Existing Optimization Methods Research Articles

Expert knowledge data-driven based actor–critic reinforcement learning framework to solve computationally expensive unit commitment problems with uncertain wind energy

With the expansion of power grid, unaffordable computational cost and time will pose serious challenges of time-efficient scheduling in unit commitment problem (UCP). However, existing optimization methods, i.e., mathematical programming methods and meta-heuristic algorithms, are powerless and time-consuming to handle computationally expensive UCP (CEUCP). Thus, reinforcement learning methods with strong inference and time-saving performances are motivated to solve the computationally expensive challenges in tackling CEUCPs. In this paper, a novel expert knowledge data-driven based actor–critic (AC) reinforcement learning methodology is proposed for solving CEUCPs. Specifically, in the proposed AC reinforcement learning methodology, expert knowledge, data-driven surrogate model, and improved meta-heuristic algorithm are integrated for further performance enhancement. Firstly, a novel action selection mechanism (based on the expert knowledge of thermal units characteristic) is integrated into AC to improve the efficiency of network training. Secondly, an improved extreme learning machine (ELM) data-driven surrogate model is proposed to build reward function in AC. In detail, original cost function in reward is replaced by a lightweight ELM model. Shape distance is integrated into ELM for enhancing accuracy. Finally, original marine predators algorithm (MPA) is improved for obtaining optimal dispatching decisions and rewards of AC method quickly and correctly. Original search pattern is replaced by quantum based representation for boosting convergence. The excellent performances of the proposed AC framework are verified by simulations of 10-units, 100-units, and 100-units with wind energy test systems.

An Adaptive Dual-Mode Task-Oriented Resource Management Strategy for GEO Relay Systems

With the fierce global competition on satellite networks, the building of satellite constellations grows explosively. Sharply increasing on-orbit data will face the challenge of satellite-ground data transmission. GEO satellites become the top choice for satellite data relay due to their stable satellite-ground link. Most existing spectrum resource management for GEO relays is equipment-oriented and benefit priority, which may lead to a waste of spectrum resources. In this paper, we propose a real-time task-oriented resource allocation strategy for GEO relay systems. We model the spectrum allocation problem as a distributed non-cooperative Stackelberg game process. We prove that when both sides of the game pursue the maximization of personal revenue, the system will enter a Nash equilibrium state, whereas spectrum resources are not fully used. Based on the maximization of individual utilities (U-prior) and spectrum utilization (S-prior) methods, we design an adaptive dual-mode pricing mode to maximize the spectrum resources within a certain loss of revenue. The simulation results show that the S-prior and U-prior have better performance than the baseline method and existing optimization methods. Our proposed dual-mode strategy is making more throughputs and has less delay with little loss of utility values than that of individual utility maximization.

Joint cooperative caching and power control for UAV-assisted internet of vehicles

In view of the current problems of spectrum resource scarcity, return congestion, and insufficient energy utilization in the unmanned aerial vehicle (UAV)-assisted Internet of Vehicles (IoV), this paper investigates the cooperative caching and power control, and proposes a joint optimization method to improve the overall Energy Efficiency (EE) . In this method, we first propose a communication establishment threshold to control the V2V communication distance and serve as a joint optimization factor. Then we derive the closed form expressions of offloading ratio and EE of the UAV-assisted IoV, and formulate the optimization problem of maximizing EE. Due to the coupling relationship between caching strategy and transmission power, it is difficult for us to directly solve the optimization problem. Furthermore, we propose an alternating optimization algorithm for solving the optimization problem. Finally, the experimental simulation compare the propose joint optimization method with other existing optimization methods, and the simulation results prove the effectiveness and superiority of the propose joint optimization method.

