POBLACIÓN OCUPADA POR RAMA DE ACTIVIDAD

Although the deployment was tested with a particular hardware setup, it could have been deployed on a wide range of computing devices. The only requirement is to provide the Java 6 Runtime Environment (or higher) to be able to run the JADE platform (version 4.1 or higher). Obviously, there also needs to be enough computer memory available to run the JADE platform with the agents and a network card to provide wired or wireless communication.

For the developed protection scheme, it was assumed that the Relay Agents would run directly on the relays. Considering the cheap but powerful computing hardware of today and the low system requirements of a Java runtime environment and memory requirements in the order of 24MB, it is a reasonable and technical feasible assumption. The manufacturers of protection relays would have to provide this environment before the protection system can be fully deployed as envisaged. Alternatively, the Relay Agents can be deployed on separate hardware, such as the one used for the test setup, and interface with the relays through external data buses or communication networks depending on the relay models and their features. Many modern relays feature serial data communication over fibre-optic cable, 2-wire connections, or IP-based communication networks. For instance, the distance protection equipment of the Siemens SIPROTEC series [115] can be extended by pluggable communication modules usable for different and redundant protocols such as IEC 61850, IEC 60870-5-103, IEC 60870-5-104, and DNP3 (both serial and through TCP).

The simulation and validation of this alternative deployment would have to reflect this new setup. Although the developed system was not meant to be deployed in this way, the simulations could be easily expanded to account for it. For example, the relays can be modelled as additional agents in JADE to verify the correct communication between them and the Relay Agents. They would also account for any time delays the relays might add to the system performance. These models represent the physical relays and are only used for the simulations and to validate the system before deployment.

5.10 Summary

This chapter described the implementation and analysis of an agent-based remote backup relay supervision scheme. The aim of the supervision scheme was to help prevent un- necessary transmission line tripping due to hidden failures. It was developed for the IEEE 39 bus system and was validated with the help of the MAC-Sim platform. Dif- ferent application designs and communication technologies were analysed under various communication network scenarios. Application designs included different peer selec- tion algorithms and communication strategies such as client-server communication and peer-to-peer communication. The impact of agent communication with different agent message sizes and different transport protocols was assessed. Quality of Service (QoS) strategies, background traffic, the impact of communication link outages, and additional communication links were also considered in the scenarios. The obtained results showed clearly that the communication infrastructure has a profound impact on the MAS per- formance and clear recommendations were made on viable technologies and designs.

This chapter also identified the system requirements of MAC-Sim for the validation of the supervision scheme. The computer memory and data storage requirements were easily met with today’s computing equipment. Also, the measured simulation execution times suggest that the time overhead added to the individual federates by the additionally developed extensions, is negligible.

6 Conclusion

Multi-agent systems have been utilised to solve complex problems and build decen- tralised, flexible, fault tolerant, and extensible systems. One area where MAS have increasingly been applied to is the power industry. Massive research efforts are under- way to design the new generation of power systems, often referred to as the Smart Grid. It is the general vision to design the Smart Grid as a more decentralised system with the availability of digital communication virtually everywhere. Multi-agent systems are widely suggested as a promising method for the realisation of a wide variety of Smart Grid applications such as power restoration, protection, and control. However, the de- velopment and validation of deployable MAS has been challenging. Agent applications make extensive use of digital communication, which influences the overall behaviour and operation of the overall system by introducing additional time delays. Especially delay-sensitive and time-critical applications need to be validated by methods that fully account for the underlying communication infrastructure.

This research has presented a novel method for the accurate validation of multi-agent systems. The validation accounts for the agent code, overall MAS functionality as well as the impact of the underlying communication infrastructure. It therefore supports the assessment of different applications, communication technologies, and scenarios.

In addition, an actual implementation, called MAC-Sim, has been presented that fed- erates the FIPA-compliant Java Agent DEvelopment (JADE) platform and the network simulator OPNET Modeler. The integration of the standard-based and well-established tools meets the research aims in terms of compatibility with relevant agent standards and features the tools offer. Also, the integration of JADE provides a platform for agent-oriented software engineering, which was another requirement for this research. It utilises a standardised distribution simulation modelling architecture, which offers open interfaces that allow the integration of other tools in the future.

This work has addressed the challenges of federating existing tools that were not de- signed for this purpose such as issues related to simulation interaction, synchronisation,

and data exchange between federates. As part of this effort, necessary extensions to JADE and OPNET Modeler have been developed. The most notable of which include the addition of discrete-event capabilities for JADE, a generic means to represent agent applications within OPNET Modeler, and data exchange capabilities between the two federates. Is has also been shown that the additional extensions have linear computa- tional complexities, which suggests that they scale well.

Moreover, an agent-based backup relay supervision scheme has been developed for the IEEE 39-bus system. Its performance has been analysed for different application designs and communication technologies under different communication network scenar- ios. Application designs included different peer selection algorithms and communication strategies such as client-server communication and peer-to-peer communication. The impact of various elements pertaining to agent communication have been included such as different message sizes and transport protocols. Also, Quality of Service (QoS) strate- gies, background traffic, the impact of communication link outages as well as additional links have been considered. The obtained results show clearly that the communication infrastructure has a profound impact on the MAS performance and therefore it is critical to account for it during MAS development. Based on the results, clear recommenda- tions on viable technologies and designs have been made. The validated MAS code has successfully been deployed on a small scale which proves that the suggested method also supports the deployment and is not only a tool for validation through simulation.

It has also been shown that the computer memory and storage requirements for the validation of the supervision scheme was easily met with the computing equipment that is readily available today. Furthermore, the obtained simulation results suggest that the time overhead added to the individual federates by the additionally developed extensions is negligible.

In conclusion, a new method for MAS development and validation has been presented which provides a means to rapidly implement agent-based applications and accurately validate them for different communication technologies.

6.1 Summary of Contributions

During the course of this research several significant contributions have been made: • A new method has been suggested that enables the development and validation