Monitoring of transformers in the power grid

Abstract

A method for estimating the hysteretic apparent power over time of at least one transformer device in a power grid within a geographical area, comprising: providing training data including hysteretic apparent power data over time of a plurality of transformer devices within one or more geographical areas, hysteretic energy consumption data of the power grid within each geographical area, and hysteretic ambient temperature data for each geographical area; initializing the weights of an untrained neural network including an input layer, a plurality of hidden layers, and an output layer, and iteratively updating the weights using the training data to minimize a loss function, thereby generating a trained neural network configured to output estimated apparent power of a transformer device over a time period within a geographical area from input energy consumption data and input ambient temperature data for the geographical area over the time period; and inputting energy consumption data and ambient temperature data for a first time period within a geographical area into the trained neural network to estimate the apparent power of at least one transformer device within a geographical area over a first time period.

Description

[Technical Field]

[0001] Technical field This disclosure relates to the field of monitoring distribution transformers in power grids, and more particularly to the estimation of the remaining service life of such transformers and the estimation of parameters for use in such estimation. [Background technology]

[0002] background A transformer is a static electromachine that increases and decreases voltage while transferring electrical energy from one circuit to another. It operates on the principle of mutual inductance between the transformer's windings to enable the transfer of electrical energy between circuits. A distribution transformer is a type of transformer that provides the final transformation in a power distribution system, reducing the voltage used in the distribution lines to the level used by consumers. A distribution transformer is typically a three-phase unit with two secondary windings that have a center tap to provide single-phase or three-phase service. The primary winding is usually connected to the distribution voltage at a medium voltage level of 10kV to 36kV. The secondary winding is connected to the consumer voltage, typically a low voltage level of 240V to 600V. Distribution transformers are an extremely important component in a power distribution system.

[0003] Distribution transformers can be further classified by the type of electrical insulation and thermal cooling mechanism they use. They can be dry-type or oil-filled. This disclosure relates to oil-filled transformers. These transformers typically include a tank filled with oil that encloses the transformer's metal core together with the coil windings. The oil removes heat from the transformer and ensures the dielectric strength of the insulating material.

[0004] Thermal aging and insulation degradation affect the lifespan of a transformer. The hot spot temperature of a transformer is the highest temperature within the transformer windings that can degrade the insulating material and lead to dielectric breakdown. Therefore, the hot spot temperature is a crucial parameter in defining the thermal conditions and overload capacity of a transformer.

[0005] In reality, the hot-spot temperature of a transformer depends on the load factor, which is expressed as the ratio of the ambient temperature and the apparent power to the base load of the transformer. Measuring the hot-spot temperature of a transformer is very important for asset managers of power grid operators, such as utility companies. At a high level, the lifespan of a transformer generally depends on the service conditions in the environment where the transformer operates and the reliability of other components of the transformer. Non-exhaustive examples that can affect the lifespan of a transformer include the following. · The type of insulating material used (e.g., the type and composition of oil) · The operating temperature of the transformer · The load of the transformer · The oil and hot-spot temperature · The environment of the transformer, such as moisture, dust, and ambient temperature · The manufacturing quality of the insulating material · The historical load of the transformer

[0006] There are many challenges faced by power grid operators during the operation and management of the power grid. One of the challenges is the prioritization of distribution transformer replacement.

[0007] For example, the introduction of nonlinear loads such as rectifiers, discharge lighting, or saturated electromachines into the power grid results in the inflow of non-sinusoidal current and voltage harmonics into the power grid. Harmonics have only begun to appear in the last 30 years. Voltage harmonic or current harmonic frequencies are integer multiples of the fundamental frequency. Total harmonic distortion is the magnitude of the harmonics present in an electrical signal. It can be calculated using the ratio of the root mean square (RMS) amplitude of higher harmonic frequencies to the RMS amplitude of the fundamental frequency. When exposed to non-sinusoidal conditions due to harmonic distortion, transformers designed for a pure sinusoidal supply of the fundamental frequency experience increased iron losses, copper losses, and dielectric losses, leading to increased temperature rises. The rate of temperature rise above rated conditions depends on the increased losses due to harmonic distortion. The increased temperature rise due to harmonics leads to more rapid deterioration of insulation, resulting in a shorter lifespan. The heat generated by these losses also affects the quality of the oil in these transformers, thereby increasing the hot spot temperature mentioned above. This significantly accelerates the aging of transformers, potentially leading to premature failure or replacement.

[0008] On the other hand, modern transformers can generally cope with the detrimental effects of these losses. Transformers manufactured in the mid-1960s and earlier were not designed to accommodate harmonic loads. The increasing use of nonlinear devices in today's power grids leads to more rapid degradation of insulating materials within transformers. In fact, partly due to the increase in harmonics in the power grid, the failure rates of these older transformers have increased to a worrying degree.

[0009] To further exacerbate this problem, most power grids worldwide rely on distribution transformers installed in the 1950s and 1960s, which are already nearing the end of their lifespan. Due to limited resources, operators cannot replace all necessary transformers at once, forcing them to continue operating transformers beyond their rated service life (i.e., the time the transformer manufacturer intends for the transformer to be used with an acceptable low risk of failure). Knowing the optimal time to replace assets is essential for operators. If assets are not replaced in a timely manner, it can lead to defects, higher costs, and customer complaints. On the other hand, replacing assets too early is unsustainable and uneconomical. Opportunities to achieve long-term productivity improvements and enhance the sustainability of such assets would be beneficial for power grid operators. Therefore, understanding the current state of a transformer (i.e., how much service life it has left before it fails or the risk of failure becomes unacceptably high) allows operators to cost-effectively prioritize their replacement and maintenance schedules, which greatly facilitates operators in managing their operating and capital costs. Existing methods for determining the state of transformers are applied only sporadically over long time intervals, which can be unsafe and do not provide asset managers with sufficient information.

