Intelligent Management Of Electrical Systems in Industries

INTROUCTION

Industrial plants have put continuous pressure on the advanced process automation. However, there has not been so much focus on the automation of the electricity distribution networks. Although, the uninterrupted electricity distribution is one basic requirement for the process. A disturbance in electricity supply causing the"downrun" of the process may cost huge amount of money. Thus the intelligent management of electricity distribution including, for example, preventive condition monitoring and on-line reliability analysis has a great importance.

Nowadays the above needs have aroused the increased interest in the electricity distribution automation of industrial plants. The automation of public electricity distribution has developed very rapidly in the past few years. Very promising results has been gained, for example, in decreasing outage times of customers. However, the same concept as such cannot be applied in the field of industrial electricity distribution, although the bases of automation systems are common. The infrastructures of different industry plants vary more from each other as compared to the public electricity distribution, which is more homogeneous domain. The automation devices, computer systems, and databases are not in the same level and the integration of them is more complicated.


Applications for supporting the public distribution network management


It was seen already in the end of 80's that the conventional automation system (i.e. SCADA) cannot solve all the problems regarding to network operation. On the other hand, the different computer systems (e.g. AM/FM/GIS) include vast amount of data which is useful in network operation. The operators had considerable heuristic knowledge to be utilized, too. Thus new tools for practical problems were called for, to which AI-based methods (e.g. object-oriented approach, rule-based technique, uncertainty modeling and fuzzy sets, hypertext technique, neural networks and genetic algorithms) offers new problem solving methods.

So far a computer system entity, called as a distribution management system (DMS), has been developed. The DMS is a part of an integrated environment composed of the SCADA, distribution automation (e.g. microprocessor-based protection relays), the network database (i.e. AM/FM/GIS), the geographical database, the customer database, and the automatic telephone answering machine system. The DMS includes many intelligent applications needed in network operation. Such applications are, for example, normal state-monitoring and optimization, real-time network calculations, short term load forecasting, switching planning, and fault management.

The core of the whole DMS is the dynamic object-oriented network model. The distribution network is modeled as dynamic objects which are generated based on the network data read from the network database. The network model includes the real-time state of the network (e.g. topology and loads). Different network operation tasks call for different kinds of problem solving methods. Various modules can operate interactively with each other through the network model, which works as a blackboard (e.g. the results of load flow calculations are stored in the network model, where they are available in all other modules for different purposes).

The present DMS is a Windows NT -program implemented by Visual C++. The prototyping meant the iteration loop of knowledge acquisition, modeling, implementation, and testing. Prototype versions were tested in a real environment from the very beginning. Thus the feedback on new inference models, external connections, and the user-interface was obtained at a very early stage. The aim of a real application in the technical sense was thus been achieved. The DMS entity was tested in the pilot company, Koillis-Satakunnan Sähkö Oy, having about 1000 distribution substations and 1400 km of 20 kV feeders. In the pilot company different versions of the fault location module have been used in the past years in over 300 real faults.

Most of the faults have been located with an accuracy of some hundred meters, while the distance of a fault from the feeding point has been from a few to tens of kilometers. The fault location system has been one reason for the reduced outage times of customers (i.e. about 50 % in the 8 past years) together with other automation.