Evolutionary computation (EC) techniques, inspired by bio- logical evolution, are generally applied to global optimization. Particleswarm optimization (PSO) ,  is such a technique that simulates the social behavior of animal flocking, and that has demonstrated strong ability in searching for optimal solu- tions in a range of problem spaces. Thanks to its conceptual simplicity and high computational efficiency, PSO has gained increasing popularity since its inception and has become one of the main tools for solving real-world optimization problems. In this paper, a PSO-based approach is proposed for EISs to autonomously attain optimal solutions given imprecisely described problems. Using the recently introduced autonomouslearning multiple model (ALMMo) fuzzysystem  as the underlying implementation, the algorithmic procedure for si- multaneously optimizing both the antecedent and consequent parameters of EISs is presented. This optimization process works on data samples collected during the online learning process. It helps significantly boost the performance of EISs in terms of the prediction accuracy owing to PSO’s strong global searching ability. Numerical examples on benchmark problems as well as real-world ones are conducted for verifying the validity and effectiveness of the proposed general concept and principles.
Abstract—The antecedent and consequent parts of a first-order evolving intelligent system (EIS) determine the validity of the learning results and overall system performance. Nonetheless, the state-of-the-art techniques mostly stress on the novelty from the system identification point of view but pay less attention to the optimality of the learned parameters. Using the recently intro- duced autonomouslearning multiple model (ALMMo) system as the implementation basis, this paper introduces a particleswarm- based approach for EIS optimization. The proposed approach is able to simultaneously optimize the antecedent and consequent parameters of ALMMo and effectively enhance the system performance by iteratively searching for optimal solutions in the problem spaces. In addition, the proposed optimization approach does not adversely influence the “one pass” learning ability of ALMMo. Once the optimization process is complete, ALMMo can continue to learn from new data to incorporate unseen data patterns recursively without a full retraining. Experimental stud- ies with a number of real-world benchmark problems validate the proposed concept and general principles. It is also verified that the proposed optimization approach can be applied to other types of EISs with similar operating mechanisms.
As a population-based stochastic optimization technique, PSO can be used to ﬁnd the “good enough solution”. From the literature a num- ber of examples can be found where PSO has been used to support clustering. In Ref.  two PSO methods were proposed, one to ﬁnd the centroids of clusters and another that used K-means clustering to seed the initial swarm. In Ref.  PSO was applied to search the clus- ter centres in the arbitrary data set automatically. In Ref.  PSO was coupled with the K-means clustering to cluster document collec- tions. In Ref.  a hybrid method, FAPSO-ACO-K, was proposed which combined Fuzzy Adaptive ParticleSwarm Optimization (FAPSO), Ant Colony Optimization (ACO) and K-means so as to ﬁnd the best clus- ter partition in the nonlinear partitional clustering problem. In Refs. [20,21] a framework was proposed for Diﬀerential Evolution ParticleSwarm Optimization (DEPSO) based clustering which combined DE with PSO. However, to the best knowledge of the authors, there has been no work directed at using PSO for the purpose of density-based cluster parameter optimization. The above methods also have other limitations. Firstly, the encoding methods of clustering results based on enumerating all items are complex to search for an optimal solution. Secondly, the method can only be applied to the unsupervised learning. To the best knowledge of the authors, no work using SPSO to optimize the DBSCAN parameters for both supervised and unsupervised learn- ing using ﬁtness functions of the form presented in this paper has been conducted.
