MIKE NET AND RELNET: WIC APPROAC TO RELIABILITY ANALYSIS IS BETTER? Alexandr Andrianov Water and Environmental Engineering, Department of Chemical Engineering, Lund University P.O. Box 124, SE-221 00 Lund, Sweden ABSTRACT A water distribution system (WDS) is a very important part of every municipality. Therefore it is very important that the WDS performs well. owever, in spite of how well the system performs at present, there is always a risk of failure occurrence. That is why it is very important to be able to estimate the Reliability of WDS. Different computer programs for the reliability assessment were developed recently. Description and comparative analysis of two computer programs, used for the Reliability analysis, are presented in this paper. Analysis showed that even though both programs evaluate the system reliability they have different approaches to the reliability assessment. It is important to be aware of the computational methods used by the programs to be able to understand software behavior in different situations and to evaluate the results of the analysis. KEYWORDS Water distribution system, reliability analysis, software, RelNET, MIKE NET INTRODUCTION There is currently no universally accepted definition of reliability of WDS. owever reliability is usually defined as the probability that a system performs its mission within specified limits for a given period of time in a specified environment. Reliability models to compute system reliability have been developed since 1980s. These models allow a modeler to determine the reliability of a system and account for such factors as the probability and duration of pipe and pump failure, the uncertainty in demands, and the variability in the deterioration of pipes. Several different methods of reliability assessment exist at present. Cut-set method is one of the most popular methods because it is efficient and easily programmed on computers. The minimum cut-set is defined as a set of system components which, when failed, cause failure of the system (Mays L. W., 2002). owever, when any one component of the set has not failed, it does not cause system failure. Pipe failure can be isolated from the rest of the system by closing a pipe or pipe combination. This determines minimum cut-set for the pipe. A minimum cut-set for a node is one that causes reduction of the hydraulic availability at the node. Minimum cut-set for the system reduces hydraulic availability for any node in the system. The model also checks if the failure causes the part of the system to be disconnected from the water source. This disconnection results in failure of all disconnected nodes.
Two different computer programs, which are used for reliability analysis of WDS, are described and compared in this article: RelNET reliability analysis program and Reliability Analysis module of MIKE NET system analysis software. DESCRIPTION OF TE RELNET COMPUTER PROGRAM RelNET software calculates the impact of each pipe link failure on the total network reliability. The output from the program is ydraulic Critical Index (CI) the impact of each pipe link on the total network reliability (CARE-W, 2003; CARE-W, 2004). RelNET uses cut-set method in its computational algorithm. owever, there are two distinctions in the RelNET methodology and the methodology described in this chapter: RelNET simulates closure of only one pipe, not of pipe groups. RelNET uses minimum and required pressures (P min and P req ) in calculations and considers availability equal to zero when pressure is below minimum pressure. If actual pressure is between minimum and required values, the hydraulic availability is calculated according to some algorithm and can vary in limits (0-1) (Mays L. W., 2002). RelNET calculation algorithm consists of three main parts: Link Processing Pressure Processing CI Processing During Link Processing the program discards one link of a network in each step of the analysis and creates a map of the network. The continuity of the network is then tested. If the network is not continuous, the nodes and links that are not connected to a water source are discarded from the network diagram. Then the input file for the hydraulic analysis is created. A number of files equal to the number of links in the network is generated in this step. Pressure Processing creates output files with the results of the hydraulic analysis. Each output file contains the node ID and calculated pressure. CI Processing calculates the