基于遗传算法的土地利用优化：NSGA-Ⅱ和NSGA-Ⅲ的对比研究

土地利用优化通常要兼顾不同群体的多种要求,理论上是复杂的超多目标(4个及以上)优化问题。但实际操作中却往往被简化为多目标(2—3个)优化问题,通过一种流行的多目标优化算法第Ⅱ代非支配排序遗传算法(NSGA-Ⅱ)求解。究其原因是对超多目标优化算法认知的缺失和与多目标优化算法理论对比的匮乏。对NSGA系列中应用最广泛的多目标优化算法NSGA-Ⅱ和最新提出、面向超多目标优化的算法NSGA-Ⅲ进行探究,从理论和实验两方面对Ⅲ和Ⅱ进行对比,从而探究二者进行土地利用优化时的优劣。在理论上,对比两种算法原理的异同。在实验中,分别设计多目标(3个目标)和超多目标(13个目标)土地利用优化问题,利用两种算法进行求解。对实验结果采用四层架构、六大指标进行全面评价,以对比两种算法的可用性。理论对比发现,两个算法只有种群多样性保护的方法不同,其中NSGA-Ⅲ是基于与固定的参考点的距离,而NSGA-Ⅱ则是基于相邻解间的距离。通过实验对比发现,NSGA-Ⅲ在超多目标优化时运算速度快,且产生的最优方案实用价值更高,NSGA-Ⅱ在算法的有效性方面更有优势。     

Integrated framework of system dynamics and meta-heuristic for multi-objective land use planning problem

Land use optimization as a resource allocation problem can be defined as the process of assigning different land uses to a region. Sustainable development also involves the exploitation of environmental resources, investment orientation, technology development, and industrial changes in a coordinated form. This paper studies the multi-objective sustainable land use planning problem and proposes an integrated framework, including simulation, forecasting, and optimization approaches for this problem. Land use optimization, a multifaceted process, requires complex decisions, including selection of land uses, forecasting land use allocation percentage, and assigning locations to land uses. The land use allocation percentage in the selected horizons is simulated and predicted by designing a System Dynamics (SD) model based on socio-economic variables. Furthermore, land use assignment is accomplished with a multi-objective integer programming model that is solved using augmented ε-constraint and non-dominated sorting genetic algorithm II (NSGA-II) methods. According to the results of the SD model, land use changes depend on population growth rate and labor productivity variables. Among the possible scenarios, a scenario focusing more on sustainable planning is chosen and the forecasting results of this scenario are used for optimal land use allocation. The computational results show that the augmented ε-constraint method cannot solve this problem even for medium sizes. The NSGA-II method not only solves the problem at large sizes over a reasonable time, but also generates good-quality solutions. NSGA-II showed better performance in metrics, including number of non-dominated Pareto solutions (NNPS), mean ideal distance (MID), and dispersion metric (DM). Integrated framework is implemented to allocate four types of land uses consisting of residential, commercial, industrial, and agricultural to a given region with 900 cells.

     

Multi-objective optimization of an ecological assembly model

The aim of the present work is to use multi-objective evolutionary algorithms (MOEA) to parameterise an ecological assembly model based on Lotka–Volterra dynamics. In community assembly models, species are introduced from a pool of species according to a sequence of invasion. By manipulating the assembly sequences, we look at the structure of the final communities obtained by a multi-objective process where the goal is to optimize the productivity of the final communities. The MOEA must also meet the constraint that the communities constructed in this fashion have a specified connectance. The Non-dominated Sorting Algorithm (NSGA-II) and the Strength Pareto Evolutionary Algorithm (SPEA2) were employed to optimize sequences according to the multi-objective optimization problem. The results show that the assembly process using optimized sequences generated different community structure than those generated via random sequences. First, the assembled communities are much more productive than those obtained from random sequences. We show that this increase of productivity is due to the degree distribution of the community food web, which was reshaped by the optimization process. In addition, using identical regional species pools the MOEAs were able to generate communities of different expected connectances. These results demonstrate the effectiveness of NSGA-II and SPEA2 for optimizing parameters in ecological models.     

