This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.
Electric utilities have a need to keep a comprehensive and accurate inventory of their physical assets, both as a part of normal service provision (extending the network, undertaking maintenance, etc.) and as a part of their obligation to inform third parties about their facilities. Complexity of electrical distribution power system is a good reason for introducing new information technology - GIS (Geographic Information System) that carries out complex power system analyses (e.g., fault analysis, optimization of networks, load forecasting) in acceptable amount of time. By using modern GIS, in conjunction with his own in-house developed software, in less time and more accurately, the utility engineer is able to design and to analyze electrical distribution network. This paper presents the idea of the project CADDiNÂ© (Computer Aided Design of Distribution Network) currently under development at the Power Systems Department of the Faculty of Electrical Engineering, University of Zagreb.
Importance of Distribution Network in Energy Supply
One of the primary contribution to the advancements and improvements in man's life-style over the years has been the ability to use and control energy. Man's use of energy can be seen in everyday operations such as mechanical motion and the production of heat and light.
Large amounts of power are generated at power plants and sent to a network of high-voltage (400, 220 or 110 kV) transmission lines. These transmission lines supply power to medium voltage (e.g. 10 or 20 kV) distribution networks (distribution primary system), which supply power to still lower voltage (0.4 kV) distribution networks (distribution secondary system). Both distribution network lines supply power to customers directly. Thus, the total network is a complex grid of interconnected lines. This network has the function of transmitting power from the points of generation to the points of consumption.
The distribution system is particularly important to an electrical utility for two reasons: its proximity to the ultimate customer and its high investment cost. The objective of distribution system planning is to ensure that the growing demand for electricity, with growing rates and high load densities, can be satisfied in an optimum way, mainly to achieve minimum of total cost of the distribution system expansion. Therefore, the distribution system planner partitions the total distribution system planning problem into a set of subproblems that can be handled by using available, usually heuristic methods and techniques [T.Gonen, 1986].
The design of electrical distribution networks is an everyday task for electric utility engineers, specially in R&D department. Such design was carried out few years ago manually. This classical approach usually result in overdesign distribution system, which is now considered as a waste of capacity that can be used instead of investing in system expansion. Four years ago a PC program package (CADDiNÂ©) for optimal planning of distribution network was put in operation in Elektra - Zagreb (Electric Utility of City of Zagreb). It is a result of joint R&D of Power System Department of Faculty of Electrical Engineering and
[End Page 1857]
Elektra - Zagreb. Based on the experience or PC-CADDiNÂ©, at the end or 1992. the prototype of new project CADDiNÂ© was started conceptually organized as a part of the Geographic Information System.
The role of GIS in Distribution Networks
Database plays a central role in the operation of planning, where analysis programs form a part of the system supported by a database management system which stores, retrieves, and modifies various data on the distribution systems. The thing that distinguishes an electrical utility information system from an other information system - such as those used in banking, stock control, or payroll systems - is needed to record geographical information in the database. Electrical utility companies need two types of geographical information: details on the location of facilities, and information on the spatial interrelations between them. The integration of geographically referenced database, analytical tools and in-house developed software tools will allow the system to be designed more economically and to be operated much closer to its limits resulting in more efficient, low-cost power distribution systems. Additional benefits such as improved material management, inventory control, preventive maintenance and system performance can be accomplished in a systematic and cost-effective manner (Z.Sumic, et al, 1993). Before graphical workstations were developed, many electric utilities have built technical information systems based on relational database management systems (E.Jorum, et al, 1993.). Technical information system is designed to cover the requirements of power supply utilities considering network expansion and operation planning, maintenance management and system documentation. In advanced utilities all information systems are built around same RDBMS and constantly updated. Establishing links between these information systems and geographical information system is only in defining relationship between objects in the two systems. The problem that has risen is in a number of different information systems in the same utility (technical information system, customer information system, etc.) or even several overlapping technical information systems and some of these are not updated.
The objective of the distribution network design process can be divided into three independent parts. These parts are:
1. load growth of the geographical area served by substation;
2. determination of load magnitude and its geographic location;
3. customer load characteristics;
Design of secondary system (low voltage distribution network)
4. optimal substation allocation and transformer sizing;
5. secondary circuitry routing and sizing;
Design of primary system (medium voltage distribution network)
6. optimal substation allocation;
7. primary circuitry routing and sizing;
To reduce a problem complexity each part of the design process is divided in functional subproblems. Each of these subproblems can be then much easier to manage. Although only independent some parts of design process interact, i.e. placement of substation will influence secondary routing which in turn will influence primary routing. The number of possible design solutions that might satisfy a given set of spatial, technical and economic constraints is quite numerous. Multiple, interdependent goals and constraints make conventional procedural optimization methods inappropriate for distribution network design. Due to the complexity of the design process, heuristic methods and AI techniques must be applied to find "near optimal" [S.Krajcar, 1988] or "satisfying" solutions [Z.Sumic, 1993]. The main reason for this simplification is regarding work-force and computer time for finding optimal solution that in high percentage could not be applicable in real situation.