A practical optimization framework for political redistricting: A case study in Arizona

Following the decennial census, each state in the U.S. redraws its congressional and state legislative district boundaries, which must satisfy various legal criteria. For example, Arizona’s Constitution describes six legal criteria, including contiguity, population balance, competitiveness, compactness, and the preservation of communities of interest, political subdivisions and majority–minority districts, each of which is to be enforced “to the extent practicable”. Optimization algorithms are well suited to draw district maps, although existing models and methods have limitations that inhibit their ability to draw legally-valid maps. Adapting existing optimization methods presents two major challenges: the complexity of modeling to achieve multiple and subjective criteria, and the computational intractability when dealing with large redistricting input graphs. In this paper, we present a multi-stage optimization framework tailored to redistricting in Arizona. This framework combines key features from existing methods, such as a multilevel algorithm that reduces graph input sizes and a larger local search neighborhood that encourages faster exploration of the solution space. This framework heuristically optimizes geographical compactness and political competitiveness while ensuring that other criteria in Arizona’s Constitution are satisfied relative to existing norms. Compared to Arizona’s enacted map (CD118) to be used until 2032, the most compact map produced by the algorithm is 41% more compact, and the most competitive map has five more competitive districts. To enable accessibility and to promote future research, we have created Optimap, a publicly accessible tool to interact with a part of this framework. Beyond the creation of these maps, this case study demonstrates the positive impact of adapting optimization-based methodologies for political redistricting in practice.

Improving classification accuracy of fine-tuned CNN models: Impact of hyperparameter optimization

The immense popularity of convolutional neural network (CNN) models has sparked a growing interest in optimizing their hyperparameters. Discovering the ideal values for hyperparameters to achieve optimal CNN training is a complex and time-consuming task, often requiring repetitive numerical experiments. As a result, significant attention is currently being devoted to developing methods aimed at tailoring hyperparameters for specific CNN models and classification tasks. While existing optimization methods often yield favorable image classification results, they do not provide guidance on which hyperparameters are worth optimizing, the appropriate value ranges for those hyperparameters, or whether it is reasonable to use a subset of training data for the optimization process. This work is focused on the optimization of hyperparameters during transfer learning, with the goal of investigating how different optimization methods and hyperparameter selections impact the performance of fine-tuned models. In our experiments, we assessed the importance of various hyperparameters and identified the ranges within which optimal CNN training can be achieved. Additionally, we compared four hyperparameter optimization methods—grid search, random search, Bayesian optimization, and the Asynchronous Successive Halving Algorithm (ASHA). We also explored the feasibility of fine-tuning hyperparameters using a subset of the training data. By optimizing the hyperparameters, we observed an improvement in CNN classification accuracy of up to 6%. Furthermore, we found that achieving a balance in class distribution within the subset of data used for parameter optimization is crucial in establishing the optimal set of hyperparameters for CNN training. The results we obtained demonstrate that hyperparameter optimization is highly dependent on the specific task and dataset at hand.

Synthesis of ship systems optimal regulators based on matrix inequalities

The output and state regulators synthesis for ship control systems based on linear matrix inequalities using the CVX software package for solving optimization problems by the method of semi-definite programming is considered. Estimates parameters for regulators providing asymptotically stable systems regimes on Lyapunov inequalities solutions sets are obtained. An algorithm for controlling the ship course using a robust regulator in state feedback, followed by the use of CVX technologies to implement the minimum energy consumption control algorithm, is presented. The control vector estimation is reduced to the minimization of the L2 norm and it is performed in two stages, namely, the best mode is first determined, and then the mode under constraints is defined. The use of Lyapunov inequalities with the calculations implementation in CVX format on proprietary solvers has allowed us to obtain compact program texts and expand the set of restrictions introduced. In the algorithm for solving the boundary value problem of controlling a discrete model of an object, in contrast to existing optimization methods, a mode of transfer from the initial state to the final one with the right boundary, not necessarily equal to zero, is implemented. For a class of observable and controllable time–invariant systems, using the Krylov’s function contained in the MATLAB gallery of functions, L2‑norms to be minimized are formed. Examples of regulator calculations with a given degree of stability and optimal systems, confirming proposed solutions correctness, are given.

