METHOD FOR OPERATING A TRANSFORMER

Abstract

The aging (33) of a transformer is predicted taking into account environmental influences (13) and load requirements (14) on the transformer, and a financial parameter (41) of the transformer operation is derived from this. By optimizing with respect to this financial parameter (41), suitable operating settings (51, 52) of the transformer are determined. The determined operating settings (51, 52) are set by controlling (100) the transformer. The optimization can take maintenance planning into account.

Description

[0001] The invention relates to a method for operating a transformer in which operating settings of the transformer are made with regard to economical transformer operation.

[0002] The average age of components such as transformers in power grids is increasing, which is bringing the economic viability of continued operation, replacement, or specific maintenance measures into sharper focus. At the same time, electricity generation from relatively fluctuating sources such as wind or solar power is increasing through comparatively decentralized installations. This results in more variable load demands on transformers in the corresponding power grid, which can accelerate aging processes in the transformer, for example, due to more varied thermal stress. Assessing the economic viability of transformer operation is thus made more difficult. Established industry standards for transformer operation, such as IEC 60076-7 or IEEE C57.91, are insufficient in this context.

[0003] The object of the invention is to provide a method that makes it possible to operate a transformer in the most economical way possible.

[0004] This problem is solved by a method according to claim 1. The dependent claims relate to advantageous further developments.

[0005] The inventive method for operating a transformer comprises at least the following steps: External attributes for the transformer are predicted. External attributes are potential influencing factors on the transformer that originate outside of the transformer itself. At a minimum, the external attributes considered include local weather data at the transformer's installation site and a load requirement for the transformer. The predicted load requirement describes, in particular, a load that is expected to be present at the transformer at a specific time or over a specific period to which the prediction refers. The load requirement for the transformer results from the demand of the power grid supplied (in part) by the transformer in question and is therefore external in nature.The load demand forecast can, for example, take into account empirical data on daily fluctuations in the demand for electrical energy in the power grid, as well as longer-term fluctuations; the load demand can also be determined by specific instructions from a control center. The predicted load demand can also consider the weather forecast. The weather data at the transformer's installation site is local in the sense that the weather conditions are considered as precisely as possible at the transformer's location. Buildings or terrain features (such as hills, depressions, and trees) in the vicinity of the transformer can also have an influence.

[0006] Next, at least one internal attribute of the transformer is predicted. This prediction uses the predicted external attributes and predefined operating settings for the transformer. Internal attributes of the transformer are properties and parameters of the transformer that originate from or are caused by the configuration and operation of the transformer itself. One or more internal attributes can be predicted. At a minimum, a winding temperature and / or an oil temperature within the transformer are considered internal attributes.

[0007] Based on the predicted at least one internal attribute and the specified operating settings for the transformer, transformer aging is predicted.

[0008] Subsequently, a financial indicator for the transformer operation is determined, based on the predicted aging of the transformer.

[0009] By varying the predefined operating settings and repeating the steps described above, at least from the prediction of at least one internal attribute onwards, the financial indicator is optimized. This also yields optimal operating settings, meaning those transformer operating settings for which an optimal value for the financial indicator is obtained.

[0010] The transformer is controlled to set the optimal operating settings determined through optimization.

[0011] The method achieves economical transformer operation by reliably identifying and ultimately adjusting the transformer's operating settings for which the relevant financial parameter is optimal. The method provides a comprehensive evaluation of all factors relevant to transformer operation, considering their interaction, rather than focusing solely on individual factors.

[0012] The transformer aging considered in this process can be expressed, for example, by the decrease in the expected remaining service life of the transformer. This decrease is determined, firstly, by the aging of the paper used to insulate the transformer windings. The latter can be quantified, for instance, by a decrease in the degree of polymerization of the paper, which can be predicted deterministically. Key influencing factors here are the water and oxygen content in the oil and their behavior; for example, diffusion between the oil and paper is a temperature-dependent process. Another influencing factor on the transformer's aging is the failure probability of its components. This allows for statistical predictions about the failure of individual components.Such failure probabilities can result, for example, from empirical data and depend on the operating time and operating mode of the respective component; the operating mode of a component can depend on the specified operating settings.

[0013] Within this process, approaches such as model-based predictive control can be used to achieve optimization. Other approaches or combinations of different approaches are also conceivable.

[0014] In a general embodiment, the prediction of the external attributes and / or the at least one internal attribute includes determining a respective probability distribution for future values ​​of the corresponding external or internal attribute.

[0015] However, configurations are also conceivable in which specific values ​​(instead of a distribution of values) are used for some or all external and / or internal attributes. In this context, it would also be conceivable that some or all of the specific values ​​are determined as expected values ​​from a corresponding probability distribution or as the value at which the corresponding probability distribution reaches its maximum.