[End Page 1858]
GENERAL DESCRIPTION OF GEOGRAPHIC INFORMATION SYSTEM OF PROJECT CADDIN
Pilot-project CADDiN was started at the beginning of 1993 as a research project inside the main research project "Research and Development of Electric Power System" supported by the Ministry of Science and Technology of the Republic of Croatia. The development of optimization and design procedures of electric distribution network is a parallel process with building database by Cadastral Office of the City of Zagreb, and therefore some other available examples of basic map databases are used for research purposes (see Figure 1). The strategy employed emphasized only the data composed of basic map databases for technical applications (scales of 1:500 to 1:5000).
There is no unique definition for Geographic Information System (GIS) but a commonly accepted one is that it is a system with computer hardware and software functions for the spatial data input, storage, analysis, and output [T. Bernhardsen 1992]. Many textbook definitions go further and identify analysis as the one activity which differentiates GIS from other computer-based systems for handling geographic data, such as automated cartography.
[End Page 1859]
Modern GIS, stores information on the geometry, attributes and topology of geographic features in one relational database management system. SYSTEM 9 used in the pilot-project CADDiN is a feature-oriented GIS which organizes geography-related information into a topology-structured, object-oriented, relational database system.
A project is the highest level of data organization of GIS used in CADDiN [Computervision, 1992]. It represents the entire database that has been set up for a particular geographic area - for example, a town, a municipality, or a service district. It comprises two components: a data store that contains all the geographic and attribute data relating to features; and a database definition that specifies the structure of the project through feature classes and themes.
Theme definition determines which features and attributes are to be used and the ways in which are to be displayed. Independently stored geometry of a feature, and its graphic representation enables position and representational data to be changed without reference to each other. The link between the geometry and the representation is provided by the theme. It comprises a list of feature classes, feature class attributes, and a link to a separate list of graphic transforms.
An important safety aspect of used GIS is that it does not allow users to make changes to the database at project level. A user may only query it. The database is created and updated by means of the next lower level of data structure: the partition. This is a copied, working subset, or portion of a project. It is at this level that a user interacts with the system to enter, edit, update and manipulate data. Partitions are extracted from a project based on the type of work to be done and the data that will be required to perform that work. When editing is completed, the partition is merged into the project database, effecting the update. Partitions are created by means of a partition definition that describes the spatial extent, the contents, and the representation. The system uses the partition definition to extract the required geometric and attribute data and then allocates them into the required partition. The merit of the partition structure is that it allows different departments within an organization to work safely on the data from the same project.
[End Page 1860]
All geometric features in the data model are built up from geometric primitives, referred to as nodes, lines, surfaces and spaghetti. A node is stored as a set of X, Y, and optionally Z coordinates in 3D database, and might be used to represent e.g. transformers, switchgears, MV - LV buses, etc. A line primitive is a geometric element defined by two end-nodes (allowing intermediate points), and might be used to describe transmission lines, cables, etc. A surface consists of one or more line segments that together form a closed polygon. A forest, lakes, parks, a portion of network, or area covered by a lot of buildings could be described by this kind of polygon. Spaghetti enables to model features where no topological structure is required. Nodes are the only geometric primitives that have coordinate information directly associated with them. Lines are not defined in terms of geographic coordinates, but by pointers to their topological nodes. Surfaces are defined by pointers to the lines surrounding the surface. All these pointers are created and maintained automatically.
Geographic objects are stored as collections of nodes, lines, surfaces or spaghetti, but they can be referred to as geometric primitives as well as some group of objects which can be identified and named in the real world - 'roads'. 'cables', 'transformers', buildings', and so on. These categories are represented by `feature classes'', and the individual instances of geographic objects as `features'. Such features at last consist of one or more geometric primitives. All features within a particular feature class will have the same topological structure, and the same set of attributes.
Feature classes could be also identified as objects in groupings of related objects that may be established on the basis of location, spatial relationships or common attributes. These logical groupings of features are called complex features. They are defined as features that contain other features. All complex features of particular type, comprise a complex feature class. A useful application of complex feature classification would be in forming logical groupings such as MV bus, transformer, LV bus, protection devices into 'substation'. Complex features can also have attributes associated with them (for example name, number). It would eliminate duplicating of feature attributes which properly relate to the substation. Definition of complex feature is not restricted to include only simple features as constituent components. For example, 'distribution network' could be defined as a complex feature containing a number of 'substations', 'cables', which are themselves complex features.