It has been shown that as capacity or delay increases, the traditional AQM schemes  eventu- ally become oscillatory and prone to instability . It is further pointed out  that these schemes demonstrate instability with the introduction of high bandwidth-delay links, because they still require a careful conguration of non-intuitive control param- eters. As a result, they are non-robust to dynamic network changes. They exhibit greater delays than the target mean queuing delay with a large delay variation, plus large buer uctuations and, consequently, cannot control the router queue. It has also been stressed that in practical queuing systems, the mean arrival rate and the mean service rate are frequently fuzzy, i.e. they cannot be expressed in exact terms [14,15]. The European Network for Intelligent Technologies (EU- NITE) Roadmap  points out that the application of fuzzy control techniques to the problem of congestion in IP-based networks is suitable, due to diculties in obtaining a precise mathematical model using conven- tional analytical methods. Fuzzy Logic AQM algo- rithms, presented in [13,17], use instantaneous queue length and the variation of the queue (trac incoming rate) as inputs. Three membership functions, for the two inputs and the output, are used. The system output is a probability with which packets are either dropped or marked if Explicit Congestion Notication (ECN)  is enabled. ECN is a means of explicitly notifying end-hosts of network congestion by marking, instead of dropping, packets. The performance of these fuzzy AQM algorithms is generally better than that of traditional approaches, such as PI and Adaptive RED. However, their major shortcoming lies in the fact that their control rules and membership functions are obtained through a manual tuning process, which
Intrusion detection based on ANN is built by using gathered features about several types of attacks. Usually, building knowledge based on gathered data required sufficient amount of data with comprehensive nature. Unfortunately, in the application of intrusion detection, it is not feasible to create a sufficient knowledge for learning or at least balanced learning between the different classes (refer to the problem described in KDD99 in the previous section). Therefore, learning algo- rithm has to be carefully optimized according to the nature of the dataset. This leads us to investigate about how to identify the optimization parameters of the learning algorithm. In this work, the problem will be formulated as an optimization problem. More specifically, the problem is how to find the optimal values of the hidden layer neurons in both SLFN, and FLN in order to maintain highest accuracy of testing. Such problem is addressed in the literature as a heuristic searching in the space of solutions considering the aim is to minimize an objective function represents the accuracy of the classification of attacks. Mathematically, assuming that the accuracy of the testing is the function f (x), where x = (x 1 , x 2 , . . . x n ) denotes the random selected different
Raja, M. A. Z., Shah, F. H., Tariq, M., Ahmad, I., & Ahmad, S. ul I. (2016). Design of artificial neural network models optimized with sequential quadratic programming to study the dynamics of nonlinear Troesch’s problem arising in plasma physics. Neural Computing and Applications. https://doi.org/10.1007/s00521-016-2530-2
The next stage for optimizing the component function from every input and output variables used PSO. For this research, the PSO process is used refer to . Next the parameter explanation that is used to PSO. The particle amount used is 28,24 particles from the input variables and 4 particles from the output variables. The particle used is the component function in every variable. The dimension of the particle is defined from the problem that will be optimizing. C1 (learning factor for particle), C2 (learning factor for swarm). The value used is 2. The particle maximum change while the iteration is happening, with the limit used is -1 to 1. The condition will stop while the maximum iteration can be reached. The iteration used is 100 times. The weight inertia (𝑤) is used to keep the balance between the global and local exploration capability.
Abstract: Negative output elementary Luo converter performs the conversion from positive DC input voltage to negative DC output voltage. Since Luo converters are non-linear and time-variant systems, the design of high performance controllers for such converters is a challenging issue. The controller should ensure system stability in any operating condition and good static and dynamic performances in terms of rejection of supply disturbances and load changes. To ensure that the controllers work well in large signal conditions and to enhance their dynamic responses, soft computing techniques such as Fuzzy Logic controller (FLC) and ParticleSwarm Optimization based FLC (PSO-FLC) are suggested. In recent years, Fuzzy logic has emerged as an important artificial intelligence tool to characterize and control a system, whose model is not known or ill defined. Fuzzy logic is expressed by means of if-then rules with the human language. In the design of a fuzzy logic controller, the mathematical model is not necessary. However, the rules and the membership functions of a fuzzy logic controller are based on expert experience or knowledge database. To ensure better performance of fuzzy controller, membership functions, control rules, normalizing and de-normalizing parameters are optimized using PSO. The main strength of PSO is its fast convergence than the other global optimization algorithms. To exhibit the effectiveness of proposed algorithm, the performance of the PSO based fuzzy logic controller has been compared with FLC and the necessary results are presented to validate the PSO for control purposes. Comparative study emphasize that the optimized PSO based fuzzy controller provide better performance and superior to the other control strategies because of fast transient response, zero steady state error and good disturbance rejection under variations of line and load and hence output voltage regulation is achieved. Simulation studies have been performed using Matlab-Simulink software.
Based on the above-mentioned problems, in [2-3], the variable universe fuzzy controllers are proposed, which introduce the contraction-expansion factor combined with relevant optimization algorithms, for universe’ adjusting, to improve the dynamic and static performances of the system. Through designing static and dynamic coupling of distributed fuzzy adaptive control law,  proposes a distributed fuzzy adaptive control approach based on Lyapunov theory with proving its stability.  elaborates a new sliding mode fuzzy control (SMFC), a kind of sliding mode fuzzy control （ SVM ） combined with space vector modulation (SVM) technology. study the design of the controller of T-S fuzzysystem with unknown parameters, and promotes to combine the Lyapunov theory with T-S fuzzysystem, which could ensure accuracy and stability of the system. [7-8],with the problem of limited control precision and self-adaptive ability of the traditional fuzzy control, optimizes the relevant parameters or rules of the controller with the particleswarm optimization, which improves the robustness and control precision.