[0016] In a general embodiment, predicting the aging of the transformer includes determining a probability distribution for the aging of the transformer. Likewise, in a general embodiment, determining the financial parameter of the transformer operation includes determining a probability distribution of the financial parameter.

[0017] Here too, it would be conceivable in other configurations to work with a specific value instead of a distribution for aging and / or the financial indicator.

[0018] Using specific values ​​for external attributes, internal attributes, aging, and financial metrics can simplify the calculation methods. Using probability distributions for these variables leads to more reliable results.

[0019] In one embodiment, the local weather data includes one or more of the following: outside temperature, type of precipitation, amount of precipitation, precipitation intensity, wind speed, wind direction, solar radiation, and cloud cover. The outside temperature clearly influences the temperature of the transformer. Precipitation, such as rain or snow, falling on the transformer has an additional cooling effect. This cooling effect depends on the type of precipitation, the amount of precipitation, and the temporal distribution of the precipitation, i.e., its intensity. Wind can dissipate heat from the transformer. Solar radiation, which is influenced by cloud cover, can cause additional heating of the transformer.These examples illustrate the importance of the locality of weather data for a reliable procedural result: For the effects of solar radiation, it is not only important whether it is a sunny day in the area where the transformer is located. Rather, it is also necessary to consider whether, and to what extent, the transformer is shaded by a neighboring building. The same applies to wind effects; in the lee of a building, there are fewer or no cooling effects from wind.

[0020] In one embodiment, the local weather forecast comprises combining forecast data provided by a weather service with a correlation between past forecast data provided by the weather service and corresponding past locally measured weather data. While the forecast data provided by the weather service generally refers to a geographic area in which the transformer is located, which may have a diameter of one kilometer or more, the local weather data is measured at the transformer's installation site. A correlation between the forecast data and the corresponding locally measured weather data can be determined by comparing past forecast data with the relevant locally measured weather data.This correlation can take the form of a probability distribution, such as a probability distribution for specific local weather data like outside temperature or wind speed, depending on forecasts for corresponding weather data from the weather service. This correlation can then be used to predict the local weather data at the transformer's installation site, based on a current forecast from the weather service.

[0021] In a training course, the correlation between previous forecast data provided by the weather service and corresponding earlier locally measured weather data is maintained and updated using machine learning. This means that the machine learning is not limited to an initial learning phase based on a dataset of forecast data from the weather service and corresponding local weather data available at a given point in time, but rather that further forecast data from the weather service and corresponding local weather data generated during ongoing operation are used as additional training data.

[0022] In one embodiment, the transformer aging prediction takes into account measurement data about the transformer. This measurement data includes at least the water content and / or oxygen content of the oil in the transformer. Considering measurement data about the transformer, particularly water content and / or oxygen content, enables more accurate predictions.

[0023] In one embodiment, the prediction of the at least one internal attribute involves machine learning. Here, too, the corresponding machine learning system can be continuously updated. Model parameters that change over time can be used to monitor the condition of the transformer.

[0024] In one embodiment, the prediction of at least one internal attribute takes into account the temperature dependence of a characteristic parameter of the oil in the transformer. This parameter could be, for example, the oil's viscosity, thermal conductivity, gas solubility, or water absorption capacity. The oil's viscosity is particularly relevant to the process due to its influence on convective heat transfer within the oil. The temperature dependence of the viscosity can also be used to deliberately avoid local temperature peaks in the transformer. For this purpose, the oil is heated, or heating is not counteracted, in order to achieve improved convective heat transfer, for example, away from winding sections with peak temperatures, through the resulting reduction in viscosity. At the same time, an increase in the oil temperature in the transformer can sometimes improve the moisture distribution within the oil.Such approaches are among the predefined operating settings of the transformer.

[0025] In one embodiment, the predefined operating settings include a switching strategy for a transformer's on-load tap changer. It is conceivable, for example, that a load requirement for the transformer can be met in more than one switching position of the on-load tap changer; the crucial requirement here is to maintain the voltage supplied by the transformer within a predefined tolerance range around a target value of the grid voltage. However, a different amount of heat can be generated in the transformer in each of these switching positions, depending on what portion of a transformer coil is carrying current. The heat generated in the transformer affects its aging. On the other hand, maintenance work on the on-load tap changer is required after a certain number of switching operations.A lower number of switching operations per unit of time therefore allows for such maintenance work to be carried out later. However, a lower number of switching operations per unit of time also means accepting sometimes greater fluctuations in the voltage supplied by the transformer and operating the transformer more frequently in a switching position that is not optimal in terms of heat generation within the transformer and its associated aging. This method makes it possible to consider these sometimes conflicting effects of switching operations during optimization and thus determine the timing of switching operations and the switching positions of the on-load tap changer to be used in each case—in other words, a switching strategy.