Adaptive population-based simulated annealing for resource constrained job scheduling with uncertainty

Transporting ore from mines to ports is of significant interest in mining supply chains. These operations are commonly associated with growing costs and a lack of resources. Large mining companies are interested in optimally allocating resources to reduce operational costs. This problem has been previously investigated as resource constrained job scheduling (RCJS). While a number of optimisation methods have been proposed to tackle the deterministic problem, the uncertainty associated with resource availability, an inevitable challenge in mining operations, has received less attention. RCJS with uncertainty is a hard combinatorial optimisation problem that is challenging for existing optimisation methods. We propose an adaptive population-based simulated annealing algorithm that can overcome existing limitations of methods for RCJS with uncertainty, including pre-mature convergence, excessive number of hyper-parameters, and the inefficiency in coping with different uncertainty levels. This new algorithm effectively balances exploration and exploitation, by using a population, modifying the cooling schedule in the Metropolis-Hastings algorithm, and using an adaptive mechanism to select perturbation operators. The results show that the proposed algorithm outperforms existing methods on a benchmark RCJS dataset considering different uncertainty levels. Moreover, new best known solutions are discovered for all but one problem instance across all uncertainty levels.

Automatic multi‐objective particle swarm optimization method for effective Doherty power amplifier design

AbstractIn this paper, an automatic multi‐objective particle swarm optimization (AMOPSO) method is developed for the design of Doherty power amplifiers (DPAs). In comparison to the well‐known built‐in optimizer available in commercial simulators, the proposed method not only reduces the optimization time, but also provides superior power added efficiency (PAE) for the final power amplifier. According to the reported measurements, the output PAE of the fabricated DPA exceeds 50% at 6 dB backoff, the saturated PAE is more than 61%, and the saturated output power (Pout) is over 43 dBm in the target frequency range of 1.9–2.1 GHz. As compared with existing optimization methods, the proposed method allows reducing optimization time by more than 37%.

Modified African Buffalo for Optimal Accommodation of Distributed Energy Resources (DERs) for Annual Energy Loss Minimization in Active Distribution Systems

Abstract. The paper is acquainted with a new optimization technique. Modified African Buffalo optimization (MABO), to solve optimal Distributed Energy Resources (DERs) accommodation problem formulated for annual energy loss minimization esteeming multiple distributed generations and shunt capacitors contemporaneously. To demonstrate the proposed MABO, primitively, the optimal endowment of DERs is obtained for power loss minimization of benchmark 33 – bus test active distribution system (ADN) and simulation resultants are compared with well - established optimization techniques. The optimal DER allocation is also predetermined for depreciation of annual energy loss in conjuration with node voltage deviation over three different load levels. The simulation results obtained are further compared with some of the existing optimization methods. The two-level comparability of optimization methods shows that the Modified African Buffalo method has better searching potency to determine optimal solutions for optimal DERs allocation problems of ADN.

Hybrid optimization approach for power scheduling with PV-battery system in smart grids

This manuscript proposes a hybrid method for the smart-grid (SG) optimization, which combines automatic demand-response (DR) shedding with load classification. The integration of the Mexican Axolotl Optimization (MAO) and Honey Badger Algorithm (HBA) constitutes the proposed hybrid approach. The HBA method improves the axolotls' updating behavior. It is commonly referred to as the Enhanced MAO (EMAO) approach. The proposed energy-management framework optimizes customer power consumption patterns to minimize carbon-emissions, electricity-costs, and peak-power-consumption. By integrating utility generation, PV-battery systems, and dynamic price signals using the EMAO approach, it reduces power consumption costs, minimizes peak-fluctuations, and lowers carbon emissions. The EMAO control-topology is rigorously evaluated through MATLAB simulations, demonstrating superior performance compared to existing optimization methods such as HGPO, PSO, and GA. The results showcase the EMAO algorithm consistently achieving the lowest cost at 310 cents, minimizing carbon emissions to 1.8 pounds, and achieving a high load classification accuracy of 98.2 %. With a moderate performance-to-cost ratio of 1.7, the EMAO algorithm excels in energy management, effectively balancing cost considerations, environmental impact, and load classification objectives. The proposed hybrid method effectively integrates DR shedding and load classification to optimize SG-operation, achieving significant improvements in cost, emissions, and load-classification accuracy compared to traditional methods.

