Study Of Reconfiguration Method Under Self-healing Mechanisms Of Distribution Systems With Microgrid

Dual pressure from environmental protection and energy security, are pushing the sustainable development of renewable energy in energy system. But currently our country is lack of unified technical standard on renewable energy connected to the grid, combined with not enough specification on project construction. So there are increasingly prominent problems on a large number of completed systems, which are unable to be connected for operation. As distributed power supply and micro grid connected, the distribution network will be confronted with the problem of network reconfiguration, fault location, isolation and recovery, and supply restoration. Taking distribution network self-healing as the goal, and operation problems after distributed power supply and micro power grid are connected to distribution network as study object, in view of the basic problem as distribution network reconstruction, a "distribution network-micro grid-local unit" three-stage multi-agent system is established using nodal load prediction (NLP) technique in this paper, which covers following contents:In the self-healing mechanism, distribution network reconfiguration needs to accomplish optimum operation under normal circumstances and autonomous recovery after fault. On the premise of permission of island in reconstruction and NLP, a generalized distribution network reconfiguration model different from conventional ones is proposed. This model including the following content:1) dynamic optimum reconstruction based on NLP under normal circumstances;2) fault recovery reconstruction in the permission of island under fault circumstances;3) island division reconstruction after being disconnected from the external power grid;4) emergency reconstruction of micro grid based on NPL in the island. The model converts distribution network reconfiguration into the conventional island division problem for the extreme case of fault reconfiguration, namely, being disconnected from the external power grid by failure, which fully solves the disadvantages of conventional ones.Combination of measured data and predicted information helps to obtain the optimum cumulative effect topology in reconstruction. Combining link function with Gaussian Process Regression (GPR) method, nodal power prediction model is designed, which refers to historical time sequence of nodal power to establish appropriate mixed Gaussian link function model. And the posterior probability of each Gaussian component is calculated based on the Bayesian inference. After local Gaussian process regression model is established for each Gaussian component, their posterior probability as weight factor of local model are integrated into the global model for wind power prediction. Since time sequence model using power data for characteristic representation as multiple non-Gaussian component through mixed Gaussian link function, it is applicable to a variety of different types of nodes, including load node, distributed power supply node, micro node, and load plus distributed power grid node, etc.Selecting the optimal value of input weight and threshold of neurons, the improved extreme learning machine is proposed in this paper combining with the optimization algorithm of electromagnetic-like mechanism. And its effectiveness is verified by the actual data set from UCI database. Compared with other incremental-type extreme learning machines and fixed-type ones, the test result suggests that such indicators as convergence rate, convergence effect and so on are obviously superior to other extreme learning machines. Finally, EMO-ELM algorithm is applied to the above generalized distribution network reconfiguration problem for case study in this paper. And the effectiveness of the method proposed in this paper is verified by the simulation result.The control of distribution network containing distributed power aimed at self-healing is achieved in the multi-agent environment. Control strategy for both island operation in micro grid and planned island operation is respectively designed, which plays agent's autonomy and interoperability to implement the stability of bus voltage in the mode of island and operation, or during the transient process of switching to island mode of micro grid. A control agent is designed for inverter-type micro power. In detail, the feed-forward compensation based on the droop control is adopted to improve dynamic coupling of the micro power and micro grid, which strengthens the stability of system. A recursive least squares estimation (LSE) algorithm is used to calculate the system steady-state operation point, making feed-forward compensation adaptive.The self-healing mechanism of distribution network manifests in its capacity for self-prevention and self-recovery, which can fully embody and meet the requirements of the current user in reliability. In the self-healing mechanism, distribution network reconfiguration bases on the global measurement information, makes full use of the power prediction data, and uses distributed control agent, which has the function of autonomy and coordination to make up reconfiguration scheme by decision algorithm, which is adaptive to the micro power grid connection. Micro grid can also control each internal unit, and conducts coordination or independent optimization control to be adaptive to task requirements of reconfiguration scheme and the change of the external power grid environment while reconfiguration.