impact of each link on the total network reliability. Two main parameters for CI calculation are required ( req ) and minimum ( min ) pressures (Valkovic Pavel, 2002). Algorithm of CI calculation: Calculation of actual pressure and demand in each node of the WDS in the original state (none of the pipe links is discarded). Results are act (actual demand), act (actual pressure) and tot (total network demand). Calculation of pressure in each node ( ) and demand ( ) when one link of the network is discarded. CI of the discarded link is calculated from the volume of undelivered water in the entire network. The amount of undelivered water in each node depends on the calculated pressure value ( ). If < min then = 0 If min < < req then the amount of delivered water in the node is reduced and is calculated according to the following equation (1):
= act (1) act If > req then = act Delivered water is calculated by this method for each node of the network. CI is calculated according to the following formula (2): CI tot = (2) tot tot total demand of the network (all links are opened) Σ sum of all the node demands when one link is closed A higher value of CI means a higher impact of the discarded link on the total network reliability. If the sum of = 0 then no demand is satisfied in all nodes of the network and CI = 1. If sum of = tot, CI = 0 then demand is fully satisfied at the required pressure (Valkovic Pavel, 2002). This CI is calculated for each link in the network and shows its importance. In case of branched systems CI has usually higher values for links situated in the beginning of branches, because in case of link closure the whole branch is disconnected from the source of water. In case of looped systems the links presenting the big transmission mains usually have higher values of CI. DESCRIPTION OF TE MIKE NET RELIABILITY ANALYSIS MODULE MIKE NET Reliability Analysis Module can evaluate the reliability and risk of the water distribution network. The reliability analysis module was developed by DI Water & Environment jointly with Professor Ruggero Ermini, Università della Basilicata, Dipartimento di Ingegneria e Fisica dell'ambiente, Italy. Reliability Analysis module consist of two parts: Demand Adjusted EPANET Analysis Demand Availability Demand Adjusted Analysis This part of the Reliability Analysis Module adjusts demands in the nodes according to available pressure in these nodes. The computational algorithm is a modification of classical hydraulic model and it is not a strictly demand driven analysis. The demand in the node depends on available pressure in the node and if this pressure is less than some limiting value ( max,i ) consumer demand is not satisfied fully and it is recalculated according to the algorithm: If i < min,i then i = 0 If min,i < i < max,i then the amount of delivered water in the node is reduced and is calculated according to the following equation (3):
i i min, i = max, i (3) max, i min, i Where i pressure at a node i min,i minimum required pressure at a node i max,i maximum pressure at a node i (defined as the pressure above the house height) i demand at a node i max,i user specified (requested) demand at a node i If i > max,i then i = max,i Maximum pressure at a node must be defined in meters of water column above the house height. This method to define the pressure makes it possible to perform demand adjustment in the networks with several pressure zones or several districts with different average house heights. Minimum pressure at a node usually doesn t depend on house height and can be expressed in absolute values (m 2 O). This value usually is defined by minimum acceptable pressure at a fire hydrant. Demand Availability This part of the Reliability Analysis module doesn t make any corrections in node demands but only calculates system reliability and amount of undelivered water due to pressure drop below the minimum pressure or pressure increase above the maximum pressure. The performance is evaluated with respect to the time period during which threshold values ( max and min ) are not satisfied. For each node the threshold values are defined and the failure frequency evaluated. The reliability is calculated according to equation (4): R = 1 (4) i f i Where R i reliability at node f i failure frequency at node ti i 0 fi = 1 (5) T Where t i total failure time T total time period of simulation The method of calculation of total failure time and total time period is presented in Fig. 1.