The continuous stochastic gradient method: part II–application and numerics

In this contribution, we present a numerical analysis of the continuous stochastic gradient (CSG) method, including applications from topology optimization and convergence rates. In contrast to standard stochastic gradient optimization schemes, CSG does not discard old gradient samples from previous iterations. Instead, design dependent integration weights are calculated to form a convex combination as an approximation to the true gradient at the current design. As the approximation error vanishes in the course of the iterations, CSG represents a hybrid approach, starting off like a purely stochastic method and behaving like a full gradient scheme in the limit. In this work, the efficiency of CSG is demonstrated for practically relevant applications from topology optimization. These settings are characterized by both, a large number of optimization variables and an objective function, whose evaluation requires the numerical computation of multiple integrals concatenated in a nonlinear fashion. Such problems could not be solved by any existing optimization method before. Lastly, with regards to convergence rates, first estimates are provided and confirmed with the help of numerical experiments.

Period Cycle Optimization of Integrated Energy Systems with Long-Term Scheduling Consideration

The economy and energy saving effects of integrated energy system dispatch plans are influenced by the coupling of different energy devices. In order to consider the impact of changes in equipment load rates on the optimization and scheduling of the system under long-term operation, a method for energy and component cycle optimization considering energy device capacity and load has been proposed. By improving the initial parameters of the components, energy economic parameters, and operational optimization parameters, the system is subjected to long-term scheduling and multi-cycle operational optimization analysis to evaluate the energy saving and emission reduction potential as well as the economic feasibility of the system. Finally, through numerical analysis, the effectiveness of this optimization approach in achieving energy savings, emission reductions, and cost benefits for the system is validated. Furthermore, compared to existing optimization methods, this approach also assesses the economic feasibility of the system. The case study resulted in a pre-tax IRR of 23.14% and a pre-tax NPV of 66.38 million. It is inferred that the system could generate profits over a 10-year operation period, thereby offering a more rational and cost-effective scheduling scheme for the integrated energy system.

Software-Defined Satellite Observation: A Fast Method Based on Virtual Resource Pools

In recent years, the proliferation of remote sensing satellites has dramatically increased the demands of Earth observation and observing efficiency. Designing a promising satellite resource scheduling method is a pivotal way to meet the requirements of this scenario. However, with hundreds or more satellites involved, the existing optimization methods struggle to address the NP-hard resource scheduling problem effectively. In this paper, an approach named software-defined satellite observation (SDSO) is proposed. First, adopting the new design ideology, we define a unified specification based on a discrete spatial grid to describe the observation capability of all satellites. The observation resources are virtualized using the virtual resource pool technique and then stored in the database in advance, implementing on-demand acquisition for observation resources. Next, we designed a model of the remote sensing satellite resource scheduling problem based on a virtual resource pool and designed a solution method for searching information within the virtual resource pool. Finally, the experimental results show that the computational efficiency of the proposed SDSO methodology has a substantial advantage over the traditional methods. Meanwhile, with the growing number of satellites involved in scheduling, there is only a slight degradation in the execution performance of our method, while the time complexity of optimization-based approaches increases exponentially.

A method for real-time optimal heliostat aiming strategy generation via deep learning

Optimal aiming strategies are essential for efficient solar power tower technology operation. However, the high calculation complexity makes it difficult for existing optimization methods to solve the optimization problem in real-time directly. This work proposes a real-time optimal heliostat aiming strategy generation method via deep learning. First, a two-stage learning scheme where the neural network models are trained by genetic algorithm (GA) benchmark solutions to produce an optimal aiming strategy is presented. Then, an end-to-end model without needing GA solutions for training is developed and discussed. Furthermore, a robust end-to-end training method using randomly sampled flux maps is also proposed. The proposed models demonstrated comparable performance as GA with two orders of magnitude less computation time through case studies. Among the proposed models, the end-to-end model shows significantly better generalization ability than the pure data-driven two-stage model on the test set. A robust end-to-end model with data enhancement has better robustness on unseen flux maps.

