Article Reference
Zoong Woo Geem, Joong Hoon Kim, G V Loganathan, A New heuristic Optimization Algorithm : Harmony Search
Introduction
The paper tries to focus on the new heuristic algorithm called “Harmony Search” (HS) which is explained using three examples – TSP, academic optimization problem and a least cost pipe network design problem. Authors explain the drawbacks of using the normal optimization methods i.e LP, NLP and DP. In LP there are considerable losses associated as the real world non linear model is being expressed as a linear model; In DP increased variable increase the evaluation of the recursive functions; in NLP if the functions are not differentiable then finding optimum is not possible and attention is required to select the initial values in order to guarantee global optima instead of local optima. To overcome the above mentioned drawbacks, heuristic method of optimization is developed which results in near optimum solution within reasonable computation time and within reasonable use of memory. Since 1970 there have been many heuristic algorithms which have been developed combining rules and randomness mimicking the natural processes. The techniques are:
• Simulated annealing (SA) – It is based on the analogy of physical annealing process. In physical annealing as the materials temperature increases the molecules mobility increases and gradually these molecules form the crystalline structure i.e. the optimal solution.
• Tabu search (TS) – it explores the search space of feasible solutions by sequence of moves. To escape from the local optima and to prevent cycling some moves are classified as tabu.
• Evolutionary algorithms (EA) – it is based on the principle of evolution. Consists of four heuristic algorithms- genetic algorithm, evolutionary strategies, evolutionary programming and genetic programming.
• Genetic Algorithm – based on the natural selection and the mechanism of population genetics. It has three operators – reproduction, crossover and mutation. Main advantage of GA when compared to other optimization techniques or heuristic techniques is that GA evaluates different solutions simultaneously unlike the other methods which evaluate on solution at a time.
• Evolutionary strategies
• Evolutionary programming
• Genetic programming
These simulation based heuristic search techniques have powerful search abilities which overcome the drawbacks of the traditional optimization techniques. The paper proposes a new heuristic search technique which produces better solution in less number of iterations than the other techniques.
New Heuristic technique (Harmony Search)
The steps to be followed in this new technique are:
• Step 1 – Initialize a Harmony memory (HM)
• Step 2 – Improvise a new harmony from HM
• Step 3 – If the new harmony is better than the minimum harmony in HM then include the new harmony in the HM and exclude the minimum harmony from HM
• Step 4 – if the stopping criteria are not met then repeat the Step 2.
To improve the solutions and escaping from the local optima there two methods which are introduced, they are:
• Harmony memory considering rate (HMCR): indicates the probability of the algorithm which selects the variable from the harmony memory onto the next step
• Pitch adjusting rate (PAR): shifting the values to nearby neighboring values in the specified range which is called as PAR. PAR 10% means the new value is selected in the probability of 10% (5% of higher value or lower value with 5%)
Major difference between HS and GA is that HS makes a new vector from the existing vectors (i.e from all harmonies in the Harmony memory) where as GA makes new vector only from the two existing vectors (parent).
Examples
1. Travel Salesman Problem – Harmony memory consists of the existing tour lengths, if the new tour length is shorter than the existing HM tour lengths then the new tour length is replaced in the HM. Different HMCR and PAR parameters are considered and the iterations are performed to avoid the local optima solution. For faster convergence two new operators have been introduced – Neighboring city (i.e salesman visiting the closest neighboring city) and city inverting (i.e inverting the city if the former is shorter than the later city).
2. Simple constrained minimization problem – Initially decision variable x1 and x2 have been selected randomly based on the given range later the range has been narrowed to the values in the HM. This problem was solved using generalized reduced gradient (GRG), EP and GA, HS has evolved much better solutions than the other methods.
3. Design of pipe network for water supply – The problem has 32 nodes, 32 links and 3 loops, no pumping station is considered. Minimum pressure requirement at all the nodes is specified. This problem was solved using Nonlinear programming gradient and GA. HS gave a minimum least cost of the pipe network when compared to the other heuristic methods.