Fig. 1 Total time period and total failure time in reliability analysis In Fig. 1 X max and X min are maximum and minimum limits for the pressure in the system. Time periods t 1 and t 2 are periods when the real pressure exceeds the limiting values and t i =t 1 +t 2 is total failure time (Ingeduld P., 2003). In this method of calculation it is considered that water is undelivered to customer in two cases: pressure at a node is below the minimum or above the maximum. Pressure increase above the upper limit doesn t cause a demand drop; it only increases the risk of failure in the node. Probably it was the only way to express the situation when any of the pressure threshold values are not satisfied in terms of undelivered water. COMPARATIVE ANALYSIS OF RELNET SOFTWARE AND MIKE NET RELIABILITY ANALYSIS MODULE Both RelNET and MIKE NET adjust the network demands in the process of reliability analysis. The structures of these two algorithms are similar: If pressure at a node is below the minimum pressure no demand is satisfied. If pressure at a node is above the required pressure customer demand is fully satisfied. If pressure at a node is between these two values customer demand is adjusted according to some method; these methods are different for RelNET and MIKE NET. RelNET and MIKE NET computer programs use different symbols to present the same values: Required pressure in RelNET is req and in MIKE NET is max,i Actual demand at a node in RelNET is act and in MIKE NET is max,i Adjusted demand in RelNET is and in MIKE NET is i It was decided to use symbols used by RelNET in this comparative analysis to avoid confusion and because they better reflect the physical sense of presented parameters. To make a comparison of these two methods, a - diagram of demand change due to pressure decrease is created. In Sweden minimum and maximum acceptable pressures at a node are 15 and 70m 2 O; pressure above the highest tap in the house should be at least 15m 2 O, that is why required pressure must be around 30m 2 O (Svenska Vatten och Avloppsverksföreningen, 2001). The following values are taken for calculations: req = 30m (in case of MIKE NET it is assumed that house height is zero) min. = 15m
act. = 50m Equations for demand adjustment are the following: RelNET = act (6) act MIKE NET req min = req (7) min These two equations give two different curves for - diagrams (Fig. 2)., m 2 O act =50m A ( act ; act ) Rel NET MIKE NET req =30m B ( req ; req ) min =15m 0 0,55* req 0,77* req act ( req ) Fig. 2 Node demand change due to the pressure drop calculated by RelNET and MIKE NET It is seen from the diagram in Fig. 2 that the MIKE NET method of demand adjustment gives more smooth demand change when pressure at a node crosses required and minimum pressure threshold values. In case of RelNET, demand changes significantly when pressure at a node crosses these threshold values, for example: If is just above 15m, = 0,55* act. But if = 15m then = 0. That means that difference is about 55% of actual (required) demand. If is just below 30m, = 0,77* act. But if = 30m then = act. That means that difference is about 23% of actual (required) demand.
CONCLUSIONS The following conclusions can be drawn from the comparative analysis of two computer programs: Both computer programs RelNET and MIKE NET evaluate the system reliability; however, they have different approaches to the reliability assessment. It is extremely important to be aware of the computational methods used by the programs to be able to understand the software behavior in different situations and to evaluate the results of the analysis. Both programs adjust the network demands in the process of reliability analysis and use similar computational algorithms for this purpose. owever they use different equations in the process of calculation. MIKE NET points out problematic nodes in the network but RelNET highlights the most important of the weakest links in the network. RelNET discards one link of a network in each step of the analysis and creates a map of the network. The program calculates the impact of each link on the total network reliability. MIKE NET Reliability Analysis module doesn t simulate closure of any links and makes analysis for the entire network. It is difficult to use RelNET for analysis of a WDS with several pressure zones because it is possible to define only one set of parameters for req and min. That means that it is necessary to divide the network into sub models and make several calculations for each particular pressure zone. owever CI will be difficult to estimate in this case. In MIKE NET this problem is solved by defining required pressure in terms of meters above the house height. RelNET doesn t use time in calculations. Calculations are made for steady state analysis or for the initial conditions of the network for extended period simulations. Consequently it is necessary to make analysis in RelNET for the peak water demand in order to obtain reasonable results. In case of a branched system RelNET computer program calculates how much water is undelivered due to the fact that some users are disconnected from the source of water. In case of a looped system the program calculates how much water is undelivered due to the pressure decrease in the WDS. In case of a mixed system, due to both of the factors mentioned. RelNET calculates undelivered water when pressure at a node is lower than required pressure. MIKE NET calculates undelivered water when pressure in the node is lower than minimum of higher than maximum. REFERENCES CARE-W. 2003. Tests and validation of Technical Tools. Cemagref, INSA Lyon, NTNU, Brno University. Report. CARE-W. 2004. Guidelines for the use of Technical Tools. Cemagref, SINTEF, INSA Lyon. Report.
Ingeduld P. 2003. Modelling of Water Distribution Systems with MIKE NET. DI Water & Environment. Mays L.W. 2002. Urban Water Supply andbook. McGraw-ill, USA. Svenska Vatten och Avloppsverksföreningen. 2001. Allmänna vattenledningsnät. Anvisningar för utformning, förnyelse och beräkning. Stockholm, Sverige. Valkovic Pavel. 2002. RelNET elp file. Brno University of Technology, Czech Republic.