[0010] For example, one known method uses a process known as dissolved gas analysis (DGA). DGA provides detection of early fault conditions that can lead to major failure modes of transformers. Dissolved gas analysis is typically performed by manually taking a sample of transformer oil and sending it to a laboratory. The sample is then injected into a gas chromatograph that separates various dissolved gases and identifies their concentrations. The results are then analyzed to determine whether any fault conditions are present. The gases present in transformer oil are formed by the decomposition of the transformer insulation. The most common gases found in transformer oil are ethane, ethylene, acetylene, hydrogen, methane, and other greenhouse gases (GHGs). These gases are formed by the decomposition of the cellulose insulation within the transformer. Analysis of a transformer oil sample provides useful information for determining the health of the transformer, which the DSO can then use to determine the schedule for further maintenance activities and transformer replacement.

[0011] The gases in transformer oil can convey many details about the health of the transformer. For example, a high concentration of hydrogen gas may indicate that the transformer is under high electrical stress. High concentrations of methane and ethane may indicate that the transformer is under high thermal stress. Thermal stress arises from temperature fluctuations and the different thermal expansion coefficients of materials. Thermal stress can cause cracks in the insulation, thereby reducing its dielectric strength. Thermal stress can also cause expansion or collapse of the insulation, creating air pockets within the insulation and increasing the risk of electrical failure. Electrical stress arises from the application of high voltage to the insulation. These stresses can cause dielectric breakdown of the insulation, allowing current to flow through it.

[0012] However, the problem with the DGA approach is that it is an intrusive method that requires maintenance technicians to actively open transformer housings to take oil samples, exposing themselves and the public to the opened transformers. Furthermore, doing so would require employing hundreds of technicians to perform manual sampling within the power grid, and given the costs that would arise from the logistics of regularly testing a large number of samples, it would not be feasible to do it continuously, daily, or weekly for all transformers in the power grid.

[0013] As a result, many power grid operators either simply avoid regular monitoring and maintenance of distribution transformers, or minimize the frequency of such monitoring. In some extreme cases, distribution transformers may not be monitored at all, and if maintenance is performed, it may consist only of occasional substation cleaning, visual inspection, and DGA to check oil quality. However, the consequences of transformer failure can be catastrophic.

[0014] Therefore, asset managers of power grid operators will benefit from being able to remotely and regularly monitor the health of transformers. Furthermore, power grid operators are expected to ensure high availability and rapid restoration after power outages.

[0015] To attempt to standardize the management and maintenance of transformers globally, the Institute of Electrical and Electronics Engineers (IEEE) has developed several standards for determining transformer life. The standards relating to this disclosure are used to calculate life loss. Life loss is a metric that indicates how much of the transformer's intended life has already been used up, and this, therefore, indicates how much of the transformer's service life remains before replacement is required.

[0016] Specifically, lifetime loss is defined as the decrease in the ability of the transformer's insulating material to perform its function of providing electrical isolation. This is characterized by an increase in the dielectric loss of the insulating material, which leads to a corresponding increase in the transformer's operating temperature. Lifetime loss in mineral oil-filled transformers is formalized in the ANSI / IEEEC57.91 and IEC-354.91 standards.

[0017] According to the IEEE standard ANSI / IEEEC57.91, the normal life of a transformer is 50 years. This life corresponds to continuous operation at a design hotspot temperature of 110°C. In contrast, the IEC standard IEC-354.91 does not define total life, but it is usually equal to 30 years, depending on the aging rate determined by the hotspot temperature. The calculation of life loss is given by the IEEE load general rule, which assumes that insulation degradation can be modeled as a unit amount at a reference temperature of 110°C. Accelerated aging F AA The equation is given by Equation 1, which depends on the hot spot temperature of the transformer.

number

[0018] Equivalent aging of transformers (F EQA Equation (2) can be used to calculate ).

number

[0019] The remaining service life in hours can be calculated using Equation 3. Remaining service life = Normal insulation life - F EQA *Operating hours Equation 3

[0020] On the other hand, the IEC standard loading rules are mainly applicable to non-thermally improved paper, and the hot spot temperature is limited to 98 °C at an ambient temperature of 20 °C.

[0021] The aging rate given by the IEC model is referred to in Equation 4.

Number

[0022] In practice, the above-mentioned IEEE and IEC approaches for determining life loss and remaining service life generally have two stages. In the first stage, the ambient temperature and load factor are obtained as input parameters. In the second stage, transformer parameters and real-time electrical data (e.g., voltage, current, and / or power in the transformer) are obtained, and the life loss is calculated using the hot spot temperature calculated by running a thermal model. The IEC-354 loading rules for oil-immersed transformers shown above describe the most common thermal models used for calculating oil and hot spot temperatures. For example, the above-mentioned IEC-354 loading rules give the steady-state temperature relationship. The reached hot spot temperature of a transformer under any load is equal to the sum of the ambient temperature, the oil temperature rise above the ambient, and the hot spot temperature rise above the oil. Note that the oil temperature in this case is usually measured as the top oil temperature. This is because this is where the maintenance technician takes oil samples. This can be represented by Equation 5 below. θ H = θ A + Δθ TO + Δθ H Equation 5 Here, θ A is the average ambient temperature during the considered load cycle, in °C, θ TO is the top oil temperature, in °C.

