Self learning expert system (SLES) for power transformers

Abstract

In a conventional approach the capacity of a power transformer is based on rated power according to the IEC. Therefore continuous load and fixed environmental and operation parameters are used. However, these parameters as well as the load change considerably during the operation of the power transformer. These parameters have significant influence on the temperature behaviour inside the transformer. Our research has shown that the standard IEC-model does not describe the transformers’ thermal behaviour with enough accuracy. Therefore the Dutch DSO Alliander is developing a self-learning expert system (SLES), to define the capacity more accurate. An automatic optimization routine has been added to the SLES to estimate the thermal parameters of each transformer. This system learns from the thermal behaviour in the past and adopts it’s thermal models. With the system Alliander can predict the real capacity of its power transformers at any time and under every condition in a safe and flexible way. As a further step in the development of this SLES, next to the top-oil temperature measurements a couple of new transformers have been provided with fiber optics to get a permanent and direct measurement of the winding temperature. These transformers will serve as test objects to get more information about the thermal behaviour of power transformers in the formation of hot spots. Integrating the SLES power grids will allow guaranteeing a safe operation and efficiently planning grid investments. INTRODUCTION The standard methods to determine the capacity of power transformers and other components in the grid are based on static rating. Continuous load and fixed environmental and operational parameters are used for this purpose. However, environmental parameters (solar radiation, wind, ambient temperature, etc) as well as the load change considerably during the operation of the power transformer. The environment and the load, together with the own thermal behaviour of the transformer play a significant role in its temperature variation. Because the loadability of the transformer depends on its maximal allowed temperature, accurate monitoring of these environmental parameters and load profiles is essential. Using the data obtained from this monitoring in combination with an adequate modelling, the real capacity of the transformer can be determined. This so called dynamic rating system allows guaranteeing a safe operation and efficiently planning grid investments. DYNAMIC LOADING OF POWER TRANSFORMERS: The benefits The Dutch DSO Alliander attempts to operate its transformers in a safe and flexible way. To achieve this goal, a method to define the capacity of each single power transformer at any time and under every condition is necessary. Using such a method it is possible to predict the temperature of the transformer winding and in this way the possible arise of hot-spots. Preventing damage of the transformer while exploiting it optimally will be the result. An important aspect were dynamic rating systems offer significant advantages is the investment planning. Figure 1 shows how investment decisions are taken, based on rated capacity and predicted load. Inves tmen t po lic y bas ed on rated c apac ity ‐12 ‐11‐10 ‐9 ‐8 ‐7 ‐6 ‐5 ‐4 ‐3 ‐2 ‐1 0 1 2 3 4 5 6 7 8 9 10 11 12 T ime [years ] S [M V A ] Predicted Load [MVA] Rated Capacity [MVA] bottleneck