03-08-2012, 10:48 AM
Implementation and evaluation of a distributed load flow Algorithm for networks with distributed generation
Implementation and evaluation of a distributed load flow Algorithm for networks ......pdf (Size: 3.67 MB / Downloads: 64)
ABSTRACT
In this semester project, an existing algorithm for the calculation of power flow in a distribution network with dispersed generation is implemented in MATLAB and a given network is analyzed. MATLAB with its toolboxes such as Simulink and SimPower Systems is one of the most popular software packages used by educators to enhance teaching the transient and steady-state characteristics of electric machines and power system . Special consideration was given to the voltage profile of the network, and active and reactive power flows. Different types of representative distribution generation profiles and load profiles were adopted. Comparisons were made in case with and without DG units.
The topic was chosen and was carried out after discussions with our mentor, the faculty and after doing a secondary research for a variety of projects on distribution networks. It involved the following steps :-
Phase 1 : Network without dispersed generation
Phase 2 : Network with dispersed generation
Chapter 1
“Load Flow And Distribution Networks”
1.1 Introduction
Load flow is an important tool for proper design and operation of distribution networks. At the design stage, load flow analysis can be used to check whether the voltage limits are satisfied all over the network. At the operation stage, it can be utilized to check the different possible supplying arrangements to fulfill the required voltage profile and to achieve minimum system losses. In recent years, introduction of system automation, dispersed generation and the power industry deregulation have increased the need for powerful tools, including load flow, for the system analysis. Although the conventional load flow methods, e.g. Newton- Raphson and Fast Decoupled, are very efficient for transmission networks, they are not so efficient or even may be unusable in the some cases for distribution networks. The inefficiency of these methods is the result of inherent features of distribution networks. The inefficiency is the cause of the divergence, since distribution networks are placed in ill conditioned networks for these methods. Some of the troublemaking features of distribution networks are: radial or near to radial structure, high R/X ratio, unbalanced multi-phase operation, distributed loads and generation. Regarding to the inherent difference between transmission and distribution networks, the modeling methods of the transmission system components are not suitable for the distribution networks. Therefore, especial modeling methods are required for the distribution networks.
Many methods are presented for distribution load flow solution. The presented methods can be categorized in some groups. Some of these methods are modified versions of Newton-Raphson, and the method deals with formation of Jacobians and computation of power mismatches. Some others are based on the backward and forward sweep method involving branch current flows computing. Moreover, a method is presented for direct solution of the network, by modelling loads as constant impedances. The above mentioned methods are very accurate involving detailed modelling of the system. Most of them are dealing with deterministic conditions and accept only fixed inputs. Therefore, their result accuracy greatly depends on the validity of inputs. The input variables are normally unknown, so they would be obtained by estimation according some past and present data.
~ 8 ~
1.2 SPECIAL FEATURES OF DISTRIBUTION SYSTEMS
1. Radial or near to radial structure. Similar to transmission systems, distribution networks have an interconnected structure, but they are operated in radial mode unlike the transmission systems. One or few rings may exist in the distribution networks, but the number of rings in the whole system is very low. This is because of the arisen difficulties in protection of the ringed networks.
2. High R/X ratio. This ratio is low in the transmission systems, e.g. 0.1 or less, whereas in distribution systems it is in the range of 0.5 to 2 or even higher for the lines with low cross section conductors.
3. Unbalance multi-phase operation. Transmission system are balanced and the loads are supplied through the three phases. Loads in distribution systems can be fed through three-phases, two-phase or single-phase, and so distribution systems may be unbalanced. Therefore, the methods base on the balanced three-phase loads may not be able to solve the distribution networks load flow problem.
4. Distributed loads. Transmission system loads are lumped and they are located on the buses of the system. In the case of each transmission line, there are two loads at each end and there is rarely tap-off load along the line. On the other hand, loads are tapped along the length of the distribution feeder. If each load is considered as a node, it may result in an excessive number of nods. In distribution systems so many laterals are fed via the main feeder. Therefore, each distribution feeder contains some laterals as well as some distributed load points.