A strength of this approach is that it can be used to minimize the level of data redundancy of both attribute and geometric information. Users interact with the database via an object handler, and they are assisted in that interaction by a structured query language that incorporates extended spatial and reference operators.
Behind analytical tools available inside GIS environment, a set of standalone functions is available from UNIX shell. This set of functions is called Application Tool Box (ATB). ATB offers an environment in which data can be managed directly, without first having to extract meaning from map representations of those data. Under this approach a user can develop analytical models according to specific requirements by integration of ATB functions, in-house developed software (C and FORTRAN programs) and shell programming. To speed up complex analysis by Development Libraries of ATB new processing functions of ATB could be developed. Applications of project CADDiN are developing by ATB functions in conjunction with C and Corn shell programs.
ATB data management and viewing comprise processing functions, dataflow management and graphics viewing system. Processing functions perform the actual analysis operations on sets of data called data flows, each of which corresponds to a relational table in the database. All manipulation of data flows takes place in a special temporary work area called a clipboard. Processing functions involve the following operations: information management (i.e. selecting information from database and placing it into a dataflow, communicating with external software packages), attribute processing (i.e. generating values for attributes based on classification rules or formula), geometry processing (spatial functions - union, adjacent, etc.) and arithmetic processing (i.e. calculating the area of surface entities, or length of linear entities). Dataflow management is used to create, display and delete data flows and views. Graphic viewing system allows user to see the intermediate or final results and generate a plot of those results.
Compatible to ATB functions are standalone functions of Network Trace Analysis module. By those functions network tracing can be carried out using the information on network connectivity and component characteristics that are already stored in database. Special function is used for network generation that is stored as dataflow on the clipboard. On this dataflow several networks tracing functions can be performed
[End Page 1861]
(path optimization, range finding, path finding) or can be used by external software. As a result of that analysis a dataflow is produced on the clipboard. Original and resultant networks can be queried simultaneously. The user can keep or delete resulting data flow on the clipboard or retrieved in database.
OPTIMIZATION OF DISTRIBUTION NETWORKS IN GIS
Optimal Location of TS x/0.4 kV in Secondary Distribution Network
The procedure for finding optimal configuration of secondary system consists of two possible optimization steps:
optimization of new area secondary system and
optimal connection of the particular customer(s) to existing secondary system.
Regarding urbanistic plans, ecological and esthetic constraints as well as previous load growth analyses possible locations of substations are known in advance. These assumptions make planning of secondary system more simple because only routing process must be applied for several locations of substations and fixed locations of customers.
The first step of routing process begins by connecting customer to the nearest routing corridor. After that procedure, the secondary system network is generated by network module. On this network "any path analysis" is applied and as results of analysis there are all possible connections between substation and customers. These results are used as input for external, CADDiN module of optimization of radial structured networks. During this process of optimization the set of rules is used to satisfy standard practices employed by designers. The optimized network is then saved on clipboard in dataflow and can be graphically viewed. The cost for the secondary system is mainly the capital investment cost consisting of cable laying cost and cost of cables. For each location of substation optimization process must be repeated. Solution with minimal investment costs and satisfactory technical constraints is the best regarding secondary network. All solutions that are technically satisfied must be taken into account during the primary network optimization. It is necessary because the local optimum of secondary system does not imply the optimum of primary system, and global optimum of distribution network.
The optimal connection of the particular customer to existing secondary system must fulfill next two technical as well as economical constraints:
the shortest possible length of connection due to voltage drop that may be permitted;
reserve in load capacity of substation due to customer load.
The new customer must be connected to the nearest neighbor customer satisfying previously mentioned constraints. The few nearest customers are found in a buffer zone with new customer as a center of this zone. The shortest path between new customer and possible connection node is found in two steps: both nodes are connected to the nearest routing corridor, and after that by GIS network function "find best path analysis" shortest path between nodes is found.
Optimal Structuring of Predefined Primary Distribution Network Configuration
Due to the load characteristics, requested availability and quality of energy supply two main configurations of secondary system are used in optimal planning There is a ring structure (starting and ending node is the same HV/MV substation and routing nodes are MV/LV substations) and a link structure (starting node is one HV/MV substation routing nodes are MV/LV substations and ending node is other HV/MV substation). Regarding the usage of GIS technology the optimization procedure of these two network configurations is very similar. In optimization process three different problems are considered:
optimization of the new primary system;
reconfiguration of the existing primary system regarding predefined structure, and
reinforcement of the existing primary system with defined structure by installing additional capacity in demand nodes or including the new MV/LV substation in the network.