Reliable classification of two-class cancer data using evolutionary algorithms

In the area of bioinformatics, the identification of gene subsets responsible for classifying available disease samples to two or more of its variants is an important task. Such problems have been solved in the past by means of unsupervised learning methods (hierarchical clustering, self-organizing maps, k-mean clustering, etc.) and supervised learning methods (weighted voting approach, k-nearest neighbor method, support vector machine method, etc.). Such problems can also be posed as optimization problems of minimizing gene subset size to achieve reliable and accurate classification. The main difficulties in solving the resulting optimization problem are the availability of only a few samples compared to the number of genes in the samples and the exorbitantly large search space of solutions. Although there exist a few applications of evolutionary algorithms (EAs) for this task, here we treat the problem as a multiobjective optimization problem of minimizing the gene subset size and minimizing the number of misclassified samples. Moreover, for a more reliable classification, we consider multiple training sets in evaluating a classifier. Contrary to the past studies, the use of a multiobjective EA (NSGA-II) has enabled us to discover a smaller gene subset size (such as four or five) to correctly classify 100% or near 100% samples for three cancer samples (Leukemia, Lymphoma, and Colon). We have also extended the NSGA-II to obtain multiple non-dominated solutions discovering as much as 352 different three-gene combinations providing a 100% correct classification to the Leukemia data. In order to have further confidence in the identification task, we have also introduced a prediction strength threshold for determining a sample's belonging to one class or the other. All simulation results show consistent gene subset identifications on three disease samples and exhibit the flexibilities and efficacies in using a multiobjective EA for the gene subset identification task.     

Multi-objective optimization of an industrial penicillin V bioreactor train using non-dominated sorting genetic algorithm

Bulk of the penicillin produced is used as raw material for semi-synthetic penicillin (such as amoxicillin and ampicillin) and semi-synthetic cephalosporins (such as cephalexin and cefadroxil). In the present paper, an industrial penicillin V bioreactor train is optimized for multiple objectives simultaneously. An industrial train, comprising a bank of identical bioreactors, is run semi-continuously in a synchronous fashion. The fermentation taking place in a bioreactor is modeled using a morphologically structured mechanism. For multi-objective optimization for two and three objectives, the elitist non-dominated sorting genetic algorithm (NSGA-II) is chosen. Instead of a single optimum as in the traditional optimization, a wide range of optimal design and operating conditions depicting trade-offs of key performance indicators such as batch cycle time, yield, profit and penicillin concentration, is successfully obtained. The effects of design and operating variables on the optimal solutions are discussed in detail.     

Multi-objective optimization of cascade reservoirs using NSGA-II: A case study of the Three Gorges-Gezhouba cascade reservoirs in the middle Yangtze River,China

A multi-objective optimization model of cascade reservoirs was developed to maximize the power generation and minimizie the appropriate ecological flow shortage index (AEFSI) downstream from the reservoir. Additionally, the non-dominated sorting genetic algorithm (NSGA-II) was used to search for multi-objective Pareto optimal solutions. The paper took the Three Gorges-Gezhouba cascade reservoirs as a case study. After validating the model, data from three typical years were used in the optimization. The results indicated that maximizing power generation by adjusting the optimal rules increased power generation by 1.07%, 0.91%, and 1.03% in normal, wet, and dry hydrological years, respectively, while increasing the AEFSI by 22.12%, 11.78%, and 14.67% (compared to real operations). The AEFSI was improved (decreased) by 21.90%, 10.27%, and 18.52% when the optimal rules favored the downstream ecology, but power generation decreased by 1.61%, 1.06%, and 2.29%, respectively, in the different hydrological years. Moreover, the results provide a set of well-distributed optimal solutions along the Pareto front that allow decision-makers to easily determine the best compromised solutions based on the trade-offs between the economic and ecological benefits. The results of this study provide guidance for decision-makers to improve the comprehensive benefits of the Three Gorges-Gezhouba cascade reservoirs.     

