Presentation on LSMS2017

1 | Ladies and gentleman, It is my honor here to introduce you the work of our team 2 | Here is the content of this reportsection 1...section 2...section 3...section 4. is the proposed consensus algorithm which is utilized to distribute the centralized algorithm, it is the center of the report, section 5 3 | First of all, let's look at the features of the proposed algorithm featue 2 A double layer multi-agent model is proposed in this paper to discribe the power sysytemfeatue 5, in this paper, the DC Optimal power flow problem is chosen to illustrate the effectiveness and performance of the proposed algorithm 4 | The disadvantages of traditional centralized generation, such as high pollution, high transmission loss urgess the development of distributed generation(DG).The DG entities, such as CCHP, PVs，WTs and fuel cells are generally applied in the power grid, which motives the control and coordination of power system to be more distributed. so we propose the distributed algorithm to 5 | Now, let's talk about the DC OPF problem, the objective of this problem is... 6 | I can’t introduce this algorithm in detail because the reason of time, you can read our paper for more information. By research the procedure of the PDP algorithm, we found that, In each bus, if we can locally get the voltage angle estimations of all buses in the system , then we can distribute the PDP subgradient algorithmin the following anaysis, let's set the 6 buses power sysytem as an example. 7 | For further analysis, we have the following equations.xij is the impedance of transimission line btween bus i and bus jpisp is the real power injection of bus iinverse matrix of Baccording to the equations in the red frame, we can calculate the voltage angle of each bus, just like.and unfold this equation, we found that if we can calculate。。。in each bus locally, then we distribute the PDP algorithm, so that is kernel of this problem[B-1]ij is the element of B-1 in i row and j column 8 | To achieve this goal, here we propose a double layer multi-agent model.We still set the 6-bus power system in the left graph as an example, and the corresponding double layer multi-agent system is show in the left graph, here we can see, each generator or load bus is considered as a bus agent, and connect to the grid through a network agent. We assume that part of the the structure information of the grid, that is the inverse of matrix B, is accessible to the network agent. 9 | Now, let's talk about the consensus algrithm. According to our analysis, each network agent in the system can get an voltage angle estimation vector locally, let's set network agent 2 as an example, the elements of the vector are angle estimations from bus 1 to bus 6 in bus 2. For all network agents in the system, there has the followung matrix, then, we take jth row of the matrix and make it the initial status of the corresponding network agent. 10 | Then each network agent communicate only with its neigbors and update its status(特尔) iteratively by the consensus protocol on the right of the screen.with iteration increase, each network agent reaches a consensus value on the estimation of voltage angle in bus j, which equal to the average value of all initial estimations. and then multiple it by 6, we acctually get a jth voltage angle estimation in every network agents of the system locally.And in this way, it's obvious that each node can acquire the voltage angle estimation of all network agents in the system.and in this way, we can distribute the refered PDP subgradient algorithm 11 | this table is the optimal solution comparision between centralized and distributed PDP algorithm, we can see the error is very small 12 | also the time-varing demand test and plug and play capability of the proposed algorithm is verified. which shows the good performance of the algorithm.in each case the system can reaches steady state fastly 13 | that is all of my report, thank you for your listening. for more disscussion connect me by email

zhongyuanyang / -

Presentation on LSMS2017 #2