The active distribution networks (ADNs) are becoming more complex due to the incorporation of distributed generation (DG) sources and frequent topological changes that make traditional methods of protection obsolete. Despite the rapid progression made in the domain of control strategies, the process of protection has proved to be much more complex. The switching point dynamics add another level of complexity to the process of protection because each change in the switching point status changes the electrical environment upon which the process of relay coordination relies. This paper proposes an adaptive protection scheme based on a multi-agent system and deep neural network (DNN) architecture. The IEEE 34-bus test model is selected and modified to include switching points to facilitate network reconfiguration in real-time conditions. Though flexible in nature and thereby improving the load factor, reducing losses, and leveling the voltage, the system requires an adaptive protection scheme to adjust to the changes in the network structure. The procedure involves two collaborating elements of the system. Firstly, a topology identification agent weighs the statuses of the switching points and promptly changes the protective parameters of intelligent devices based on a fault occurrence to prepare for the next event. Secondly, the DNN-based local backup agent is involved whenever the intelligent devices need backup, and its focus is mostly on the identification of a single-phase ground fault, a frequent fault that continues to be an issue in the Protection devices in an active network environment. The simulations and implementation works were performed using the DIgSILENT platform, MATLAB, and Python programming environment. The results show that the proposed scheme is able to maintain high levels of accuracy and coordination irrespective of the network topological variations. The solution has a significant potential for implementation in the hardware domain and marks a constructive initiative towards the development of more robust and efficient ADNs.