Ecological reservoir operation based on DFM and improved PA-DDS algorithm: A case study in Jinsha river,China

Abstract

This study aims to explore the optimization of reservoir operation in the Jinsha River, accounting for both economic and ecological needs. Power generation was set as an objective to express economic needs, while water supply, shipping, and flood control were considered as restrictions. The distribution flow method (DFM) based on Pearson-III curve was proposed to calculate the most optimal ecological flow (MOEF). Additionally, minimal difference between the outflow and MOEF was set as an objective to reflect ecological needs. Results indicated that the DFM was more objective, effective, and convenient than the improved Tennant method and monthly frequency computation method in calculating the MOEF. An improved Pareto archived dynamically dimensioned search (PA-DDS) algorithm was introduced to address completed reservoir optimization. Results highlighted that the application of the improved PA-DDS produced more convergence, uniformity, and stability in nondominated solutions (NDS), compared with the non-dominated sorting genetic algorithm-II (NSGA-II). Separate optimization and joint optimization of cascade reservoir operation were studied using the improved PA-DDS. The results indicated that joint optimization produced higher power generation in the dry year and normal year, and significant reduction in the difference between the outflow and MOEF in the wet year.     

Resonance assignment of the NMR spectra of disordered proteins using a multi-objective non-dominated sorting genetic algorithm

A multi-objective genetic algorithm is introduced to predict the assignment of protein solid-state NMR (SSNMR) spectra with partial resonance overlap and missing peaks due to broad linewidths, molecular motion, and low sensitivity. This non-dominated sorting genetic algorithm II (NSGA-II) aims to identify all possible assignments that are consistent with the spectra and to compare the relative merit of these assignments. Our approach is modeled after the recently introduced Monte-Carlo simulated-annealing (MC/SA) protocol, with the key difference that NSGA-II simultaneously optimizes multiple assignment objectives instead of searching for possible assignments based on a single composite score. The multiple objectives include maximizing the number of consistently assigned peaks between multiple spectra (“good connections”), maximizing the number of used peaks, minimizing the number of inconsistently assigned peaks between spectra (“bad connections”), and minimizing the number of assigned peaks that have no matching peaks in the other spectra (“edges”). Using six SSNMR protein chemical shift datasets with varying levels of imperfection that was introduced by peak deletion, random chemical shift changes, and manual peak picking of spectra with moderately broad linewidths, we show that the NSGA-II algorithm produces a large number of valid and good assignments rapidly. For high-quality chemical shift peak lists, NSGA-II and MC/SA perform similarly well. However, when the peak lists contain many missing peaks that are uncorrelated between different spectra and have chemical shift deviations between spectra, the modified NSGA-II produces a larger number of valid solutions than MC/SA, and is more effective at distinguishing good from mediocre assignments by avoiding the hazard of suboptimal weighting factors for the various objectives. These two advantages, namely diversity and better evaluation, lead to a higher probability of predicting the correct assignment for a larger number of residues. On the other hand, when there are multiple equally good assignments that are significantly different from each other, the modified NSGA-II is less efficient than MC/SA in finding all the solutions. This problem is solved by a combined NSGA-II/MC algorithm, which appears to have the advantages of both NSGA-II and MC/SA. This combination algorithm is robust for the three most difficult chemical shift datasets examined here and is expected to give the highest-quality de novo assignment of challenging protein NMR spectra.     

Effect of deacetylation time on the preparation, properties and swelling behavior of chitosan films

The objectives of the study were to propose a two-stage microfluidization combined with an ultrafiltration (UF) treatment for chitosan mass production and the manipulation of molecular weight and its distribution. The proposed methods are based on the degradation rate and rate constant of various process variables studied. Results obtained were that the rate constants were faster during the earlier reaction period, were higher for those operating at a higher pressure, were better for using concurrent UF treatment to remove small degraded fragments, and the degradation rate constants were faster for 30 °C solutions than that for 50 or 0 °C. A two-stage microfluidization process is proposed. The first stage constitutes of the highest possible concentration solution with concurrent UF treatment at 50 °C, and recycled 5 times. The second stage consists of the highest possible concentration of solutions with concurrent UF treatment at 30 °C, and recycled 5 times.     