Conclusions and Discussion
HS is advantageous than the other methods
- as it makes a new set of vectors based on all existing vectors instead of just two vectors (parent),
- it does not require initial values of the decision variables
- It outperformed the other heuristic techniques (GA, EP etc)
- Faster convergence
Friday, July 31, 2009
Discussion # 5 (22 July 2009)
Article Reference
Q. Xiao, E G McPherson, J R Simpson and S L Ustin, Hydrologic processes at the urban residential scale, October 2006
Introduction
The paper focuses on building up a physical numerical model to understand hydrologic processes in the urban scale and its interaction with different BMP’s. The model was applied to a single family residential lot in Los Angeles, CA. Model was calibrated and validated by the data collected for over 2 yrs. They have modeled rain gutters, driveway interceptors and lawn retention basin. It was observed that there was about 65% reduction in storm water by using driveway interceptor. But using the rainwater gutters the annual irrigation demand was provided by 9%. They modeled trying to replace the existing loam soil with clay soil and it was observed that annual runoff discharge to the street was increased by 63%. In this study they have considered the case study of Los Angeles which has the polluted runoff being flown into local watershed and thereby into the coastal recreation areas. LA experience sever summer deficit of water demand where they import almost 85% of the water to meet their demands. Author focuses on how the policy makes are considering the implementation of the decentralized approach in the urban watershed i.e creating a mini reservoir system which retains runoff and reducing the irrigation demand but they lack the quantitative data on the effectiveness of the usage of different BMP’s. Most of the urban hydrologic models focus on the water balance and not on the effects on the BMP’s on the hydrologic processes which are dependent on the land use.
Objective of the study:
Develop, calibrate and validate the physical model at the residential scale for simulating the runoff for both long term events and storm term storm events. The study is carried out to see the effectiveness of BMP and simulating the effects of retrofitting the control site with the BMP’s.
Methodology:
• Site: residential housing unit
• BMP’s considered: cistern, swale, rain gutter, lawn retention basin, driveway interceptor and drywell
• Designed for 50 yr storm event
• Runoff from rooftops flowed into cistern and lawn retention basin
• Runoff from driveway flowed into the interceptor and further into the drywell
The data was collected from Jan 2001 to Jan 2003 and the model was calibrated and modeled based on the measurements of land covers, hydrologic processes and micro climate data. The runoff from treatment Site Street was collected in the measurement system and then through the water outlet drainages where as the runoff from the control site was drained directly into the drive way and thereby into the street. Each land cover was considered as a individual modeling unit. The model was studied for both long term simulation and short term events. For the long term simulation the focus of study was taking into consideration the annual irrigation use and runoff to the street and for short term simulation the focus was on the storm runoff to the street and the storage of the BMP.
• Precipitation data taken as the input of the rainfall data
• Two strategies of irrigation (first using the annual ET and second one was based on the soil and water deficit) were used in the model to determine the timing and the amount of water to be applied.
• Interception losses are also taking into consideration in the model
• Surface flow is calculated using the manning’s equation
• Infiltration is estimated using Green Ampt’s equation
• Percolation
• Evapo-transpiration
• Lateral subsurface flow
• Cistern storage is based on the total amount of water entering the cistern and the cistern restricted area
Numerical simulation focused on the landscape irrigation, storm runoff, ET and percolationof runoff into the deep layers, sensitivity analysis of the landscape retrofit was simulated using the date of 2001 and the 50 yr storm event (1988). The purpose of the sensitivity analysis was to distinguish which BMP was more effective with respect to runoff reduction and conservation of landscape irrigation water use. For the control site there were not BMP retrofitted which was considered as the base condition and the treatment site was varied with 6 different BMP’s. Other type of sensitivity analysis was carried by replacing with the clay soil instead of sandy soil.
Results:
Both treatment site and control site had about 50% impervious cover. Field measurement revealed that both cistern and retention basin were more effective in conserving municipal water supplies and reducing the total storm runoff. Surface runoff was effectively reduced by the drive way interceptors and lawn retention basin.
