Mei Congli

An Improved Particle Swarm Optimization With Fuzzy c-Means Clustering Algorithm

This paper introduces a novel velocity equation of particle swarm optimization algorithm (PSO) based on fuzzy cmeans (FCM) cluster analysis of the current particles’ position. Besides the previous best location and the global best point, the cluster weighted centers could also be important biological force in the evolution of particles. And local information could be transferred among individuals by a cluster center points. In contrast to standard PSO (SPSO) and PSO with constriction factor (CPSO), the proposed approach is tested with a set of six benchmark functions with different dimensions. Experimental results indicate that this enhancement make the algorithm converge rapidly to good solutions on benchmark functions.

A Novel NN-Based Soft Sensor Based on Modified Fuzzy Kernel Clustering for Fermentation Process

With massive data of a fermentation process, a single data-based soft-sensor modeling method suffers from heavy features and bad accuracy. A novel soft sensor using multi-model neural network (MNN) based on modified kernel fuzzy clustering is proposed. Firstly, Features of sample data are extracted using principal component analysis (PCA) and the secondary variables are determined by PCA. Secondly, a kernel fuzzy c-means clustering algorithm based on particle swarm optimization (PSO) is applied to group the principal data into overlapping clusters, and neural network (NN) is used to construct sub-models based on the clusters. Finally, the estimation of every sub-model is fused by computing the weighted sum of the local models. The proposed modeling method is used to construct a novel soft sensor model for an erythromycin fermentation process. Case studies show that the approach has better performance compared to the conventional single model.

Soft Sensor Modeling Using SVR Based on Genetic Algorithm and Akaike Information Criterion

Support vector regression (SVR) is one of the new methods of soft sensor modeling for estimating the products of metabolism in microorganism fermentations. The accuracy of SVR is mainly impacted by two factors: input variables selection and parameters set in SVR training procedures. But it is difficult to select the input variables and set the parameters. A novel method of soft sensor modeling is proposed based on Akaike Information Criterion (AIC) and Genetic Algorithm (GA) to overcome the difficulties. Moreover, a real experiment process—erythromycin fermentation process is used to evaluate the performance of the proposed soft sensor modeling method. Results show the accuracy of the estimation is improved and the number of the input variables is reduced by the proposed approach, and the presented method could have a promising application in industrial process.

Static Decoupling Control Based on Hammerstein Model and Neural Network for Induction Motors

A novel static decoupling control strategy based on Hammerstein model and neural network for induction motors was proposed in this paper. Hammerstein model, consisting of a static nonlinear module and a dynamic linear module, can be used to model many nonlinear systems. In the proposed method, firstly, neural network and auto-regressive moving-average (ARMA) model were employed to construct the static nonlinear module and the dynamic linear module respectively. Further, neural network inverse model of the static nonlinear module can be trained on the static dataset collected in the framework of the Hammerstein model. Finally, the inverse model was utilized to offset the nonlinear characteristic of an induction motor, decoupled into a rotor speed subsystem and a rotor flux subsystem. Simulations show that the proposed static decoupling control strategy has satisfactory decoupling performances and robustness to load disturbance in close loop control.

Hybrid Power System Composed of Photovoltaic and Commercial Power

As an important application of solar energy, photovoltaic generation has attracted more and more attention. But there are some disadvantages in the perspective of solar energy, such as dispersion, intermittent, and randomness, which cannot provide stable and consistent power and should be equipped with additional energy. This chapter chooses commercial power as the back-up energy and proposes a hybrid power system with both photovoltaic and commercial power. The system is composed of a solar cell, the commercial power, a DC–DC(direct current) converter, a power factor correction (PFC) converter and DC load. In order to utilize the solar energy as much as possible, a power management is necessary for the hybrid power system. The core of the energy management strategy is to keep the system work under suitable mode to control the energy flow of the system according to the work status of the solar cells and the load. The experimental results verified the effectiveness of the energy management strategy.

On the identifiabiliy of gross errors in process industries

Gross error detection and discarded are the foundations of data reconciliation in process industries. High effective methods of gross error detection for both steady and dynamic systems have not be presented. So the proposition on the identifiability of gross errors is performed. The result of discussion on gross error identifiability and the theory of equivalent set are reviewed. And a new criterion of gross error identifiability is proposed based on the analysis of measurements correlation matrix. Theory analysis and simulation show the efficiency of the presented criterion.

Data reconciliation by two-step risk analysis of modeling

A new method for data reconciliation by risk analysis of modeling is presented in this paper. Yamarura designed an integer programming model for gross error detection and data reconciliation based on Akaike information criterion. But much computational cost is needed for its combinational nature. To reduce computation burden, a new method by two-step risk analysis of modeling is proposed. Measurement modeling risk is analyzed in the first step. Then gross error modeling analyzed based on the minimum measurement modeling risk is considered. The proposed method could effectively reduce the scale of the integer programming problem. Simulation shows the efficiency of the proposed method.