[0023] The terms of Equation 5 are well known to those skilled in the art and are defined in the IEC-354 general load rules and the IEEEC57.91 standard for oil-filled transformers, for example, in Appendix G. Methods for measuring the physical parameters of the terms of Equation 5 and calculating those terms are also known to those skilled in the art. For example, the physical and structural parameters of the transformer being tested are identified by measuring the top oil temperature using a sample collection rod and referring to the manufacturer's data sheet and the metal nameplate attached to the transformer, on which this information is also recorded.

[0024] For ease of reference, one exemplary set of standard thermal models from the IEEE standard C57.91, which is widely used in transformers, is presented below.

[0025] Specifically, Appendix G of IEEE standard C57.91 describes a method for estimating the hottest spot temperature of a transformer, given ambient temperature and load profile. The transformer temperature model estimates the internal transformer temperature using an alternative temperature calculation method proposed in Appendix G of the IEEE General Rules. The transformer temperature model estimates the internal transformer temperature at 1-minute intervals using forward Euler integrals. As mentioned above, transformer aging results from the breakdown of dielectric strength in the transformer insulation, which is directly related to the temperature of the transformer windings. Therefore, the hottest spot temperature of the windings estimated by this model is necessary to determine the equivalent aging of the transformer. The formula in Appendix G of the IEEE General Rules takes into account the type of fluid, cooling mode, winding duct oil temperature rise, power loss due to harmonics, resistance and viscosity changes, ambient temperature, and load changes during load cycling. The transformer temperature model estimates the hottest spot temperature of the windings at each time interval using the IEEE General Rules outlined below. The overall model is based on the fluid flow conditions that occur within the transformer during transient conditions. The maximum spot temperature includes the following components, given by Equation 6. θ H =θ A +Δθ BO +Δθ H / WO +Δθ WO / BO formula 6 Here, θ AThis is the mean ambient temperature in °C during the load cycle under consideration. Δθ BO This is due to the rise of the lower fluid above the surrounding area, and the temperature is °C. Δθ WO / BO This is the temperature rise in °C of the oil at the winding hot spot, where the temperature rise exceeds that of the lower oil. Δθ H / WO This is the winding hotspot temperature rise, which exceeds the oil temperature adjacent to the hotspot location, in °C.

[0026] The overall model is based on three separate, widely used standard thermal models. (i) Average oil temperature calculation (ii) Calculation of average winding temperature (iii) Winding maximum temperature spot temperature calculation

[0027] These IEEE standard models precisely describe the heat transfer and fluid flow phenomena that occur in oil-filled transformers during transient loads.

[0028] For the first model, the average oil temperature of a given transformer can be calculated by evaluating equations 7-9.

number

number

number

number

[0029] For the second model, the average winding temperature of a given transformer can be calculated by evaluating equations 10-11.

number

number

[0030] For the third model, the temperature of the highest temperature spot (i.e., hot spot) on the transformer winding can be calculated by evaluating equations 12-13.

number

[0031] As described above, these IEEE standard formulas are well known to those skilled in the art and, when combined with physical manufacturer parameters of a given transformer, obtained from datasheets and / or nameplates attached to the transformer, allow for the estimation of the transformer winding hotspot temperature from simple temperature measurements.

[0032] European Patent No. 3061107 proposes a method for determining the remaining service life of an electrical device from measured electrical and physical factors (e.g., temperature and humidity) to define the current actual state of the transformer at a given time (e.g., what the accelerated aging rate is based on the current actual state). This can then be used to estimate how much the transformer may have aged based on existing factors, and to estimate its remaining life. Thus, if the current actual temperature and electrical measurements indicate that the transformer is currently under heavy load, the method of European Patent No. 3061107 returns the conclusion that the transformer is likely to have a short remaining life. Conversely, if the current actual measurements indicate that the transformer is not under heavy load, it is likely to have a longer remaining life and not need immediate replacement. However, the actual values ​​being measured now do not provide any indication of what kind of load the transformer has been subjected to in the past. Furthermore, in the case of older transformers from the 1950s and 1960s, this information cannot be determined from past data because such data has not been collected at all and therefore does not exist. [Overview of the Initiative] [Problems that the invention aims to solve]

[0033] Improved methods for monitoring transformers in the power grid are desired. [Means for solving the problem]

[0034] overview In general, this disclosure solves these and other problems by simulating historical transformer load data from historical ambient temperature data and historical energy consumption data from different geographical regions, and using this simulated historical data to calculate a more accurate remaining life of the transformer. The method of this disclosure is a significant improvement over the method of European Patent No. 3061107, which does not and cannot consider the historical load of the transformer when calculating the remaining life.

[0035] Specifically, the term hysteretic load refers to the apparent power of a transformer over a hysteretic period of time. This apparent power indicates the hysteretic temperature (e.g., hot spot temperature) over the time the transformer is operating, which therefore indicates the extent to which the transformer's original service life has been exhausted. Thus, this disclosure is directed toward three main aspects: estimation of the hysteretic apparent power of a transformer over time, estimation of the hysteretic temperature of a transformer over time over that period, and calculation of the remaining service life of a transformer using these estimations.

[0036] Historical apparent power over time is estimated by generating simulated data using trained neural networks to fill in any gaps in this unavailable data, which is a particularly significant problem for older transformers from the 1950s and 1960s. Specifically, for a given set of geographical regions, as long as at least some historical apparent power data can be obtained, it is possible to predict what the historical apparent power data will be like in these regions where real-world data is incomplete or missing.

[0037] More specifically, the neural network is trained with data including apparent power data for several transformer devices in different geographical regions, which is available, obtained, for example, from historical records or national archives; ambient temperature data for these regions over a period of time; and energy consumption data for the corresponding power grids within these regions. The inventors have found that ambient temperature data and energy consumption data for a region over a period of time provide a good indicator of what the apparent power of transformers within the region was during that period. This is advantageous because ambient temperature data and regional energy consumption data are collected by national archives and are therefore generally more widely available. Thus, the neural network can be trained to output this simulated apparent power data from these much more widely available inputs.