Discussion:
It was observed that all the BMP’s installed in the treatment site were effective in reducing the storm runoff and reducing the irrigation demand from the municipal water demand. Using this strategies enabled the improvement of urban eco system. BMP has reduced the maximum runoff flow to the streets and the maximum annual evaporation and increased the deep percolation. Increased percolation increases the downstream base flow and groundwater recharge. For runoff reduction driveway interceptor and drywell combination was the best but this BMP carries the risk of carrying pollutants from the surface runoff into the ground water. Cistern provided additional irrigation demand supply but for more effective use of cistern it should be appropriately designed for the catchment areas. Few of the sites will not follow the sub surface balance which focuses on the importance of more detailed model with such kind of topographic situations.
Conclusions
• Most effective BMP is driveway interceptor
• Cistern provided 9% of the annual irrigation demand
• Annual irrigation demand was reduced by 53% by adjusting the irrigation efficiency and application rates based on the plant water demand
• Simulation studies concluded that infiltration and surface runoff is the function of the soil properties and its effective depth
• Selection of the BMP is based on testing the soil properties
Q. Xiao, E G McPherson, J R Simpson and S L Ustin, Hydrologic processes at the urban residential scale, October 2006
Introduction
The paper focuses on building up a physical numerical model to understand hydrologic processes in the urban scale and its interaction with different BMP’s. The model was applied to a single family residential lot in Los Angeles, CA. Model was calibrated and validated by the data collected for over 2 yrs. They have modeled rain gutters, driveway interceptors and lawn retention basin. It was observed that there was about 65% reduction in storm water by using driveway interceptor. But using the rainwater gutters the annual irrigation demand was provided by 9%. They modeled trying to replace the existing loam soil with clay soil and it was observed that annual runoff discharge to the street was increased by 63%. In this study they have considered the case study of Los Angeles which has the polluted runoff being flown into local watershed and thereby into the coastal recreation areas. LA experience sever summer deficit of water demand where they import almost 85% of the water to meet their demands. Author focuses on how the policy makes are considering the implementation of the decentralized approach in the urban watershed i.e creating a mini reservoir system which retains runoff and reducing the irrigation demand but they lack the quantitative data on the effectiveness of the usage of different BMP’s. Most of the urban hydrologic models focus on the water balance and not on the effects on the BMP’s on the hydrologic processes which are dependent on the land use.
Objective of the study:
Develop, calibrate and validate the physical model at the residential scale for simulating the runoff for both long term events and storm term storm events. The study is carried out to see the effectiveness of BMP and simulating the effects of retrofitting the control site with the BMP’s.
Methodology:
• Site: residential housing unit
• BMP’s considered: cistern, swale, rain gutter, lawn retention basin, driveway interceptor and drywell
• Designed for 50 yr storm event
• Runoff from rooftops flowed into cistern and lawn retention basin
• Runoff from driveway flowed into the interceptor and further into the drywell
The data was collected from Jan 2001 to Jan 2003 and the model was calibrated and modeled based on the measurements of land covers, hydrologic processes and micro climate data. The runoff from treatment Site Street was collected in the measurement system and then through the water outlet drainages where as the runoff from the control site was drained directly into the drive way and thereby into the street. Each land cover was considered as a individual modeling unit. The model was studied for both long term simulation and short term events. For the long term simulation the focus of study was taking into consideration the annual irrigation use and runoff to the street and for short term simulation the focus was on the storm runoff to the street and the storage of the BMP.
• Precipitation data taken as the input of the rainfall data
• Two strategies of irrigation (first using the annual ET and second one was based on the soil and water deficit) were used in the model to determine the timing and the amount of water to be applied.
• Interception losses are also taking into consideration in the model
• Surface flow is calculated using the manning’s equation
• Infiltration is estimated using Green Ampt’s equation
• Percolation
• Evapo-transpiration
• Lateral subsurface flow
• Cistern storage is based on the total amount of water entering the cistern and the cistern restricted area
Numerical simulation focused on the landscape irrigation, storm runoff, ET and percolationof runoff into the deep layers, sensitivity analysis of the landscape retrofit was simulated using the date of 2001 and the 50 yr storm event (1988). The purpose of the sensitivity analysis was to distinguish which BMP was more effective with respect to runoff reduction and conservation of landscape irrigation water use. For the control site there were not BMP retrofitted which was considered as the base condition and the treatment site was varied with 6 different BMP’s. Other type of sensitivity analysis was carried by replacing with the clay soil instead of sandy soil.