Battery thermal management system optimization using Deep reinforced learning algorithm

The temperature of the battery plays a critical role in the working conditions and a suitable temperature environment can help the battery extend its capacity. Air cooling and water cooling are prevalent as positive heat dissipation methods used in research. Phase change material (PCM) has a good heat dissipation capacity because of its latent heat property. Therefore, PCM is chosen as the battery thermal management method in this study. Adding two water channels in the center of the pack strengthens the heat dissipation capacity. Deep reinforcement learning (DRL) is used to explore the multi-optimal solution of the battery thermal arrangement, and its results are compared with the classical methods such as NSGA-ii and MOPSO. Max temperature, temperature difference of the battery, and average temperature of PCM are selected as the optimization targets. A comparison between DRL optimization and the initial system shows that the increased max temperature decreased by 0.33 ℃ (2.1%), and the average temperature of PCM decreased by 0.3 ℃ (2.3%). The temperature difference of DRL optimization decreased by 1.3% compared with that of NSGA-ii and MOPSO. The max temperature decreased by 0.2% and the average temperature of PCM decreased by 0.3% compared with those of NSGA-ii. In summary, DRL has the potential application in battery thermal management system (BTMS) optimization. The most important contribution of this work is utilizing DRL in the optimization process, which shows better results than existing optimization methods.

Generative design: Integrating rent and retail compatibility goals for automated tenant mix layout

Tenant mix is a central issue in the development and operation of shopping malls. An effective tenant layout can enhance the consumer experience and increase shop sales. So it is crucial to comprehensively consider the rent and retail compatibility when crafting the tenant mix. Developers mainly considered the rent income from anchor and non-anchor tenants. But many studies ignored the externality effect of anchor tenants on the rent of non-anchor tenants. Furthermore, the relationship between different tenant type and retail compatibility under customers' behavioral preferences is not adequately studied. In addition, existing optimization methods is inefficient, poorly integrated and cannot rapidly explore design options. Therefore, this paper aims to achieve automatic generation and optimization of tenant mix layouts considering the rent and retail compatibility objective. Firstly, a bi-objective model consisting of the rent and retail compatibility objectives is developed to evaluate the performance of the generated schemes. The rent objective function quantifies the externality of anchor shop on the rent of non-anchor shops in the shopping center. The retail compatibility function quantifies the relationship between the consumers’ behavior preferences and retail types. Then this paper proposed a generative mechanism consisting of parametric design, scheme optimization and decision making processes. The results of the case study show that the optimized scheme outperforms the original scheme in terms of both rent and retail compatibility objectives. The proposed generative mechanism achieves automatic generation and optimization of tenant mix layouts, and greatly improves the efficiency of the design process.

Enhancing robustness and noise rejection in flexible joint manipulators: an optimized sliding mode controller with enhanced gray wolf optimization for trajectory tracking

The objective of this study is to enhance the performance of a nonlinear three-rigid-link manipulator (RLM) with a focus on trajectory tracking, robustness against disturbances and noises, and adaptability to joint flexibility. To achieve this, we have employed an optimized sliding mode controller with a proportional integral derivative (PID) sliding manifold. The tuning process involves selecting the critical gains of the controller that minimizes the integral time absolute error (ITAE), serving as the objective function (OBJF) to optimize the performance of the robot manipulator. To identify the optimal gains of the controller, we have utilized a new optimization algorithm known as memory enhanced linear population size reduction gray wolf optimization (MELGWO). The efficacy of this algorithm is compared to other existing optimization methods in the literature. Moreover, this research has delved into the impact of joint flexibility on the robot system’s performance. Encouragingly, the results demonstrate that the optimized SMC–PID with MELGWO adaptation can effectively address joint flexibility while maintaining acceptable performance levels.