Given the above, mitigation of frequency fluctua- tions in a PV- EWH-DEG-BESS based isolated hybrid power system has been investigated. An efficient control strategy together with artificial intelligence (AI) technique has been used for effectively maintaining the active power balance. Thus, the proposed strategy is expected to mitigate the oscillations which is caused due to the mismatch in the active power, so as to maintain the system frequency within an acceptable range of the proposed isolated hybrid power system . Different controllers and several optimization techni- ques for tuning the parameters of the controllers have been reported in the previous works [6-14]. In hybrid system studies, PI controllers have been used for maintaining the energy balance [6-8]. The parameters of the PI controllers in  has been tuned by Ziegler and Nichols method. GA optimized PI/PID [10,11], Fuzzy based PI controller [12,13], and neural network based PI controller  are also used. Firefly algorithm is optimized PID controllers  were successfully emp- loyed for an interconnected power system for load frequency control.
Due to its simplicity in implementation, PSO has gained popularities in engineering applications, such as in image processing  and in system modeling . A number of publications have also been reported in using PSO to automatically tune the FLC parameters [15, 16, 17, 18]. These publications are focused on tuning the parameters involved in the TS-type fuzzy controllers. In general, the PSO is used to perform the learning tasks that are usually associated with the NN in the TS FLCs. A PSO based fuzzy MFs optimizing method is also introduced to a fixed point control problem, i.e. parking a car into a predefined garage location [19, 20]. Although there are research results in the area of automatic fuzzy MFs optimizing, most of them are in the area of TS type of fuzzy controllers. To the best of our knowledge, there is no report on using PSO for the Mamdani-type of fuzzy controller MFs tuning.
At small servo valve spool displacements, leakage flow between the valve spool and body dominates the orifice flow through the valve . In precision positioning applications, where the servo valve operates within the null region, this flow, if ignored, may severely degrade the performance of a conventional servo hydraulic design. In this study, an accurate model of leakage flow  is used. It includes both leakage flow and orifice flow, and makes smooth transition between them would likely improve precision of the servo hydraulic system design and performance. The model used is a nonlinear servo valve model that accurately captures the servo valve leakage behaviour over the whole ranges of spool movement. The leakage behaviour is modelled as turbulent flow with a flow area inversely proportional to the overlap between the spool lands and the servo valve orifices.
The neuro-fuzzy inference system is optimized by adapting the antecedent parameters (membership function parameters) and consequent parameters (the polynomial coefﬁcients of the conse- quent part) so that a speciﬁed objective function is minimized. The adaptation methods of most fuzzy inference systems rely on the backpropagation algorithm that is generally used to recursively solve for parameter optimization. This conventional optimization algorithm is susceptible to get stuck at local optima. To overcome this drawback, evolutionary techniques such as genetic algorithm have been used. However, these techniques require much compu- tation time if there are many parameters to be optimized. There- fore, the least-squares method that is a one-pass optimization method is combined for optimizing a part of the parameters. The evolutionary technique is used to optimize the antecedent parameters c ij and s ij , and the least-squares algorithm is used to
Particleswarm optimization, PSO, is a stochastic optimization algorithm based on the population of solutions. It is originally developed by J. Kennedy and R. C. Eberhart  and presented as a new optimization method inspired by the phenomenon of biological and social- cognitive behaviour of different types of group-living creatures. Living in groups helps individuals in self- realization by means of social influence and social learning. Interaction with others produces the change of beliefs, views, and ultimately of the behaviour of individuals. These changes in the limited socio-cognitive space are represented as the movement of individuals. In other words, self- knowledge and knowledge of the group is optimized by the social contact because each individual can contribute to the overall knowledge of the group, the group transfers its knowledge by means of social contact to the individual members of the group. Clearly, these mechanisms are simplified descriptions of complex natural phenomena, but can faithfully illustrate the movement of swarms in search of better solutions. PSO algorithm uses a population-based search in which the particles change their position (state) by means of iterative algorithms. During the performance of the algorithm the particles change their positions in a multidimensional search space, i.e. the particles move in solutions space in search for the best positions.
