基于神经网络的柔性互联装置功率自主调控方法OA

[Objective]The soft open point(SOP)is a key solution for addressing challenges such as distributed photovoltaic(PV)integration,as it can connect multiple low-voltage distribution networks and offer functions including flexible power regulation.However,existing SOPs require power data from distribution transformer service areas to formulate their operating strategies,necessitating additional detection devices,energy management systems,and communication infrastructure,which significantly increases system complexity.A pressing problem lies in enabling SOPs to operate autonomously without reliance on external commands.[Methods]This paper proposed a novel neural network-based power control method for SOPs.First,a neural network model is constructed using low-time-resolution historical power sampling data and local voltage sampling data from the SOP.Subsequently,the neural network is deployed within the SOP controller,enabling the device to predict the transformer service area power deficit using real-time voltage sampling data.Finally,the power outputs of the multiple ports are regulated according to the predicted power deficit to achieve real-time power balancing among the connected distribution transformer service areas.[Results]The neural network model is trained using real-world transformer service area data and deployed in both a field-installed SOP and a laboratory test platform.Experimental results demonstrate that the proposed method successfully achieves automatic power distribution among the distribution transformer service areas,with an average deviation of less than 2 kW compared to the ideal equal power distribution values.[Conclusions]The findings indicate that this method enables power balancing using only local voltage sampling,effectively reducing the operation and maintenance costs and complexity of SOPs.This facilitates the broader adoption of SOPs,and promotes the local consumption of distributed PV generation.