TY - GEN

T1 - Calibration of C-Town network using harmony search algorithm

AU - Kim, Joong Hoon

AU - Chung, Gunhui

AU - Yoo, Do Guen

PY - 2012

Y1 - 2012

N2 - This study adopts the Harmony Search (HS) meta-heuristic algorithm (Geem et al., 2001; Kim et al., 2001) to calibrate a hydraulic simulation model (EPANET) for C-Town. The algorithm conceptualizes a musical process of searching for a perfect state of harmony (optimal solution) and allows a random search without initial values, thus removing the necessity for information on derivatives. The calibration procedure is implemented in a manner that the sum of errors between two data sets: the measured SCADA data and the values derived from the HS algorithm. The calibration process is implemented into three stages - preprocess, rough tuning, and fine tuning. In the preprocess stage the original SCADA data and pump control rules are modified to complement missing data and operation rules. In the rough and fine tuning stages the C-town network is divided into five sub-networks based on supply zones. Pipe roughness factors in each sub-network are calibrated by pipe diameters which are divided into five groups. Therefore, total number of decision variables in each sub-network is 168 nodal water demand factors and 5 pipe roughness coefficients. The entire C-town network is also calibrated using each already calibrated sub-network as the initial populations. It is found that the simple network in the downstream end can be calibrated separately; thus, the network division does accelerate the convergence speed. Several parameters of the HS algorithm such as HMCR and PAR can be adjusted to find better calibration results

AB - This study adopts the Harmony Search (HS) meta-heuristic algorithm (Geem et al., 2001; Kim et al., 2001) to calibrate a hydraulic simulation model (EPANET) for C-Town. The algorithm conceptualizes a musical process of searching for a perfect state of harmony (optimal solution) and allows a random search without initial values, thus removing the necessity for information on derivatives. The calibration procedure is implemented in a manner that the sum of errors between two data sets: the measured SCADA data and the values derived from the HS algorithm. The calibration process is implemented into three stages - preprocess, rough tuning, and fine tuning. In the preprocess stage the original SCADA data and pump control rules are modified to complement missing data and operation rules. In the rough and fine tuning stages the C-town network is divided into five sub-networks based on supply zones. Pipe roughness factors in each sub-network are calibrated by pipe diameters which are divided into five groups. Therefore, total number of decision variables in each sub-network is 168 nodal water demand factors and 5 pipe roughness coefficients. The entire C-town network is also calibrated using each already calibrated sub-network as the initial populations. It is found that the simple network in the downstream end can be calibrated separately; thus, the network division does accelerate the convergence speed. Several parameters of the HS algorithm such as HMCR and PAR can be adjusted to find better calibration results

KW - Harmony Search

KW - Model Calibration

KW - Water Distribution System

UR - http://www.scopus.com/inward/record.url?scp=84862924238&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84862924238&partnerID=8YFLogxK

U2 - 10.1061/41203(425)143

DO - 10.1061/41203(425)143

M3 - Conference contribution

AN - SCOPUS:84862924238

SN - 9780784412039

T3 - Water Distribution Systems Analysis 2010 - Proceedings of the 12th International Conference, WDSA 2010

SP - 1610

EP - 1628

BT - Water Distribution Systems Analysis 2010 - Proceedings of the 12th International Conference, WDSA 2010

T2 - 12th Annual International Conference on Water Distribution Systems Analysis 2010, WDSA 2010

Y2 - 12 September 2010 through 15 September 2010

ER -