Global optimization of the Saccharomyces cerevisiae: Fermentation process

In this article, steady‐state optimization of the Saccharomyces cerevisiae fermentation process problem is performed revealing the existence of multiple optimum solutions. The globally optimum solution was determined using the NEOS global optimization solver LINDO. A branch and bound strategy (bnb20.m) and the global search and multistart algorithms in the MATLAB global optimization toolbox were successful in determining locally optimum solutions and these results are validated by plotting the objective function against the decision variables. While in some cases all the strategies were successful in obtaining the globally optimum solutions, an example is presented where the most beneficial product value, which is not a stationary point and lies on the feasible boundary, is obtained by the LINDO global optimization solver (but not the other routines) as the globally optimum solution. The Jones–Kompala model was used to model the steady‐state of the fermentation process. While several articles have been published demonstrating the existence of nonlinearities and bifurcations in this model, the challenges posed by this model to optimization has never been investigated so far and this work attempts to do so. Both dilution rate and the oxygen mass transfer coefficient were used as the decision variables individually and together. © 2013 American Institute of Chemical Engineers Biotechnol. Prog., 29:917–923, 2013     

A two-stage cultivation process for the growth enhancement of Chlorella vulgaris

This study proposes a two-stage cultivation process with an autotrophic growth followed by a mixotrophic process. The results indicated that a two-stage cultivation process using a daily dose of 3 g/L of glucose could achieve 7.4 g/L of biomass, which was about a 64 % increase over simple autotrophic cultivation. In the second stage of mixotrophic cultivation, glucose was regarded as a better carbon source for cell growth, than was glycerol. Linoleic acid (C18:2) would be the primary component in the two-stage cultivation as in the autotrophic cultivation. Even carbon source was provided in the second stage of mixotrophic cultivation; lower light intensity limited the mixotrophic growth, which indicated that photosynthesis still plays an important role in the second stage of mixotrophical cultivation. The final biomass was higher after this two-stage cultivation process, which made it suitable for application in the production scale-up of algal biomass.     

Improvement of fumigaclavine C production in a two-stage culture of <Emphasis Type="Italic">Aspergillus fumigatus</Emphasis> with molasses as a cost-effective ingredient

Fumigaclavine C (FC), which is produced by Aspergillus fumigatus, is a conidiation-associated ergot alkaloid with significant medical benefits. However, its application is restricted by low yields from submerged cultures. In this study, the technical feasibility of using molasses as a cost-effective ingredient for FC production in a two-stage culture of A. fumigatus was evaluated. The results indicated that molasses supplementation significantly enhanced FC accumulation by promoting conidiation and up-regulating hydroxymethylglutaryl-CoA reductase activity. Via the optimization of the two-stage process in the presence of molasses, FC production in shake flasks reached 226.9 mg/L, which was approximately three times that in the original medium (75.9 mg/L). The use of molasses as a cost-effective ingredient for FC fermentation was also successfully reproduced in a lab-scale bioreactor system in which the maximum FC production reached 215.0 mg/L. The FC production obtained in this study is the highest ever reported. This increased efficiency will enable large-scale production of FC and extend the application of molasses as a low-cost substrate for producing other conidiation-related secondary metabolites.     