[0026] In a further training course, the prediction of at least one internal attribute also takes into account heat generated by the operation of the transformer's own load tap changer.

[0027] In one embodiment, the predefined operating settings include the speed of a transformer fan. A high fan speed means high cooling capacity, but also increased energy consumption and increased fan wear. Regarding transformer cooling, it can also be considered to increase the cooling capacity in anticipation of an increased load demand, even before this demand actually occurs; here, the transformer is, in effect, pre-cooled to mitigate thermal stress. This approach is also among the predefined operating settings of the transformer.

[0028] In one embodiment, the specified operating settings include an operating mode for a transformer dehumidifier. The dehumidifier serves to dehumidify air flowing into the transformer when the oil volume in the transformer decreases due to a drop in oil temperature. The dehumidifier must be regenerated periodically, which is achieved by heating it up. This requires energy and can result in heat being introduced into the transformer. However, the timing of the regeneration is crucial. This heating process should ideally occur during a phase in which air is flowing out of the transformer, thus increasing the oil volume in the transformer due to a rise in oil temperature. If, on the other hand, the regeneration takes place while air is flowing into the transformer, the water extracted from the dehumidifier also enters the transformer.The less moisture stored in the dehumidifier material (e.g., diatomaceous earth) of the air flowing into the transformer, the more efficient the dehumidification of the air is. Dehumidification efficiency is therefore particularly high after a regeneration cycle. However, the dehumidifier material must be replaced after a limited number of dehumidification cycles. Frequent regeneration thus means consistently high dehumidification efficiency over time, but also frequent replacement of the dehumidifier material. The process can also take this trade-off into account during optimization.

[0029] In one embodiment, optimizing the financial parameter considers the potential for performing maintenance on the transformer. Maintenance measures entail additional financial expenditure but can slow down the transformer's aging. For example, a component with a high failure probability due to age can be replaced with a new component whose failure probability is significantly lower. For instance, the oil in the transformer can be degassed; this reduces the aging of the insulating paper. Degassing the oil can lead to a reduction in the oxygen content of the oil and thus also in the insulating paper. In particular, a reduced oxygen content in the insulating paper can slow its aging. In a further development, the method generates a schedule for carrying out specific maintenance measures.Such a schedule is generated as part of the optimization process.

[0030] In one embodiment, the financial metric includes one of the following: net present value, return on investment, or time to amortization. The financial metric can incorporate accelerated aging as the acceleration of acquisition costs, since a replacement of the transformer becomes necessary sooner. High loads on the transformer can also manifest as an increased failure risk for various transformer components, which can be factored into the financial metric through the costs of replacing these components.

[0031] In one embodiment, the method takes into account the requirements of a power grid into which the transformer is integrated. For example, different load requirements for the transformer can be considered, corresponding to different load distributions across multiple transformers in the power grid. Maintenance work on the power grid might mean that another transformer is out of service for a period of time. This can result in additional load for the transformer under consideration, which can be taken into account in the claimed method.

[0032] The predictions made within the framework of this process can cover a period of approximately 2 to 24 hours. However, predictions over longer periods are also conceivable. For example, predictions with a time horizon of a few days to a week are advantageous for planning the deployment of maintenance technicians more efficiently, particularly with regard to their travel. Seasonal fluctuations, such as those in load and weather, can also be taken into account in the models. Changes due to climate change can also be incorporated.

[0033] In any case, the procedure can be repeated to ensure optimized transformer operation over longer periods of time.

[0034] The invention and its advantages are explained in more detail below with reference to the attached drawing. Fig. Figure 1 schematically illustrates an example of the method according to the invention.

[0035] Fig. Figure 1 schematically shows a non-restrictive example of the method according to the invention. A prediction model 10 operates on the basis of input data 11, 12. The input data includes non-local input data 11, such as forecast data provided by a weather service and other data, such as specifications from a control center for a power grid into which the transformer under consideration is integrated. The input data also includes local input data 12, such as weather data measured in the past at the installation site of the transformer, and, for example, stored load profiles for the transformer. Based on this input data, the prediction model 10 determines predictions for external attributes of the transformer, including a prediction of local weather data 13 and a prediction of a load requirement 14 for the transformer. Both predictions can be provided by the prediction model 10 as respective probability distributions.

[0036] The predictions of the external attributes serve as input for a thermal model 20. The thermal model 20 also takes into account predefined operating settings 51 for the transformer, which can be adjusted during the process. From this provided data, the thermal model 20 generates a prediction for at least one internal attribute 21, such as a winding temperature of the transformer. This prediction can also be generated as a probability distribution for each internal attribute 21.