5. Dispersed generation. In the transmission systems, the generator buses are considered as a PV bus and their voltage can be set at a desirable value in an allowed rang. In the recent years, dispersed generation has been introduced into distribution systems. This type of generation is based on the renewable energy sources such as solar, wind, or small hydrounits, or on the other hand based on the non-renewable sources such as micro-turbine and small gas-turbine. The controlling strategy of these generators is not similar to the large generators. Some of them are set in a way that they generate a fixed P and Q. On the other hand, some others are set to have a fixed power factor. Only a small fraction of dispersed generators can stabilise the voltage of their connection point, which can be considered as a PV bus.
~ 9 ~
6. Loads in distribution networks. In distribution systems, loads are supplied through very small distribution transformers, e.g. 5 kVA. If each load point is considered as a node, the number of nods would be so high. Modeling the distributed loads, as two lumped loads at the end of a selected line sections, overcomes this problem. Beside the distributed load feature, loads characteristics are also important. Normally, loads are modelled as a fixed P and Q. In some algorithms, loads are modeled as constant impedance component. But loads in a distribution network have both fixed P and Q, and constant impedance features, but the percentage of each one would vary. The validity of a load flow method can be increased, by considering both models simultaneously.
1.3 SUMMARY OF SOME OF DISTRIBUTION LOAD FLOW METHODS
Many efforts have been done on distribution load flow in recent years and some methods are presented. These can be grouped as follows:
1. Backward and forward sweep methods, or ladder network based methods. These methods are based on the radial feature of the distribution networks, i.e. there is a unique path between each node and the source. The algorithm consists of two stages; backward and forward sweep, which are repeated until convergence. Currents or powers are calculated at the backward sweep stage, using the computed voltages. Then voltages are computed at the forward stage, based on the calculated currents or powers. This method can be summarized as: Consider a flat voltage profile for all nodes Determine the impedance between each node and the source Calculate the currents at backward stage, starting from the end nodes Compute the voltages at forward stage, starting from the source Repeat the above two stages until convergence
2. Shirmohammadi method. Shirmohammadi, present a load flow method, which is three-phase version of their method presented in. This can consider the specialities of the distribution systems, especially unbalanced loading. This method can also be used for weekly meshed networks. In these cases, firstly loops are opened and each opened point is compensated by two injected current sources with opposite directions. Therefore, the
~ 10 ~
opening point is converts to two nodes with compensating current sources. After opening the loops, the resultant radial network will be solved. This method can be summarised as: Consider a flat voltage profile for all nodes Compute the current injections due to the voltage profile Determine current flow through the line sections, starting from the end nodes Compute the node voltages due to the line section current flow, starting from the source Repeat the above three stages until convergence
3. Implicit Z method. This method utilizes sparse bifactored Y bus matrix and equivalent current injections. It is based on the superposition principle for voltages. Here, voltage consists of two parts: the voltage due to the source and the voltage due to the equivalent current injections. Loads, generators, capacitors, and reactors are modeled as current injection sources. The voltage relates to the source is the no load voltage, and it can be equal to the source voltage. The other part can not be determined directly, and should be determine through the following procedure: Form the Y and estimate nodes voltages Order and factor the Y bus optimally Calculate the injected current for each node Compute the voltage mismatches using the injected bus currents and factored Y Update the voltages applying the super-position principle Repeat three above stages until convergence
4. Direct method. The previously mentioned methods are based on iterative computations. Goswami and Basu present a method to solve the load flow directly. Here, the load flow problem formulated in a way that the direct solution becomes feasible. Loads are modeled as constant impedances instead of fixed P and Q. Then, KVL's are formed in loops start from the source and last to the end loads. These equations are ordered in a way that the node voltage is on one side. Regarding these equations, Z matrix is formed. The Z matrix then converts to an up-triangular matrix, which is applied to direct solution of the network.