[0038] Next, this simulated apparent power data can be used to provide a much more accurate estimate of the remaining service life of a given transformer. This helps power grid operators better manage their power grid maintenance.

[0039] According to a first aspect, a method is provided for estimating the hysteretic apparent power over time of at least one transformer device in a power grid within a geographical area, comprising: providing training data including hysteretic apparent power data over time of a plurality of transformer devices in one or more geographical areas, hysteretic energy consumption data of the power grid in each geographical area, and hysteretic ambient temperature data of each geographical area; initializing the weights of an untrained neural network including an input layer, a plurality of hidden layers, and an output layer, and iteratively updating the weights using the training data to minimize a loss function, thereby generating a trained neural network configured to output estimated apparent power of a transformer device over a time period in a geographical area from input energy consumption data and input ambient temperature data of the geographical area over the time period; and inputting energy consumption data and ambient temperature data over a first time period in a geographical area into the trained neural network to estimate the apparent power of at least one transformer device in a geographical area over a first time period.

[0040] Optionally, the training data further includes voltage harmonic data from the plurality of transformer devices.

[0041] A second embodiment provides a method for estimating the historical temperature over time of a transformer device in a power grid within a geographical area, comprising: (i) estimating the historical apparent power of the transformer device over a first period of time using the method described above; and (ii) estimating the historical temperature of the transformer device over a first period of time using (i) the estimated historical apparent power over the first period of time and (ii) ambient temperature data for the first period of time within the geographical area.

[0042] Optionally, the temperature of the transformer device over a first time period includes the maximum temperature present in one or more coil windings of the transformer device.

[0043] Optionally, the maximum temperature present in one or more coil windings of a transformer device is calculated according to IEEE standard C57.91.

[0044] A third aspect provides a method for managing a power grid having a plurality of transformer devices in each geographical area, the method comprising: estimating the remaining service life of each transformer device by: estimating the hysteretic apparent power of the transformer device over a first time period using the method described above; estimating the hysteretic temperature of the transformer device over a first time period using the method described above; calculating the remaining service life of the transformer device by performing a lifetime loss calculation using the estimated hysteretic apparent power and estimated hysteretic temperature over the first time period, and subtracting the result of the lifetime loss calculation from a predetermined maximum service life of the transformer device; and generating a warning if the calculated remaining service life of any of the plurality of transformer devices falls below a predetermined threshold.

[0045] Optionally, the calculation of lifetime loss may include calculation of lifetime loss in accordance with ANSI / IEEEC57.91 and IEC-354.91 standards.

[0046] The calculation of the remaining life of a transformer device by optionally performing a life loss calculation further includes using real-time measurements of the apparent power and temperature of the transformer device.

[0047] Optionally, the real-time measurement of apparent power of the transformer device includes performing one or more real-time measurements of current, voltage, and / or power at the connection point of the transformer device to the power grid.

[0048] Optionally, real-time temperature measurement of the transformer device includes measuring the outer surface of the transformer device housing without opening the housing.

[0049] Optionally, this method includes continuously performing the real-time measurements.

[0050] Optionally, the predetermined maximum service life of the transformer device is 30 to 60 years.

[0051] According to a fourth aspect, a system for managing a power grid is provided, comprising a processor and a plurality of transformer devices in each geographical region, each transformer device being communicatively coupled to a server and configured to transmit data to the server indicating real-time apparent power measurements and real-time temperature measurements of the transformer device, wherein the processor is configured to input, for each transformer device, (i) received real-time data and (ii) historical power consumption data and historical ambient temperature data for the corresponding geographical region of the transformer device from a first time period collected before the real-time measurement, the trained neural network being configured to output an estimated historical apparent power of the transformer device over the first time period; and to calculate the remaining service life of the transformer device using the real-time apparent power measurement and historical apparent power estimation; and to generate a warning if the calculated remaining service life of the transformer device falls below a predetermined threshold.

[0052] Optionally, the processor is configured to perform the step at least once a day or at even smaller intervals, for example, once every 10 minutes or once every minute.

[0053] Optionally, the system includes a user interface for displaying the geographical regions and locations of multiple transformer devices on a map, and the user interface is configured to display warnings.

[0054] Optionally, at least one of the transformer devices includes an oil-filled transformer device.

[0055] A fifth aspect provides a method for training a neural network to output estimated apparent power over a time period for transformer devices within a geographical region from input energy consumption data and input ambient temperature data for the geographical region over the time period, comprising: providing training data including apparent power data over time for a plurality of transformer devices within one or more geographical regions, energy consumption data over time for power grids within each geographical region, and ambient temperature data over time for each geographical region; initializing the weights of an untrained neural network including an input layer, a plurality of hidden layers, and an output layer; and iteratively updating the weights using the training data to minimize a loss function, thereby generating a trained neural network configured to output estimated apparent power over a time period for transformer devices within a geographical region from input energy consumption data and input ambient temperature data for the geographical region over the time period.

[0056] Brief explanation of the drawing These and other embodiments will now be described with reference to the drawings. [Brief explanation of the drawing]

[0057] [Figure 1] An illustrative block diagram relating to this disclosure is shown below. [Modes for carrying out the invention]

[0058] Detailed explanation Figure 1 illustrates block diagram 100 relating to this disclosure. Block diagram 100 has a first section 101, a second section 102, and a third section 103.