Results:
Both treatment site and control site had about 50% impervious cover. Field measurement revealed that both cistern and retention basin were more effective in conserving municipal water supplies and reducing the total storm runoff. Surface runoff was effectively reduced by the drive way interceptors and lawn retention basin.
Discussion:
It was observed that all the BMP’s installed in the treatment site were effective in reducing the storm runoff and reducing the irrigation demand from the municipal water demand. Using this strategies enabled the improvement of urban eco system. BMP has reduced the maximum runoff flow to the streets and the maximum annual evaporation and increased the deep percolation. Increased percolation increases the downstream base flow and groundwater recharge. For runoff reduction driveway interceptor and drywell combination was the best but this BMP carries the risk of carrying pollutants from the surface runoff into the ground water. Cistern provided additional irrigation demand supply but for more effective use of cistern it should be appropriately designed for the catchment areas. Few of the sites will not follow the sub surface balance which focuses on the importance of more detailed model with such kind of topographic situations.
Conclusions
• Most effective BMP is driveway interceptor
• Cistern provided 9% of the annual irrigation demand
• Annual irrigation demand was reduced by 53% by adjusting the irrigation efficiency and application rates based on the plant water demand
• Simulation studies concluded that infiltration and surface runoff is the function of the soil properties and its effective depth
• Selection of the BMP is based on testing the soil properties
Friday, July 10, 2009
Discussion # 4 (10 July 2009)
Article Reference:
Kristin L Gilroy, Richard H McCuen, Spatio-temporatl effects of low impact development practices, 2009, Journal of hydrology
Introduction:
Many hydrologists are taking up the option of Best Management practices which is majorly coz of their cost effectiveness. As it is known that with the increase of urbanization, impervious surface increases which prevents water from infiltrating. This prevention of infiltration reduces the ground water recharge, lowers the water table, increasing the surface runoff and reducing the base flow during the dry periods. These add to degrading the stream channels and increasing the transport of pollutants. With storm management plans like detention pond, wet pond and infiltration basin which though they reduce the downstream peak discharge, issue of increased runoff into the stream continue which harms the aquatic life in the stream. Author points out as to the dependency of finding out the factors like optimal location, quantity of BMP and land use types needs to be investigated to achieve the optimal effectiveness of the LID’s. According to study carried out by Schneider et.al (2006) cisterns have showed to be more effective for the small storms when compared to large storms. Various studies concluded that infiltration technique is beneficial depending on the soil texture and the rainfall event size. According to Brander et. Al (2004) using numerous infiltration technique is more effective than few conventional practices with equivalent volume. Hood et. Al (2007) carried out the comparative study with LID development and traditional development which showed that the peak discharge, runoff volume and lag time is reduced and especially for smaller storms and shorter durations. The paper covers the study carried out on a spatio-temporal micro watershed which is used to study the effectiveness of BMP with respect to the location and the volume.
Methodology:
Spatial Temporal micro watershed model was developed in Matlab taking into consideration both spatial and temporal characteristics of rainfall and runoff processes of any return period on the lot sized watershed. The study was carried out for 1 yr (2.2 cm) and 2 yr (3.6 cm) events and for 4 watershed layouts (i.e. forested and three varying development patterns – townhome with 35% impervious land, residential with 32.5% impervious cover, and commercial land use with 72% impervious cover). In the study they have considered bio retention and cisterns to deal with the increased peak discharge and runoff volume. Bio retention was used majorly as the infiltration where as cisterns was above surface storage which is used to manage the runoff from the rooftop. From the long term precipitation records in Middle Atlantic region 50% of the storm events are 2 yr or less and of 1 hr duration.
Each lot was assumed to be of 0.1 ha. The model created grids of 20 x 12 cells. For the residential and townhome development, runoff from the roadways is directed onto the grass. Surface runoff of each cell is calculated differently which contained bio retention facility and with cisterns. Cisterns are used only to intercept the rooftop runoff. Outflow from the cistern is drained through the orifice. Once the storage of the cistern is reached, the outflow occurs through the cistern. In the bio retention cells once the depth of the stored water exceeds the storage it overflow where the overflow is equal to total volume of water in storage minus the maximum storage capacity of the facility and this is directed to the down gradient cell.
