 is the ambient temperature; T STC is the cell temperature under standard test conditions, taken as 25°C.

6. The method of claim 1, wherein, calculating the load shedding loss, determining the power distribution transformer life loss according to the insulation aging acceleration factor, and calculating the light abandonment loss based on the actual output of the photovoltaic output device, which comprises the following steps: According to the insulation aging acceleration factor F AA (t), determine the distribution transformer life loss S I , as follows: wherein: L0 is the rated service life of the transformer, 180 000 h according to IEEE C57.91; c trans is the purchase price of the transformer; calculating the load shedding loss, and the formula is as follows: where: P D (i) is the load at time i, c loss is the penalty price for cutting load, t is the time of transformer failure, τ is the repair time of transformer failure; calculating the light abandonment loss based on the actual output of the photovoltaic output device, and the formula is as follows: where: P PVG (i) is the photovoltaic output at time i, c PV is the curtailment price.

7. The method of claim 1, wherein, determining the comprehensive risk loss of the power distribution transformer based on the transformer dynamic failure rate, the power distribution transformer life loss, the load shedding loss and the light abandonment loss, which comprises the following steps: The specific calculation formula of the comprehensive risk loss of the power distribution transformer is as follows: R isk = S I + p outage (t e | θ hst )(S LS + S PV ) (12).

8. An RF-GCMs based power distribution transformer integrated risk assessment system, characterized in that, The method comprises the following steps: generating weather prediction data module for generating local hourly weather prediction data according to local historical weather data, local weather measurement data and GCMs typical climate daily average weather prediction data; A dynamic failure rate determining module is configured to determine an insulation aging acceleration factor, and determine a dynamic failure rate of the transformer based on the insulation aging acceleration factor; A photovoltaic actual output calculating module is configured to calculate an actual output of the photovoltaic output device according to local hourly meteorological prediction data and a rated output of the photovoltaic output device; A transformer life loss determining module is configured to calculate a load shedding loss, determine a distribution transformer life loss according to the insulation aging acceleration factor, and calculate a light abandonment loss based on the actual output of the photovoltaic output device; A comprehensive risk loss determining module is configured to determine a comprehensive risk loss of the distribution transformer based on the dynamic failure rate of the transformer, the distribution transformer life loss, the load shedding loss, and the light abandonment loss.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The program, when executed by a processor, implements the steps of the method of any one of claims 1-4.

10. An electronic device, comprising: Comprise: The computer readable storage medium of claim 9; And One or more processors for executing the program in the computer readable storage medium.

Images