The first problem is similar to the problems in optimization of secondary system. There must be known all possible connections and distances between HV/LV substation (source node) and MV/LV substations
[End Page 1862]
(demand nodes) as well as themselves. Therefore, all network nodes must be connected to the nearest routing corridor. By "any path analysis" and heuristic algorithms (presently genetic algorithms are tested) initial solution or "zero-iteration" is generated. After that by the union of GIS network function "find best path analysis" and other heuristic methods optimal solution is found.
The second problem is more complicated than the first one because existing connections in network must be considered in optimization procedure. Otherwise, same procedures are used as in the first problem. Example of this optimization procedure can be shown in the Figure 3.
In the third problem, optimization procedure is similar to the procedure of adding the new customer to the second system. Slight differences are in a way of connecting new substation to the existing network. In
[End Page 1863]
the primary system, regarding the constraint of reliability of supply of energy to the customer, each MV/LV substation must have a possibility to be supplied from two sides. Therefore, the nearest existing cable between two substations must be found for the connection of the new station, or the nearest routing corridor by which the new station could be connected to the nearest substations that are found in a buffer zone around it. When a better type of connection is found, solution is tested on several technical constraints (voltage drop, cable and route load, investment costs, etc.).
Load forecasting of TS x/0.4 kV
Small area or spatial, forecasting is the prediction of both the amounts and locations of future electric load growth in a manner suitable for distribution planning which really means with geographic resolution adequate for planning a new distribution network or extensions to the existing one. The procedure is based on dividing a utility service area into a number of sufficiently "small areas" and projecting the future load in each one. This is usually accomplished by dividing a utility service area into either a grid of uniformly sized rectangular "cells", or into "equipment oriented" areas corresponding to feeder or substation areas (H.L. Willis, 1983,1992).
Methods for computerized small area load forecasting, regarding their data requirements and analysis methods, fall into three categories:
Essentially these methods analyze past and present load growth to identify trends, patterns, or information about the process of load growth that is then used to project future load growth.
Trending methods require minimal data (they work only with historical load data, usually annual peak load) and computer resources, and are relatively straightforward in use. Because of their simplicity and generally the lowest expenses, they were the most widely used techniques in the past.
Multivariate methods require considerably more data (historical loads, geographic and demographic data on customers and usage) and much more extensive computer resources, but in return they generally provide more accurate forecasts.
Simulation methods in addition to historical loads require extensive and comprehensive data that include land use type, geographic and demographic data on a small area basis, transportation and other diverse factors that may affect load growth. They also require considerable computer resources and work-force. On the other hand they offer advantages in accuracy and analysis of load growth under changing conditions. Because of their complexity and requirements simulation models have been beyond the scope of many electric utilities.
So far one can see that the nature of small area forecasting requires heavy use of computerized analyses and manipulation of large quantity of data.
With its possibilities GIS is an excellent mean for developing and applying simulation forecast models. Of course, there is no limitation to use GIS for trending methods, at least for some very fast qualitative review, or for short range (less than five years ahead) predictions.
A service zone of a substation may be defined as a complex feature which comprises parcels, buildings on those parcels, electrical connections for every building or customer, existing interconnections between customers hookups and associated substation etc. Parcels, buildings and streets are modeled as polygons, and cadastral lot code is attached to them as one of the attributes. Statistical and census districts based on approximately equal number of inhabitants and cadastral districts are polygons, too. Second very important information is address, modeled as complex feature class comprising a street name and number. Polygonal analysis and polygon processing, which is possible in GIS, and address as a common link enables the planner to determine a substation service zone and calculate its area. Via features' attributes all necessary customers' data (annual electricity consumption, annual peak loads, type of customers, some special requests and interfering factors, etc.) are obtainable. In that way it is possible to track amounts and sort of energy used by individual customer, or substation service area or some other region. Upon these information load densities (kWh/mÂ²) or kWh sales per customer can be computed.
Procedure with built-in clustering algorithm detects groups (classes, clusters) of customers with similar past energy consumption behavior. For distribution load forecasting K-means algorithm [Hartigan, 1986] is
[End Page 1864]
recommended, with a minimum of 6-year load history [H.L. Willis, 1983]. The K-means algorithm searches for a partition, that is, a set of clusters that minimizes the "total difference" between small areas and their assigned clusters (the error of the partition). It works by moving small areas from one cluster to another. The search ends when no such movements of small areas reduce the error value.
This paper presents the concept of the pilot project CADDiN for optimization of electric distribution networks based on GIS technology. The architecture of CADDiN consists of the heuristic methods implemented within GIS and procedural programs. In such a hybrid environment, the GIS principal task is to model "real world", perform spatial analyses and ensure the high accuracy of optimization procedures. The first results obtained by the prototype database and developed procedures encourage that concepts and ideas established in this paper can be applied on the real problems that exist in the distribution system planning.