[0059] The first section 101 relates to calculating winding hotspot temperature from real-time power quality data. In the example shown in Figure 1, this involves applying the three thermal models described above with respect to widely used IEEE standards, evaluating the spot temperature, and performing optimization functions for improving transformer physical parameters. The methods described herein for calculating hotspot temperature are merely illustrative and not intended to be limiting, and those skilled in the art will be aware of other methods for calculating and measuring hotspot temperature.

[0060] Section 102 of the second section concerns estimating the hotspot temperature of the hierarchical load of a transformer. As mentioned above, for many transformers, particularly those installed during the 1950s and 1960s, hierarchical transformer data is usually unavailable or largely incomplete. Therefore, a neural network is provided that receives energy consumption and ambient temperature along with date and time, and backcasts the load on the transformer over any given period of missing data. It is trained using real-time power quality data collected over any recorded period to predict the hierarchical load.

[0061] Section 103, the third section, concerns calculating remaining life using the well-known IEEE standard formulas (1) to (3) described above. However, other known formulas capable of calculating remaining life may also be used, and any specific formula used to calculate remaining life is intended to be illustrative and not limiting.

[0062] Section 101 Calculation of hotspot temperature using real-time power quality data As mentioned above, transformers are typically manufactured by manufacturers and their ratings are determined using several key operating and physical parameters, such as those listed below. Size (kVA) Year of manufacture Mass of windings in kg core mass in kg Weight of a transformer without fluid (kg) Tank mass in kg Types of transformers Cooling methods (ONON, ONAF, OFAF, ODAF) Type of cooling / insulating fluid (oil, silicone, or HTAC) 17 Winding material (copper or aluminum) loss Winding I2R loss W Winding eddy current loss W Core (unloaded) loss W If present, stray loss W Eddy current loss at winding hotspot location per I2R loss unit 1 Current A at rated load (LV) Mass and volume Total weight of the transformer (kg) Weight of transformer fluid in kg Weight of core and coil in kg Temperature (measured at rated load P_BASE_LOAD) Transformer base load kW Rated ambient temperature (°C) Average winding temperature at rated load (°C) Maximum winding spot temperature at rated load: °C Upper fluid temperature at rated load (°C) Lower fluid temperature at rated load: °C Aging formula parameters Standard insulation lifespan (approximately 20 years) others Winding time constant s Value per unit of winding height to the location of the hotspot: 1 Constant 1 in the exponential moving average filter for Theta_DAO Constant 1 in the exponential moving average filter for Theta_TDO

[0063] Typically, these parameters, recorded on a nameplate attached to the transformer by the original manufacturer and / or provided in the manufacturer's datasheet, can be used, along with real-time power quality measurements from the transformer, to calculate the theoretical power loss of each transformer in the power grid (e.g., lost as heat generated by real-time voltage and current harmonics of the transformer, affecting the transformer's hotspot temperature).

[0064] To obtain real-time power quality measurements, one or more sensor devices installed on each transformer in the power grid may be configured to collect the following real-time data: • Voltage and current harmonics up to, for example, the 15th multiple of the fundamental frequency. • Active power • Reactive power ·Voltage ·Current ·frequency Power factor

[0065] It is known that increased harmonic distortion can lead to a corresponding decrease in the expected lifespan of a transformer. Therefore, considering harmonics in calculations is beneficial for accurately determining the current state of electrical devices.

[0066] As mentioned above, nonlinear loads are increasing in most power grids worldwide, resulting in distorted voltage and current waveforms. This contributes to the degradation of the quality of electricity flowing through the power grid. It directly affects the operation of transformers, which are sources of energy loss during their operation. Types of power loss caused by harmonics include transformer losses, no-load losses, and load losses (P LL ), Ohm loss (P I2R ), stray loss (P OSL ), winding eddy current loss (P EC ) and others are examples.

[0067] No-load losses, also known as core losses, appear due to the time variability of the electromagnetic flux passing through the core, caused by hysteresis and eddy current histories within the magnetic material. They take frequency and maximum magnetic flux density into consideration. Load losses include ohmic losses, winding losses, and other stray losses. Ohmic losses increase with increasing winding DC resistance and load current. It also takes harmonic components into consideration. Skin and proximity effects of electromagnetic forces cover the windings, increasing eddy current losses.

[0068] The voltage induced within a conductor due to the linkage between the electromagnetic flux and the conductor generates eddy currents, increasing the temperature of the windings. Eddy current losses occurring within structural parts of the transformer other than the windings are considered under stray losses. In transformers subjected to high loads with harmonic currents, excessive losses can cause high temperatures in certain locations within the windings. This can lead to a serious shortening of the transformer's lifespan and can cause immediate damage and sometimes even fire. Reducing the maximum apparent power transmitted by a transformer is often referred to as underrating.

[0069] The known formulas for calculating harmonic losses are given by equations 14-17 below.

number

number

[0070] Thermal model Three thermal models from IEEE standards for calculating the three main temperatures of a transformer (Equations 6-13, described above in the background section) are intended to be used to estimate the hotspot temperature twice: firstly, based on theoretical hotspot temperatures derived from manufacturer parameters obtained from the datasheet and / or nameplate attached to the transformer; and secondly, based on actual measurements obtained by measuring the outer surface of the transformer housing using, for example, thermocouples or IR sensors, based on an estimate of the top oil temperature.

[0071] As described above, equations 6-13 take into account factors including the type of liquid, cooling mode, winding duct oil temperature rise, viscosity changes, ambient temperature, and load changes during load cycling. As will be understood, estimating the hotspot greatly contributes to the loss calculation, thereby making power losses due to harmonics, in some cases, the primary reason for the increase in hotspot temperature. In most devices, the temperature distribution is not uniform, so the parts operating at the highest temperature usually experience the most significant degradation. Therefore, in estimating transformer aging, it is common to consider the aging effect caused by the highest (highest temperature spot) temperature. The hotspot temperature for a given time can be calculated using losses calculated considering harmonics in the power grid and temperature calculations set by the IEEE method described above. It will be understood that other methods for estimating the hotspot temperature of a transformer are also well known to those skilled in the art.

