This project deals with Intelligent Tutors aimed at training electrical networks Control Center operators in the tasks related with their dealing with serious incidents in the network, with improvements in the following areas:

1. The power system restoration problem is basically a planning problem. Taken individually none of the tasks to be performed during the restoration of a power system is fundamentally difficult. However, many constraints must be checked repeatedly as these tasks are performed, and that must be done under very stressing conditions. The current training programs are based on the use of electrical network simulators. Albeit being very useful to understand how an electrical network behaves, those tools don't provide user modeling, don't evaluate student's performance and are quite inflexible in curricula planning and training sessions preparation. On the other hand, Intelligent Tutors are flexible tools and when provided with an adequate User Model, they can adapt the training to the specific needs and characteristics of the trainee. One sub-goal of this project is the development of a user modeling component capable of enabling the Tutor to provide flexible guidance, adaptive curricula and didactic methods selection. It also allows to maintain the trainee's work evaluation.

2. The domain knowledge used by our previous attempts in this area was embodied in the knowledge base of an expert system (SPARSE) used in the diagnostic of incidents occurred in electrical networks. The knowledge it uses can be described as the domain knowledge and has several limitations when used in a tutoring environment: it doesn't contain specific knowledge neither about the student characteristics nor about didactic strategies or techniques and the way it is represented is not normally the most suited for that environment. Specific work in this area will be carried out.

3. A detailed analysis of the diagnostic process is mandatory to allow to correctly organize the tutoring process; we expect to be able to identify and characterize the different phases that constitute this process.

4. Different approaches to the student modeling component will be tested, using among others Bayesian networks and fuzzy logic techniques as well as investigating the best ways of representing and updating this knowledge.

5. The project pays special attention to the didactic aspects of the tutoring process, namely: * Selection and sequencing of the problems to be presented to the trainee based on classification techniques using Neural Networks. * Forms of curricula representation (e.g. relating incidents with underlying concepts). * Decision triggering and initiative in the tutoring process. * Progress evaluation and help support, investing efforts in error taxonomy and cataloguing.

6. The interaction with the student will be improved, using prediction tables and limited forms of natural language techniques.

7. A multi-agent system to mimic the interaction between the several operators involved in the process will be developed and the issue of the emotional behaviour modelling in critical situations will be addressed.