Investigation of multi-objective optimization algorithms for the design of urban spaces with a focus on quantitative and qualitative control of urban runoff.

Abstract

Introduction

Surface runoff is considered as one of the main components of the hydrological cycle and one of the important sources of water supply. In today's era, with the increasing trend of urbanization and as a result of changing the use of permeable surfaces to impervious surfaces, there have been adverse changes in the quality and quantity of surface runoff. However, by applying flood management methods, this resource can be used in a controlled manner and in the best possible way to meet water needs. Based on this, in recent years, a concept called the best management solutions with the abbreviation BMPs has been proposed to control the quantity and quality of runoff. In these activities, by increasing the retention time of the flood in the reservoirs, increasing the roughness coefficient and increasing the permeability of the surfaces, an attempt is made to reduce the peak discharge and the volume of runoff, as well as control the concentration of pollutants in the runoff.

Materials and Methods

Based on this and considering the importance of runoff management in a metropolis like Tehran, in this research, a part of the catchment area of the 22nd district of Tehran municipality was selected and evaluated the effects of BMPs on the amount of runoff using mathematical models of precipitation and runoff. In order to investigate the subject of the research, it has been tried by considering three objective functions of runoff quality (including BOD5 and TSS quality parameters), runoff quantity (including the volume of runoff produced in each sub-basin) and cost (including flood damage and maintenance costs of BMPs) to The comparison of two optimization models NSGAII and MOPSO should be paid.

Results and Discussion

The results of these two multi-objective evolutionary optimization algorithms conclude that the NSGAII optimization algorithm is more suitable due to the use of features such as crowding distance and the speed of performing different steps in the optimization algorithm. In addition, the use of MOPSO optimization algorithm will be easier due to the inclusion of fewer parameters than NSGAII. It is also necessary to mention that reaching the steady state in NSGAII will take place in fewer generations than MOPSO. . Also, the results of the evaluation of BMPs in the form of different scenarios showed that the application of these solutions can reduce the peak discharge from 16.3% to 1.50% and also reduce the volume of runoff from 9.2% to 37.4% depending on the type and The number of BMPs used at the basin level. Considering that, in general, the phenomenon of rainfall-runoff is a process that is strongly influenced by uncertain factors, and the inappropriate selection of design parameters leads to the incorrect estimation of the flood discharge and as a result, the selection of unfavorable dimensions for structures and technical performance becomes inappropriate or uneconomical. Designs and ultimately financial and human losses will be many. Therefore, the correct selection of design parameters is very important. In this regard, better results can be achieved by applying methods such as uncertainty analysis of inputs and effective parameters on the results of modeling or analyzing the sensitivity of the model to changing parameters. Among the input factors of rainfall-runoff models that have a noticeable effect on the results, we can mention the temporal and spatial distribution of rainfall, the continuity of rainfall and the previous soil moisture conditions.

, it is shown that the MOPSO model produces higher quality solutions in most cases compared to the NSGA-II model. Additionally, in cases where the NSGA-II model provides higher quality solutions, the execution time or distribution of solutions in the MOPSO model is better. This is while in a real-world problem and depending on the type of objective functions and their application, the results obtained from the application of these algorithms may be contrary to the experimental function results.

Conclusion

In this research, after analyzing the uncertainty of the temporal and spatial distribution of rainfall as well as the initial moisture of the soil using the Monte Carlo simulation method and analyzing the sensitivity of the flood hydrograph to the continuation of the rainfall, flood management strategies in the region were investigated. The results of the investigations showed that the highest peak flow is obtained from rainfall with a duration of 0.5 hours, in this case the range of peak flow changes is equal to 34.8 cubic meters per second, which indicates the presence of high uncertainty in the input parameters of the rainfall-runoff model.

Overall, in terms of comparing the capabilities of the NSGA-II and MOPSO optimization models in this simulation-optimization problem, it should be noted that the optimal values related to objective functions on the optimal exchange curve by the NSGA-II algorithm exhibit more dispersion compared to the MOPSO algorithm, indicating a wider range of scenarios generated by the NSGA-II algorithm. In fact, using this algorithm can provide decision-makers with more scenarios with significant diversity in objective function values. This is not observed in the results obtained from the MOPSO algorithm.