Results:
Onsite effects on the development
The comparison was made between the type of land pattern (forest, residential, townhome, and commercial) with respect to the two hydrologic metrics i.e. peak discharge and the total volume. It was observed that 1 yr storm was totally infiltrated in the forest pattern i.e. pre development condition. The development patterns have reduced the infiltration potential by increasing the impervious cover.
Effectiveness of cisterns
Effectiveness of cisterns was studied by analyzing the different land patterns with 1 and 2 yr storm events. It was observed that cisterns were more effective with smaller storm, there was about 30% reduction of peak discharge for single family lot for 1 yr storm where as it was only 10% reduction for 2 yr rainfall. If cistern is the only type of BMP in the site then they are ineffective for 2 yr storm, where as they provide some control for smaller events. Volume of cistern should be based on the volume of the runoff at the time of maximum rainfall intensity.
Effectiveness of bio retention location
For the study they have considered the location of bio retention near the cistern overflow, around the outlet of the micro watershed, along the roadway in residential/townhome lot and along the parking lot in commercial lot. There was effective reduction in the runoff volume when bio retention intercepted the roadway runoff and cisterns intercepted the rooftop runoff. But considering the bio retention at the overflow of the cisterns was more effective in reducing the peak discharge, as cisterns storage is reached before the highest storm intensity with bio retention facility there would be storage to control the peak. Location and volume of BMP is dependent on the time of distribution of the rainfall event. Location of bioretention did not make much of a reduction in the total runoff volume as the impervious land cover was greater when compared to the bio retention storage. Whereas peak discharge was decreased with the cisterns storing the runoff and overflow from the cisterns was drained to bio retention facilities.
Effect of storage volume
It was observed that effect of storage volume depends majorly on the land use type. For 1 yr storm event for commercial lot, it was noticed that when the facilities were doubled the peak discharge was reduced by 10% and volume to 30%. Author makes an observation that number of additional facilities is dependent on each watershed which might improve the hydrology.
Varying effects on hydrologic metrics
The two metrics considered to measure the effectiveness of the LID are peak discharge and total runoff volume. Effect of changing the volume of the storage on these two metrics vary as runoff volume can be controlled with the increased storage
Discussion:
I personally found this article interesting as they have tried to check the effectiveness of the LID’s considered, which would be similar to what I am trying to do in my thesis. Check the effectiveness of the RHS and PP in improving the hydrologic condition of the study area. This paper would be helpful when I take up the option of optimizing the location of the LID’s I am considering. For my thesis I would like to measure the two metrics they have mentioned and use Marcio’s HFR metric to analyze the impacts.
Kristin L Gilroy, Richard H McCuen, Spatio-temporatl effects of low impact development practices, 2009, Journal of hydrology
Introduction:
Many hydrologists are taking up the option of Best Management practices which is majorly coz of their cost effectiveness. As it is known that with the increase of urbanization, impervious surface increases which prevents water from infiltrating. This prevention of infiltration reduces the ground water recharge, lowers the water table, increasing the surface runoff and reducing the base flow during the dry periods. These add to degrading the stream channels and increasing the transport of pollutants. With storm management plans like detention pond, wet pond and infiltration basin which though they reduce the downstream peak discharge, issue of increased runoff into the stream continue which harms the aquatic life in the stream. Author points out as to the dependency of finding out the factors like optimal location, quantity of BMP and land use types needs to be investigated to achieve the optimal effectiveness of the LID’s. According to study carried out by Schneider et.al (2006) cisterns have showed to be more effective for the small storms when compared to large storms. Various studies concluded that infiltration technique is beneficial depending on the soil texture and the rainfall event size. According to Brander et. Al (2004) using numerous infiltration technique is more effective than few conventional practices with equivalent volume. Hood et. Al (2007) carried out the comparative study with LID development and traditional development which showed that the peak discharge, runoff volume and lag time is reduced and especially for smaller storms and shorter durations. The paper covers the study carried out on a spatio-temporal micro watershed which is used to study the effectiveness of BMP with respect to the location and the volume.
