Research On Power Flow Algorithm And Development Of Simulation Software In Smart Distribution Grid

Smart Distribution Grid is a complicated power distribution network characterized by the ability of supporting large amounts of DGs to insert in, fault diagnosis and self-healing and supporting a visual management of distribution grid and its equipment. The above three characteristics bring huge economic benefits and convenience to the society,but at the same time, it faces the following three problems:the first is that large amounts of DGs to insert in will have a serious impact on power flow of distribution grid, but studies on power flow algorithms including DGs are very few. The second is that existing power system simulation softwares are of high integration, low flexibility and expensive price. The last is that classical power flow algorithms are hard to respond to the complex distribution network quickly. This dissertation does research on above three problems.A power system simulation software which is exclusively served in distribution grid is developed using visual programming technology based on Visual Studio platform. In addition to many commonly used functions such as single line diagram drawing, component parameter settings, electrical node topology analysis, power flow calculation, the software has the ability to parsing a grid model based on SVG(Scalable Vector Graphics) format. Newton-Ralph algorithm and back-forward algorithms are applied to NaiMan 16-node system to calculate its power flow based on the designed software. The results show that this software has a powerful visualization analysis ability, back-forward algorithm is better than Newton-Ralph algorithm when calculate distribution grid power flow.Secondly a neural network algorithm based on back-forward method has been proposed, and meanwhile the detailed derivation process and convergence proof of the algorithm are given. In a word, BFNN algorithm has the following three merits:one is neural network’s weights are given as an expression only associated with the node voltage, and therefore, weight trainning means node voltage trainning. The second is that most weights are zero elements. The last merit is that the structure of neural network is determined. Comparing the simulation results abtained by BFNN algorithm and back-forward algorithm, a conclusion can be easily drawed that BFNN algorithm shows a greater superiority than back-forward algorithm no matter in convergence or computational time. Moreover, BFNN’s convergence is not sensitive to initial voltage value.In the last part power flow algorithms including defferent cases of DGs have been analyzed. A kind of dynamic PQ(V) node model considering wind-power characteristic and reactive power-voltage characteristic has been proposed for the first time. Then some new methods have been proposed to deal with and dynamic PQ(V) node in Newton Laphson algorithm and BFNN algorithm. Finally, Newton Laphson algorithm, back-forward method and BFNN algorithm have been applied to calculate power flow in diferent cases that there exist different numbers of DGs, different locations of DGs and different types of DGs in the system, a conclusion is drawn that BFNN algorithm has good robustness to above cases, and it can deal with distribution grid with DGs more effectively.