Optimal Cable Force Adjustment for Long-Span Concrete-Filled Steel Tube Arch Bridges: Real-Time Correction and Reliable Results

For complex structures, the existing optimization method for suspender cable forces involves extensive matrix operations during the solution process, requiring high computational power and time. As a result, obtaining a more accurate solution becomes challenging. To address this issue and improve the stress distribution of suspenders in the completed state, while minimizing the need for frequent cable force adjustments and grid beam elevation changes during construction, a novel method for cable force optimization is proposed. In this study, the Pingnan Third Bridge, which is the world’s longest large span arch bridge with a span of 575 m, is taken as the engineering background. This study combines finite element analysis and multi-objective optimization methods to develop a cable force optimization approach for real-time correction during the panel girder lifting of long-span concrete-filled steel tube (CFST) arch bridges. The optimization method involves treating the panel girder weight and displacement during construction as parameter variables, and considering the displacement and unevenness of the panel girder in the completed state as constraint conditions. The objective equation is defined based on the displacement and cable force during the lifting construction process and, through optimization, the cable forces and displacements of each lifting section are calculated. The results demonstrate the feasibility of integrating optimization theory into the cable force optimization process during panel girder lifting. In this study, we have taken into account the characteristics of real-world engineering and focused on specific key points to reduce the order of the influence matrix. Consequently, the computational costs are reduced, facilitating the development of a multi-objective tension optimization program. By minimizing segment displacement variations and ensuring even cable force distribution in the completed state, the method ensures that the bridge meets the required completion requirements without the need for repetitive iterations or cumbersome calculations. It provides higher optimization efficiency and superior outcomes, offering significant value for cable force calculations during suspender construction of similar bridge types and guiding construction processes.

Swin-FlowNet: Flow field oriented optimization aided by a CNN and Swin-Transformer based model

This paper presents a novel aerodynamic design optimization method for airfoils. The existing optimization method is transformed from being aerodynamic coefficient-oriented to flow field-oriented. By rapidly obtaining the flow fields of airfoils within the optimization domain and integrating them with aerodynamic theory, preoptimized airfoils can be obtained. To achieve this, we propose a hybrid network called Swin-FlowNet, which combines convolutional neural network (CNN) and Transformer. This network establishes a mapping between parameterized airfoils and their corresponding flow fields. The Transformer component captures long-range dependencies, while the CNN component extracts features from local information to characterize the flow field. We introduce a simple integration of convolution and attention mechanisms, resulting in the Swin-Conv Transformer Block (SCTB) that enhances local attention to improve channel and spatial attention. In our model, the multilayer perceptron (MLP) in the Transformer block is replaced with deep convolution, reducing the number of model parameters and enhancing feature extraction capability. Experimental results demonstrate that Swin-FlowNet achieves accurate and general performance in reconstructing and predicting the flow field around small variations of airfoils. Furthermore, Swin-FlowNet outperforms CNN-based and Transformer-based architectures in terms of accuracy and efficiency trade-offs.

Butterfly Algorithm for Sustainable Lot Size Optimization

The challenges faced by classical supply chain management affect efficiency with regard to business. Classical supply chain management is associated with high risks due to a lack of accountability and transparency. The use of optimization algorithms is considered decision-making support to improve the operations and processes in green manufacturing. This paper suggests a solution to the green lot size optimization problem using bio-inspired algorithms, specifically, the butterfly algorithm. For this, our methodology consisted of first collecting the real data, then the data were expressed with a simple function with several constraints to optimize the total costs while reducing the CO2 emission, serving as input for the butterfly algorithm BA model. The BA model was then used to find the optimal lot size that balances cost-effectiveness and sustainability. Through extensive experiments, we compared the results of BA with those of other bio-inspired algorithms, showing that BA consistently outperformed the alternatives. The contribution of this work is to provide an efficient solution to the sustainable lot-size optimization problem, thereby reducing the environmental impact and optimizing the supply chain well. Conclusions: BA has shown that it can achieve the best results compared to other existing optimization methods. It is also a valuable chainsaw tool.