With the increased penetration of the advanced modern loads, such as distributed renewable energy resources, heat pumps, electric vehicle chargers and energy storage, there is an urge to implement some extent of controllability in the distribution network. This would allow for the better use of existing resources, postponing the need for equipment upgrade. In order to introduce any kind of control to the system, it must be observable in the first place. Establishing observability is a challenge in today's distribution networks, which are known for their low quantity of available network measurements. In the following doctoral thesis, we consider the increase of observability in the power network, with special interest to the medium-voltage and low-voltage distribution networks.
Key features of nowadays power networks are presented in the beginning of the thesis. Due to its context, here we focus mostly to the characteristics of the medium and low voltage distribution networks and their key differences in comparison with the high voltage transmission networks. The latter are long known for a successful implementation of different methods that render them observable. With the increasing penetration of modern loads, similar task is now foreseen also for the distribution networks. Due to the significant discrepancy between the two, direct implementation of methods, used in transmission system observability is not possible in the distribution networks. This challenge can be addressed with a dedicated algorithm to assess the state of the distribution system. Recent efforts in the field of the distribution system state estimation are presented at the end of the first chapter.
Next chapter gives the mathematical background of crucial parts of the state estimation algorithm. As these represent the heart of the estimation process, their basic knowledge is of great importance for understanding the operation of the algorithm itself. So, the mathematical modelling of the distribution network is given first. As a result of different topologies an electrical characteristic, obtained model differs significantly, from the transmission network models. Additionally, also the necessary measurement functions are explained in this part of the work. These functions and their derivatives enable the estimation algorithm to relate the system measurements with the appropriate state variables, taking into account the mathematical model of the network.
The introductory part concludes with a presentation of the weighted least squares method, which is commonly used in the state estimation algorithms. Consequently, it is chosen as a reference for assessment of the developed new algorithm performance. Description of the Kalman filter method is provided next. Kalman filter represents the basis for the developed new state estimation algorithm.
A new state estimation algorithm was developed under the scope of the doctoral thesis. Its performance was demonstrated also through field testing in the autonomous estimation of the real Slovenian distribution networks. New algorithm is described in the dedicated chapter. There, the underlying tasks of measurement acquisition and mathematical model development are explained in detail. The straightforward usage of solely network measurements is not sufficient in the distribution network state estimation, as their number is not adequate for establishing the appropriate level of measurement redundancy. Mathematical model, being the other key input of the state estimator, should also be prepared in order to faithfully mimic the real network in order to calculate accurate estimate of the network state. In the process of state estimation, the algorithm attaches the obtained measurements to the mathematical model of the network and thus, in some way, checks their accuracy. This is only possible with a properly defined network model.
The presentation of the upgrades, developed and tested as part of the doctoral thesis, are given in the end of the chapter. Firstly, the simplification of the classical weighted least squares method is presented. It enables a faster calculation of the state estimation in the low-voltage distribution system. Simplification is based on the decoupling, known from the transmission system. It takes advantage of some key characteristics of the grid, which allow neglection of the individual parts of the system matrices. Taking into account the differences between the two voltage levels, we redesigned the method and proved its successful performance in the low-voltage network state estimation. However, the employed assumptions are true only in the low voltage network, so the method cannot be applied directly to the medium voltage level. Additionally, estimators based on the weighted least squares method have a problem with robustness in case of great measurement errors, which tend to happen frequently in today’s distribution networks measurement systems.
The initial investigations of the established measurement system revealed frequent measurement drops. Consequently, we had to focus ourselves to more robust methods for state estimator. Extended Kalman filter belongs to such methods and was therefore chosen as a basis for the developed new algorithm. During the initial studies it turned out that the state estimation process can be significantly accelerated with introduction of some simplifications to the calculation of system matrices. These can be kept constant when there is no significant change in the estimated system. With additional successful implementation of the system change detection we get to a fast, robust and at the same time accurate algorithm for distribution network state estimation, which is the main outcome of the presented doctoral thesis.
In the end, the results of the developed method implementation are presented. This part includes results of the laboratory testing, followed by the results from the real network field testing. Laboratory testing has proven that the developed algorithm adequately detects and overcomes the challenges, expected to emerge in the actual process of distribution network state estimation.
This was further on proven through the developed estimator field test and operation. Its brief outcomes conclude the last chapter, together with the difficulties faced during the development and practical implementation of the estimator in the field. These obstacles had to be properly addressed and mitigated prior the algorithm implementation.
The final part gives conclusions and summarizes some of the main outcomes of the work. The significant shortcomings of today's distribution networks in the light of system observability are also highlighted. These can be packed to unreliable measurement system and outdated, scarce network databases.
|