[0072] optimizer The reason for making two separate estimates of the hotspot temperature is that it allows for adjustment of the transformer's operating parameters to account for degradation and / or deficiencies in the details provided by the manufacturer. That is, if the hotspot temperature estimate based on measurements differs significantly from that based on the theoretical manufacturer's value, the theoretical manufacturer's value is likely inaccurate and / or outdated and therefore needs to be adjusted to more closely match the actual measured temperature value. This optimization process is illustrated by the feedback loop 104.

[0073] Specifically, as mentioned above, transformers, like all other electrical equipment, are designed and manufactured to function within an electrical system at specified voltages, frequencies, loads, etc. These key operating parameters change due to excessive heat, fluctuating ambient temperatures, or cooling difficulties. Therefore, the precise parameter values ​​used as inputs to the IEEE thermal model must be recalculated periodically to compensate for these changes.

[0074] The parameters used in the temperature calculation method must be recalculated to reflect the current state of the transformer. Sequential Least Squares Programming (SLSQP) is used to optimize the parameter set by minimizing the set of objective functions. The difference between the calculated theoretical top oil temperature and the measured or estimated top oil temperature can be used as the objective function to optimize the transformer parameters according to the current operating state.

[0075] However, as mentioned above, sampling oil temperature would involve an invasive method.

[0076] Alternatively, as described above, temperature sensors such as thermocouple probes and / or IR sensors can be placed on the outer surface of the transformer housing, for example, on the outer surface of the tank housing the oil-filled transformer, in order to calculate the hotspot temperature of a given transformer. Specifically, the temperature of the tank surface can be easily measured using temperature sensors such as thermocouple probes without accessing the internal transformer. If the top oil temperature can be estimated using the tank surface temperature, the winding hotspot temperature model of the IEEE load general rules mentioned above can be applied more accurately than when only ambient temperature is used. This estimation can then be used to estimate the winding temperature using the IEEE load general rules model. The relationship between the top oil of the transformer and the tank surface is given below. θ TO =α·θ TS +β

[0077] This formula uses a slope α and offset β, which depend on the overall structure and specifications of the transformer (e.g., number of windings, oil tank size), to express the coil winding temperature θ. TO and tank surface θ TS This effectively defines a linear relationship with temperature. In other words, the higher the coil winding temperature, the higher the oil tank surface temperature. Precise values ​​of α and offset β can be determined in a controlled environment for each type of transformer in the power grid and later used to easily estimate the coil winding temperature in real time from only the temperature readings of the oil tank surface of each type of transformer in the power grid.

[0078] Second category 102 Calculation of hotspot temperature under hysteretic load As mentioned above, known methods do not, and cannot, consider historical load data (and any hotspot temperatures calculated from it) because this data is largely incomplete or completely unavailable in many geographical areas.

[0079] This hysteretic load profile is also known as the K load. Hysteretic load data for transformer lifespan is a largely incomplete dataset. Many of the transformers currently in operation were installed during the 1950s and 1960s. Utilities lacked the technology to collect and store the power quality data necessary for temperature calculations according to the current state of thermal models, such as IEEE standards for oil-filled transformers in the power grid.

[0080] Instead, this disclosure attempts to overcome this problem by generating simulated historical load profiles for geographical areas where this data is unavailable. To model the historical load profiles, a neural network model was trained with the ability to predict apparent power (K ​​load × constant base load) based on inputs: energy consumption, ambient temperature, and calendar features such as year, month, weekday / weekend, and time of day. The training data consists of three data sources: active and reactive power of transformers, energy consumption of the districts where the transformers are located, and ambient temperature of those areas. These data sources do not need to be available for all time periods or geographical areas dating back to the 1950s and 1960s; as long as at least some training data is available, the model can be trained to generalize to almost any time period and geographical area without issue. Therefore, a model trained from these data sources was used to predict historical apparent power based on historical ambient temperature and energy consumption.

[0081] Figure 2 illustrates block diagram 200 relating to this disclosure, illustrating the generation of simulated historical load data. In the first step, a training dataset 201 is provided, which includes arbitrary power quality data available for a given time period t, namely K load data, ambient temperature data, and energy consumption data. This data can be segmented based on geographical area. This is advantageous because ambient temperature data is typically recorded by the National Weather Service according to geographical areas down to the town, county, state, etc., similar to the areas where energy consumption data is recorded by government agencies. This segmentation of data according to geographical area also facilitates grouping transformers in the power grid into one or more regions based on where they are physically located.

[0082] Next, the weights and biases of the untrained neural network are initialized and iteratively updated during training to minimize the loss function, thereby generating the trained neural network 202.

[0083] It will be understood that any suitable artificial neural network (ANN) architecture may be used. For example, a suitable number of neuronal layers (e.g., number of input, output, and hidden layers), the number of neurons per layer, their interconnections, their activation function (e.g., ReLU function), which loss function and optimization method is used, and which learning rate, batch size, and regularization parameters are used to train the power grid. Those skilled in the art will further understand that any of these parameters can be adjusted during training to achieve an acceptable loss curve and convergence during training with a given training dataset. Exemplary architectures considered may include recurrent neural network (RNN) or long short-term memory (LSTM) approaches.