Methodology:
Spatial Temporal micro watershed model was developed in Matlab taking into consideration both spatial and temporal characteristics of rainfall and runoff processes of any return period on the lot sized watershed. The study was carried out for 1 yr (2.2 cm) and 2 yr (3.6 cm) events and for 4 watershed layouts (i.e. forested and three varying development patterns – townhome with 35% impervious land, residential with 32.5% impervious cover, and commercial land use with 72% impervious cover). In the study they have considered bio retention and cisterns to deal with the increased peak discharge and runoff volume. Bio retention was used majorly as the infiltration where as cisterns was above surface storage which is used to manage the runoff from the rooftop. From the long term precipitation records in Middle Atlantic region 50% of the storm events are 2 yr or less and of 1 hr duration.
Each lot was assumed to be of 0.1 ha. The model created grids of 20 x 12 cells. For the residential and townhome development, runoff from the roadways is directed onto the grass. Surface runoff of each cell is calculated differently which contained bio retention facility and with cisterns. Cisterns are used only to intercept the rooftop runoff. Outflow from the cistern is drained through the orifice. Once the storage of the cistern is reached, the outflow occurs through the cistern. In the bio retention cells once the depth of the stored water exceeds the storage it overflow where the overflow is equal to total volume of water in storage minus the maximum storage capacity of the facility and this is directed to the down gradient cell.
Results:
Onsite effects on the development
The comparison was made between the type of land pattern (forest, residential, townhome, and commercial) with respect to the two hydrologic metrics i.e. peak discharge and the total volume. It was observed that 1 yr storm was totally infiltrated in the forest pattern i.e. pre development condition. The development patterns have reduced the infiltration potential by increasing the impervious cover.
Effectiveness of cisterns
Effectiveness of cisterns was studied by analyzing the different land patterns with 1 and 2 yr storm events. It was observed that cisterns were more effective with smaller storm, there was about 30% reduction of peak discharge for single family lot for 1 yr storm where as it was only 10% reduction for 2 yr rainfall. If cistern is the only type of BMP in the site then they are ineffective for 2 yr storm, where as they provide some control for smaller events. Volume of cistern should be based on the volume of the runoff at the time of maximum rainfall intensity.
Effectiveness of bio retention location
For the study they have considered the location of bio retention near the cistern overflow, around the outlet of the micro watershed, along the roadway in residential/townhome lot and along the parking lot in commercial lot. There was effective reduction in the runoff volume when bio retention intercepted the roadway runoff and cisterns intercepted the rooftop runoff. But considering the bio retention at the overflow of the cisterns was more effective in reducing the peak discharge, as cisterns storage is reached before the highest storm intensity with bio retention facility there would be storage to control the peak. Location and volume of BMP is dependent on the time of distribution of the rainfall event. Location of bioretention did not make much of a reduction in the total runoff volume as the impervious land cover was greater when compared to the bio retention storage. Whereas peak discharge was decreased with the cisterns storing the runoff and overflow from the cisterns was drained to bio retention facilities.
Effect of storage volume
It was observed that effect of storage volume depends majorly on the land use type. For 1 yr storm event for commercial lot, it was noticed that when the facilities were doubled the peak discharge was reduced by 10% and volume to 30%. Author makes an observation that number of additional facilities is dependent on each watershed which might improve the hydrology.
Varying effects on hydrologic metrics
The two metrics considered to measure the effectiveness of the LID are peak discharge and total runoff volume. Effect of changing the volume of the storage on these two metrics vary as runoff volume can be controlled with the increased storage
Discussion:
I personally found this article interesting as they have tried to check the effectiveness of the LID’s considered, which would be similar to what I am trying to do in my thesis. Check the effectiveness of the RHS and PP in improving the hydrologic condition of the study area. This paper would be helpful when I take up the option of optimizing the location of the LID’s I am considering. For my thesis I would like to measure the two metrics they have mentioned and use Marcio’s HFR metric to analyze the impacts.
Subscribe to:
Posts (Atom)