Multicriteria optimization of biochemical systems by linear programming: application to production of ethanol by Saccharomyces cerevisiae

In this study we present a method for simultaneous optimization of several metabolic responses of biochemical pathways. The method, based on the use of the power law formalism to obtain a linear system in logarithmic coordinates, is applied to ethanol production by Saccharomyces cerevisiae. Starting from an experimentally based kinetic model, we translated it to its power law equivalent. With this new model representation, we then applied the multiobjective optimization method. Our intent was to maximize ethanol production and minimize each of the internal metabolite concentrations. To ensure cell viability, all optimizations were carried out under imposed constraints. The different solutions obtained, which correspond to alternative patterns of enzyme overexpression, were implemented in the original model. We discovered few discrepancies between the S-system-optimized steady state and the corresponding optimized state in the original kinetic model, thus demonstrating the suitability of the S-system representation as the basis for the optimization procedure. In all optimized solutions, the ATP level reached its maximum and any increase in its activity positively affected the optimization process. This work illustrates that in any optimization study no single criteria is of general application being the multiobjective and constrained task the proper way to address it. It is concluded that the proposed multiobjective method can serve to carry out, in a single study, the general pattern of behavior of a given metabolic system with regard to its control and optimization.     

Factors affecting lipid accumulation by <Emphasis Type="Italic">Nannochloropsis oculata</Emphasis> in a two-stage cultivation process

Reserve lipids of microalgae are promising for biodiesel production. However, optimization of cultivation conditions for both biomass yield and lipid production of microalgae is a contradictory problem because required conditions for both targets are different. In this study, a two-stage cultivation strategy is proposed to enhance lipid production of the microalga Nannochloropsis oculata. Biomass growth and lipid production were carried out in two separate and non-interacting stages. In first-stage cultivation, microalgae were cultivated in optimal conditions for cell growth. Then, microalgae were harvested and transferred into a growth-limited environment, thus enhancing lipid production of microalgae. Here, optimization of the lipid production stage (second stage) with respect to different levels of inoculum concentration, salinity of culture broth, and intensity of irradiance was performed. The results show that irradiance exhibits a significant influence on lipid production. The highest lipid productivity of 0.324 g L⁻¹ day⁻¹ was obtained with an inoculum concentration of 2.3 g L⁻¹, a salinity of 35 g L⁻¹, and an irradiance of 500 μmol photons m⁻² s⁻¹. The final yield of lipid obtained from the two-stage process was 2.82-times higher than that from traditional single-stage batch cultivation systems.     

Development and optimization of two-stage cyclic fed-batch culture for hG-CSF production using L-arabinose promoter of Escherichia coli

The long-term process for producing human granulocyte-colony stimulating factor (hG-CSF) was developed using two-stage cyclic fed-batch culture, in which hG-CSF expressing-recombinant Escherichia coli was directed by an L-arabinose promoter system. For the optimization, the preinduction growth rate during the growth stage and the feeding strategy during the production stage were investigated. The maximum harvest volume during the production stage was predicted before long-term cyclic operation. Based on those optimized strategies, the two-stage cyclic fed-batch culture was performed for 12 cycles (86 h). The cell growths in both stages were maintained at 45-50 g/L and 71-77 g/L, respectively. hG-CSF was stably produced at a level of 8-9 g/L and the plasmid stability was maintained at more than 90%. Volumetric productivity by the two-stage cyclic fed-batch culture was 0.643 g/L/h, which was about 280% higher than that of conventional DO-stat fed-batch culture.     

Effect of modulation of enzyme inactivation on temperature optimization for reactor operation with chitin-immobilized lactase

Temperature is a critical variable to be optimized in any enzymatic process, producing opposite effects on enzyme activity and inactivation rate. Temperature functions for all kinetic and inactivation parameters were validated for chitin-immobilized yeast lactase (CIL). Enzyme inactivation was described by a two-stage series mechanism. The effect of galactose and lactose on inactivation was determined in terms of modulation factors that were positive for galactose and negative for lactose. Modulation factors were mild functions of temperature in the first stage and strong functions in the second stage of enzyme inactivation, where galactose positive modulation factors increase while lactose negative modulation factors decrease with temperature. Temperature-explicit functions for kinetic and inactivation parameters were incorporated into a scheme to optimize temperature in the simulation of a continuous packed-bed reactor operation with chitin-immobilized lactase, based on an annual cost objective function. Optimum temperature was 20°C at enzyme replacement of 25% residual activity, and increased only slightly at higher replacement frequencies. The effect of modulation factors on reactor design and temperature optimization is presented and discussed. Software for temperature optimization that allows the introduction of variations in all parameters and operational criteria to perform sensitivity analysis was developed.     