[0084] Once the power grid 202 is trained using the training dataset 201, it can output a predicted historical K-load profile 203 for a given time period, which is a good estimation simulation of the historical K-load 203 of any transformer located in the geographical area from which the input ambient temperature and energy consumption data was obtained, from the input historical data 204 for ambient temperature and energy consumption over that time period.

[0085] The output historical K-load data for a time period and geographical area can then be used to calculate the corresponding historical hotspot data for that time period and area, so that the historical hotspot data becomes a good estimate of the hotspot temperature over time for any transformer within that geographical area. The hotspot data can be calculated from the K-load data in combination with ambient temperature data for a given time period, for example, using equations 7-13. Those skilled in the art will recognize that other methods for calculating hotspots from apparent power (i.e., K-load data), such as any method described in the aforementioned IEEE standard or other standards, can also be used. By generating simulated load and hotspot data in this way, the problem of missing or incomplete historical datasets is overcome, thereby taking into account the contribution of historical loads to aging in the transformer. This is an improvement over methods such as those provided by European Patent No. 3061107, which do not and cannot take into account historical data, as this data is usually incomplete or absent.

[0086] Finally, using the estimated historical hotspot data, Equation 1 is used to calculate the remaining lifetime F for all time points from the historical data and, if present, real-time actual / measured data. AA The value can be calculated. Therefore, this provides a good understanding of how much a given transformer has actually aged over its service life, and thus a better indication of whether replacement is appropriate. [Examples]

[0087] example The inventors demonstrated the effectiveness of the above method as shown below.

[0088] Data was recorded from exemplary transformers (500 MVA and 230 / 110 kV). The transformers were in operation from 1960 to 2022, and at this point in their lifespan, the theoretical, or rated, hotspot temperature based on the manufacturer's information for these transformers should be 85°C. This theoretical hotspot estimate would typically indicate that these transformers are likely nearing the end of their lifespan and therefore should be replaced.

[0089] Figure 3 shows Chart 300 of estimated hotspot temperatures over the transformer's operating period (i.e., 1960–2022), generated using the method of the present disclosure, i.e., by a trained neural network generated from input energy consumption and ambient temperature data within the geographical area where the transformer is located.

[0090] The x-axis represents the total time the transformer was in operation, and the y-axis shows the hotspot temperature in degrees Celsius. It can be seen that the hotspot temperature has been increasing since 1960. However, the range of the hotspot temperature remains within 20°C to 35°C, which is significantly lower than the theoretical hotspot estimate based solely on manufacturer data. From this chart, it can be inferred that the upper oil in the transformer does not get very hot, and therefore the oil cooling is operating well.

[0091] Figure 4 shows Chart 400 of the estimated K-load (apparent power / base load) over time for the example transformer described above. As before, the transformer's load profile is presented over its operational period from 1960 to 202. As can be seen in Figure 4, the x-axis shows that the K-load increases over time, but not significantly.

[0092] Figure 5 shows Chart 500 of ambient temperature data over time for the geographical region where the exemplary transformer described above is located and for the period from 1960 to 2022. It can be inferred that the hotspot temperature increases in parallel with the load on the transformer, and that ambient temperature also contributes to its influence. The ambient temperature at the geographical location is relatively low, between 4 and 9°C, which actually facilitates the cooling of the transformer. Therefore, the hotspot temperature is found to be significantly lower than its manufacturer's rated value throughout its entire operating time. No significant temperature increases occurred in the transformer.

[0093] Figure 6 shows the average aging acceleration coefficient F over time for the exemplary transformer described above, calculated using the hotspot temperature data from Figure 3. AA The chart is shown below. For example, it is as follows: Regular lifespan = 50 years Operating hours = 61.9986 years Equivalent F AA =0.00188 Aging = Operating time × Equivalent F AA =0.11630 years Remaining life = normal life - aging = 49.88370 years

[0094] The remaining lifespan of the transformer was calculated to be approximately 49.8 years, in addition to its existing normal lifespan of 50 years. Therefore, if the transformer continues to operate under similar conditions as it has since 1960, it will likely have twice the service life it was originally intended to have. Thus, the transformer is in good condition, and the power grid operator can continue to operate it for a further period.

[0095] Therefore, the graph in Figure 6 is F AA The values ​​increase, but on average remain within the limits of approximately 0.002 and 0.004. Therefore, the example transformer is F AAAs the coefficients show, aging is much slower than linear. If a transformer's hotspot temperature remains below the rated hotspot temperature throughout its entire operating life, it can be concluded that less power is lost and the transformer is operating under low load. Thus, the transformer will last longer than its assumed lifespan, and the power grid will not need to be replaced immediately. This is the opposite conclusion that power grid operators would have reached if they relied solely on the rating information of the manufacturers of transformers that have been in operation since 1960.

[0096] Therefore, this disclosure provides an effective method for more accurately determining which transformers in the power grid need to be replaced and the priority of such replacements at any given time.

[0097] In summary, transformers are among the more expensive pieces of equipment included in the inventory of power grid operators. This has led to a growing demand for tools to support the protection of transformers and the intelligent monitoring of their status, activity, and history. The monitoring system itself only provides raw information that, unless processed, does not add value to the power grid operator. Therefore, the methods provided in this disclosure bring value to power grid operators in the following ways: • The remaining lifespan value provides information about the aging of transformers in the power grid. This allows power grid operators to assess the condition of transformers and return the total lifespan value. Understanding the current condition enables utilities to make more informed decisions regarding transformer replacement and, more importantly, facilitates them in determining replacement priorities. This directly leads to reductions in capital and operating expenditures. The model includes the hotspot temperature and its associated mean acceleration coefficient F. AA This allows for accurate estimation. This information can provide power grid operators with guidance on how quickly transformers are aging, enabling them to more effectively plan their maintenance activities. The output from the algorithm also provides guidance on specific parameters, such as oil change requests, load capacity reductions, or ventilation supply to further support maintenance activities. • Power grid operators can use the methods of this disclosure to evaluate hotspot temperatures and average acceleration factors to help optimize transformer operation by reducing the load on one or more transformers that appear to be in worse condition than others, for example, and thus enabling power grid operators to better manage the grid so that existing transformers last longer without needing replacement. This disclosure may enable power grid operators to understand the effect of ambient temperature on transformers and facilitate their determination of geographical locations for future transformer replacements.