2 used learning algorithm for training NN (Zweiri et al., 2003). BP algorithm is used in NN learning process for supervised or associative learning. Supervised learning learns based on the target value or the desired outputs. During training, the network tries to match the outputs with the desired target values. Other algorithm that usually use is Genetic Algorithm (GA) which is one of the famous evolutionary technique in NN learning.
In the original particleswarm optimization, there has also a lack of solution, because it is very easy to move to local optima. In certain circumstances, where a new position of the particle equal to global best and local best then the particle will not change its position. If that particle is the global best of the entire swarm then all the other particles will tend to move in the direction of this particle. The end of result is the swarm converging prematurely to a local optimum. If the new position of the particle pretty far from global best and local best then the velocity will changing quickly turned into a great value. This will directly affect the particle's position in the next step. For now the particle will have an updated position of great value, as a result, the particle may be out of bounds the search area.
Many investigations on LFC and AVR of power system have been reported in the literature. There are inherent non-linearities present in the power system components and synchronous machines, therefore most of the controllers employed in LFC and AVR are primarily composed of an integral controller . Proportional Integral Derivative (PID) controller is a powerful tool to improve both transient and steady state performance of a control system, but their tuning is a difficult process. Evolutionary computational techniques such as ParticleSwarm Optimization (PSO), Genetic Algorithm (GA), Bacterial Foraging Algorithm (BFA), Simulated Annealing (SA), Ant Colony Optimization (ACO), etc. have been reported in literature to tune PID controllers. In this paper, PSO algorithm is presented for searching the optimal gains of PID controller in LFC and AVR of a single area power system comprising a non-reheat type thermal generating unit. The LFC and AVR loops are designed to operate around normal state with small variable excursions. The loops may therefore be modelled with linear, constant coefficient differential equations and represented with linear transfer functions .
Fuzzy set is suitable to process the relevant problems to uncertainty and fuzziness and it has been extensively applied. Fuzzy clustering comes into being by integrating fuzzy clustering and the concept of fuzziness. Fuzzy clustering makes it possible for the clustering samples tobelong to multiple classes and it uses membership to refer to the size of the degree of membership. As a widely used method, FC algorithm has been successfully applied in image analysis, medical diagnosis, target identification and image segmentation. FC algorithm performs fuzzy clustering on the consistent pixels in the image through membership matrix and segments the image throughthe iterative optimization of the objective function. However, FC algorithm also has many shortcomings . For instance, it is greatly affected by noise. It is very sensitive to the initial value and it depends greatly on the selection of the initial clustering center. When the initial clustering center severely deviates from the global optimal clustering center, it may cause the algorithm trapped in local optimum, especially in the case of numerous clustering samples. Therefore, this kind of problems can be solved by improving the membership function in FC algorithm and introducing PSO with strong global optimization ability .
In this thesis, an Autonomous Group ParticleSwarm algorithm for DG placement in radial distribution networks is presented. Modal analysis and CPF are used for determining DG placement candidates, while the loading parameter is the comparison index for selecting the best DG places. The places are ranked using an MERC method, which determines a priority list of DG locations for reactive power compensation during occasions of reactive power shortage. The placement algorithm is executed on the well-known 33 bus radial distribution network, and the results show the remedial effect of DGs, both in loss reduction and voltage profile improvement in normal operation, and enhancement of the loading parameter in the case of voltage instability. The ranking method is executed over the obtained candidates to provide a priority list from the viewpoint of reactive power compensation in the case of shortage.
Swarm Intelligence (SI) is the latest of an artificial intelligence technique based around the study of collective behaviour in decentralized and self-organized systems. The idea of SI came from systems found in nature, including ant colonies, bird flocking and animal herding that can be effectively applied to computationally intelligent system. SI systems are typically made up from a population of agents interacting locally with one another and with their environment and local interactions between such nodes often lead to the emergence of a global behaviour. There are two major techniques in SI which are the Ant Colony Optimization (ACO) and ParticleSwarm Optimization (PSO). The ACO algorithm is a probabilistic technique for solving computational problems to finding good paths through graphs. They are inspired by the behaviour of ants in finding paths from the colony to food. While PSO (which is the focus of this project) is a technique where all the particles (or solutions) move to get better results. PSO is a new branch of the soft computing paradigms called evolutionary algorithms (EA). EA includes genetic algorithms (GA), evolutionary programming (EP), evolutionary strategies (ES) and genetic programming (GP). Before PSO, the most popular technique in evolutionary computing is Genetic Algorithm (GA). GA is widely used to determine BP learning parameters and weight optimization to make the convergence rate faster and avoid from being trapped in the local minima.