On Polydispersity of Plant Biomass Recalcitrance and Its Effects on Pretreatment Optimization for Sugar Production

This paper discusses a property associated with plant biomass recalcitrance to enzyme and microbial deconstructions in sugar production from cellulose and hemicelluloses. The hemicelluloses are more readily hydrolyzed to sugars than is cellulose. As a result, optimization to maximize individual glucose and hemicellulose sugar recovery is not possible. This property is an inherent feature of plant biomass and is named polydispersity of plant biomass recalcitrance (PPBR) in this study. A set of pretreatment experiments using eucalyptus and sulfite pretreatment to overcome recalcitrance of lignocelluloses was conducted. The results were used to predict the conditions for individually maximizing enzymatic glucose and xylose yields. The predicted maximal yields were used to quantitatively illustrate the PPBR concept. The effect of PPBR on pretreatment optimization and strategies for maximal sugar recovery using two-stage pretreatment are discussed.     

Optimization of Inocula Conditions for Enhanced Biosurfactant Production by Bacillus subtilis SPB1, in Submerged Culture, Using Box–Behnken Design

The inoculum age and density can influence considerably the production yield and cost of the fermentation process. Some literature studies report the use of two-stage inocula to enhance metabolite production. In the present study, optimization studies were done in order to define the best inocula conditions supporting a maximum biosurfactant production by Bacillus subtilis SPB1. Hence, by adjusting the levels of the two-stage inocula strategy, lipopeptide production was effectively enhanced to almost 3.4 g/l as estimated gravimetrically. The new defined parameters were as follows; a first inoculum age of 23 h followed by a second inoculum age and size of 4 h and 0.01, respectively. Thereby, we note an improved production as compared to the production yield described under non-optimized inocula conditions reported in our previous work.     

Multi-Objective Evolutionary Optimization of Biological Pest Control with Impulsive Dynamics in Soybean Crops

The biological pest control in agriculture, an environment-friendly practice, maintains the density of pests below an economic injury level by releasing a suitable quantity of their natural enemies. This work proposes a multi-objective numerical solution to biological pest control for soybean crops, considering both the cost of application of the control action and the cost of economic damages. The system model is nonlinear with impulsive control dynamics, in order to cope more effectively with the actual control action to be applied, which should be performed in a finite number of discrete time instants. The dynamic optimization problem is solved using the NSGA-II, a fast and trustworthy multi-objective genetic algorithm. The results suggest a dual pest control policy, in which the relative price of control action versus the associated additional harvest yield determines the usage of either a low control action strategy or a higher one.     

De novo design of potential RecA inhibitors using multi objective optimization

De novo ligand design involves optimization of several ligand properties such as binding affinity, ligand volume, drug likeness, etc. Therefore, optimization of these properties independently and simultaneously seems appropriate. In this paper, the ligand design problem is modeled in a multiobjective using Archived MultiObjective Simulated Annealing (AMOSA) as the underlying search algorithm. The multiple objectives considered are the energy components similarity to a known inhibitor and a novel drug likeliness measure based on Lipinski's rule of five. RecA protein of Mycobacterium tuberculosis, causative agent of tuberculosis, is taken as the target for the drug design. To gauge the goodness of the results, they are compared to the outputs of LigBuilder, NEWLEAD, and Variable genetic algorithm (VGA). The same problem has also been modeled using a well-established genetic algorithm-based multiobjective optimization technique, Nondominated Sorting Genetic Algorithm-II (NSGA-II), to find the efficacy of AMOSA through comparative analysis. Results demonstrate that while some small molecules designed by the proposed approach are remarkably similar to the known inhibitors of RecA, some new ones are discovered that may be potential candidates for novel lead molecules against tuberculosis.