Claims

1. A computer-based method for estimating the hysteretic apparent power over time of at least one transformer device in a power grid within a geographical area, To provide training data including time-series active and reactive power data of multiple transformer devices in one or more geographical regions, time-series energy consumption data of the power grid in each of the geographical regions, and time-series ambient temperature data for each of the geographical regions, The process involves initializing the weights of an untrained neural network and iteratively updating the weights to minimize a loss function, thereby generating a trained neural network configured to output the estimated apparent power of a transformer device over a time period within a geographical area, using input energy consumption data and input ambient temperature data for the geographical area over the said time period. The apparent power of at least one transformer device within the geographical area over the first time period is estimated by inputting energy consumption data and ambient temperature data over the first time period within the geographical area into the trained neural network. A computer implementation method including

2. The method according to claim 1, wherein the training data further includes voltage harmonic data of the plurality of transformer devices.

3. A computer-based method for estimating the historical temperature over time of transformer devices in a power grid within a geographical area, Estimating the hysteretic apparent power of the transformer device over a first time period using the method described in claim 1, (i) Estimating the hysteretic temperature of the transformer device over the first time period using the estimated hysteretic apparent power over the first time period and (ii) ambient temperature data for the first time period within the geographical area. A computer implementation method including

4. The method according to claim 3, wherein the temperature of the transformer device over the first time period includes the maximum temperature present in one or more coil windings of the transformer device.

5. The method according to claim 4, wherein the maximum temperature present in one or more coil windings of the transformer device is calculated in accordance with IEEE standard C57.

91.

6. A method for managing a power grid having multiple transformer devices within each geographical area, The remaining service life of each transformer device is Estimating the hysteretic apparent power of the transformer device over a first time period using the method described in claim 1, Estimating the historical temperature of the transformer device over the first time period using the method described in any one of claims 2 to 5, The remaining service life of the transformer device is calculated by using the estimated hysteretic apparent power and estimated hysteretic temperature over the first time period to calculate the loss of life, and by subtracting the result of the loss of life calculation from a predetermined maximum service life of the transformer device. To estimate by, If the calculated remaining service life of any of the above-mentioned transformer devices falls below a predetermined threshold, a warning is generated. A method that includes this.

7. The method according to claim 6, wherein the calculation of the lifetime loss includes calculating the lifetime loss in accordance with the ANSI / IEEE 57.91 and IEC-354.91 standards.

8. The method according to claim 6 or 7, wherein calculating the remaining life of the transformer device by performing the calculation of the life loss further includes using real-time measurement of apparent power and real-time measurement of temperature of the transformer device.

9. The method according to claim 8, wherein the real-time measurement of the apparent power of the transformer device includes performing one or more real-time measurements of current, voltage and / or power at the connection portion of the transformer device to the power grid.

10. The method according to claim 8 or 9, wherein the real-time measurement of the temperature of the transformer device includes measuring the outer surface of the housing of the transformer device without opening the housing.

11. The method according to any one of claims 8 to 10, comprising continuously performing the real-time measurement described above.

12. The method according to any one of claims 8 to 10, wherein the predetermined maximum service life of the transformer device is 30 to 60 years.

13. A system for managing the power grid, Processor and Multiple transformer devices in each geographical region, each transformer device being communicatively connected to a server and configured to transmit data to the server indicating real-time apparent power measurements and real-time temperature measurements of the transformer device, and Includes, The aforementioned processor, For each transformer device, the steps include: (i) inputting the received real-time data and (ii) historical power consumption data and historical ambient temperature data for the corresponding geographical area of ​​the transformer device from a first time period, collected before the real-time measurement, into a trained neural network, wherein the trained neural network is configured to output the estimated historical apparent power of the transformer device over the first time period; A step of calculating the remaining service life of the transformer device using the real-time apparent power measurement and the historical apparent power estimation, If the calculated remaining service life of the transformer device falls below a predetermined threshold, the step of generating a warning is performed. A system configured to perform the following actions.

14. The system according to claim 13, wherein the processor is configured to perform the step at least once a day.

15. The system according to claim 13 or 14, comprising a user interface for displaying the geographical regions and locations of the plurality of transformer devices on a map, wherein the user interface is configured to display the warnings.

16. The system according to any one of claims 13 to 15, wherein at least one of the plurality of transformer devices includes an oil-filled transformer device.

17. A computer implementation method for training a neural network to output estimated apparent power of a transformer device within a geographical area over a period of time, from input energy consumption data and input ambient temperature data of the geographical area over the said period of time, To provide training data including apparent power data over time for multiple transformer devices in one or more geographical regions, energy consumption data over time for the power grid in each of the geographical regions, and ambient temperature data over time for each of the geographical regions. The process involves initializing the weights of an untrained neural network, which includes an input layer, multiple hidden layers, and an output layer, and iteratively updating the weights using the training data to minimize a loss function, thereby generating a trained neural network configured to output the estimated apparent power of a transformer device over a time period within a geographical area, using input energy consumption data and input ambient temperature data for the geographical area over the said time period. A computer implementation method including

Images