| With the increasing number of distributed energy resources and terminal load devices in active distribution network,the communication data between information centers and nodes grows massively,and the importance of information systems cannot be overemphasized.The communication transmission methods between different locations and devices vary greatly,and the differential uncertainty of information transmission is becoming more and more obvious.However,most of the existing uncertainty models of active distribution network only consider the fluctuation of source-loads,which do not take the uncertainty,ambiguity and interruption of information transmission into account.In the actual system,the uncertainties of information transmission and the prediction of the source-loads are interwoven.It is of great significance to ensure the safety,highquality,economical and flexible power supply of active distribution network by analyzing the multiple coupled uncertainties in the system,performing perceptual prediction and optimal control of the system.Based on the above problems,this paper carries out the following work.Firstly,the optimal path selection of information transmission in hybrid communication network is studied.To solve the problem that information deviates from the actual value in the transmission process,the component states,delay,packet loss and bit error of information transmission are modeled.And the transmission deviation of hybrid communication network under different load rate and information failure rate are quantified.Based on Q-learning,a routing algorithm that minimizes the data deviation is proposed to ensure that the optimal transmission path can still be found even in the case of poor communication quality,and the validity of the model is verified by an example.Secondly,a two-stage stochastic control method considering the uncertainties of information transmission in active distribution network is studied.The data of each sampling point in the trend optimization period is discretized into different scene sets based on the possible uncertainties of information transmission.With the goal of minimizing the total cost,stochastic model predictive control is used to perform rolling optimization of each scene in the cycle,and the output of each distributed energy resource is coordinated.Then,in a shorter real-time correction period,the controllable resources are finely adjusted according to the schedulable ability,which further reduces the influence of the uncertainties of information transmission on the distribution network without increasing the calculation amount of the distribution center.The advantages of the two-stage model in terms of voltage stability and economy are verified by an example.Thirdly,the intraday trend coordination optimization method of distribution network considering multiple coupled uncertainties of cyber-physical system is studied.The cloud model based on conditional value-at-risk is used to model both the information transmission uncertainties and source-loads uncertainties in the active distribution network.By sensing the changes of randomness and fuzziness of the current source-loads of the system,the certainty and confidence interval are set adaptively to determine the worst scenario under the current parameters.The objective function including tail risk is solved in the trend optimization cycle using the adaptive robust model predictive control.A numerical example is used to verify the adaptability and economy of the proposed method in the face of multiple uncertainties.Finally,a real-time correction method of distribution network considering direct load control and interruptible load in the case of uplink communication interruption is studied.When the uplink communication is interrupted,the fuzzy C-means clustering and long-term and short-term memory neural network algorithm are used to finely predict the adjustment range of commercial,residential,building and different categories in industrial loads in each node.In order to ensure the rapidity of real-time operation and the ability to deal with multiple uncertainties,the deep deterministic policy gradient algorithm is used to minimize the sum of expected cost and uncertainty cost.By setting the injection upper limit of the distributed generation,along with the adjustments of controllable resources,the direct load control and interruptible load of each node,these methods can ensure the stability of the system and reduce the peak load of the system when the uplink communication is interrupted.The calculation example shows that the above methods have significant effect in ensuring the safety,economy and flexibility of the active distribution network when the uplink communication is interrupted. |
