| Distribution systems are an important part of electric power systems for two reasons: (1) Capital deployed in distribution systems constitutes about half of the total amount spent in entire power system; (2) Distribution systems are directly connected to the customers.; Due to the large scale of distribution systems, it is not economically justified to place meters on every branch in order to fulfill the automation requirements. The main goal of this dissertation is to present and implement a method of meter placement for load estimation, using as few meters as possible.; An approach for solving the probabilistic distribution power required for state estimation. First, apply the backward and forward sweeps to solve the deterministic power flow. Second, use the similar backward and forward sweeps to calculate the variance values. Sets of equations for variance power flow and load conversion are derived. In order to implement the approach efficiently with Microsoft Visual C++ tree representation and traversal are used. The tests on sample systems show the approach is robust, fast and suitable for solving large distribution system problems.; The features of the distribution system loads, such as load class and time of day, have also been investigated. The load estimation method is based on the customer billing data (or transformer capacity) and typical daily load curves in different load classes. The method is implemented with Microsoft Visual C++. Results are analyzed in order to combine load estimation with the meter placement approach.; The meter placement method used is a two-stage approach. The basic steps of the first stage are: (1) Conduct probabilistic distribution power flow analysis by using minimum loads as inputs. The resulting branch power flows are the meter readings for load estimation; (2) Inversely traverse the tree from leaves to root and use the meter readings obtained from step 1, the down stream loads are estimated based on the maximum loads. Depending on the load estimation error, either a meter is placed on the down stream branch, or the load estimation is continued on the upstream branches; (3) After a meter is placed, the upstream meter readings are updated by removing the sub-trees.; At the second stage, the confidence interval is calculated by using meters placed at the first stage when the loads vary between the maximum and minimum. If the meters give a satisfactory confidence interval, the meter placement is done. Otherwise, it is necessary to go back to the first stage to place meters under a stricter criterion. Sample system analysis and testing results show the approach is efficient for finding tentative meter locations. Real application constraints such as meter failure backup, availability of space, automated switch locations and unbalanced systems, also are considered. (Abstract shortened by UMI.)... |