[0037] The predicted at least one internal attribute 21 serves as input for an aging model 30. The aging model 30 also receives measurement data for the transformer, including a water content 31 of the oil in the transformer and an oxygen content 32 of the oil in the transformer, as well as predefined operating settings 52 for the transformer. From this, the aging model 30 determines an aging 33 of the transformer. This aging 33 can also be generated as a probability distribution.

[0038] The results of the thermal model 20 and the aging model 30 are summarized and evaluated by a financial model 40. As a result of this summary and evaluation, the financial model 40 outputs a financial parameter 41 for transformer operation to an optimizer 50. The financial parameter 41 can be expressed as a probability distribution.

[0039] Optimizer 50 is used to optimize transformer operation with respect to the financial parameter 41. To this end, optimizer 50 varies the operating settings 51 and 52, and the procedure performs a new determination of the financial parameter 41, as described above. In this way, an optimal financial parameter 41 and corresponding optimal operating settings 51 and 52 are ultimately obtained.

[0040] These optimal operating settings are set on the transformer by controlling 100.

[0041] In the Fig.In the example shown, the thermal model 20 is assigned a set of operating settings 51, and the aging model 30 is assigned a set of operating settings 52. It is conceivable that the sets of operating settings 51 and 52 are disjoint. However, it is also conceivable that there is a non-empty cut between the sets of operating settings 51 and 52, such that some operating settings are included in both sets. Likewise, the sets of operating settings 51 and 52 can be completely identical. As an example, consider the switching strategy for a load tap changer. Because the heat generation in the transformer depends on the switching position of the load tap changer, the switching strategy is relevant for the thermal model 20 and advantageously forms part of the operating settings 51.Because of the wear and tear of the load tap changer associated with switching operations, the switching strategy is also relevant for the aging model 30 and advantageously forms part of the operating settings 52.

[0042] The process can be carried out by a data processing system, which may include, for example, a computer or several interconnected computers. The data processing system has suitable communication interfaces for receiving, for example, non-local input data 11 and suitable storage media for storing the local input data 12. Furthermore, such a data processing system can be in communicative contact with sensors to obtain, for example, current local weather data and measurement data from the transformer. In addition, such a data processing system can have an interface through which the transformer is controlled for setting operating parameters.

[0043] Predictive models, thermal models, and aging models can each utilize different, but essentially well-known, methods to fulfill their respective tasks. For example, Monte Carlo simulations are suitable for handling quantities given as probability distributions.

[0044] In addition to the input data, measurement data, predictions and operating settings described above, the models also use additional information, for example on the configuration of the transformer and on restrictions or specifications for the operation of the transformer.

[0045] For example, a maximum permissible winding temperature or oil temperature may be specified, above which gas bubbles would form in the oil, potentially leading to failure of the winding insulation. Related to this is the overload capacity of a transformer; this is the maximum overload at which the transformer can be operated, possibly for a defined limited period, without exceeding a maximum permissible temperature.

[0046] It is conceivable that the transformer will be used in scenarios where specific requirements must be met. These requirements may impose limitations on the transformer; for example, certain values ​​or value ranges may be excluded for some operating settings because these values ​​or value ranges would violate the aforementioned limitations. Such values ​​or value ranges are then not considered in the models, even if the transformer would theoretically be capable of handling them.

[0047] Examples of other possible parameters that can be taken into account by the models include, for the financial model, regulatory requirements, internal company guidelines (e.g. for investments, risk tolerance), inflation scenarios, depreciation rules, market conditions (e.g. procurement costs for spare parts or a complete new device, including delivery, installation or commissioning), and CO2 emission costs.

[0048] For the thermal and aging models, parameters describing the transformer's configuration are of primary importance, such as oil volume, losses in the windings and core. It is also conceivable to consider the overall age of the transformer or its components, i.e., the time elapsed since its manufacture or initial use. Furthermore, the type of insulator used can be relevant to the models.

[0049] It is conceivable that the parameters in the models will be adjusted if they change in reality. For example, regulatory requirements and company policies may change. Replacing components in the transformer can alter its configuration in ways relevant to the models. Reference symbol list 10 Predictive model 11 non-local input data 12 local input data 13. External attribute: local weather data 14 External attribute: Load request 20 thermal model 21 Internal attribute: Winding temperature 30 Aging Model 31 Water content (in the transformer oil) 32 Oxygen content (in the transformer oil) 33 Aging 40 financial model 41 financial indicator 50 optimizers 51 Operating mode 52 Operating mode 100 Control (of the transformer)

Images