Liu Zaiwen

The Method of Soft Sensing Based on RBF Neural Network and Intelligent Control for Sewage Disposal Process

Soft sensing method was proposed for determination of effluent BOD from SBR and its principle was introduced, the method was based on the radial basic function (RBF) artificial neural network. The RBF neural network was trained and simulated by a lot of observed data, and the result showed that the RBF neural network may be used to fulfil soft sensing for effluent BOD from SBR, so as to create condition for real-time control of sewage disposal process, showing abroad perspective in application. The fuzzy control method for DO in SBR sewage disposal process was represented, and the aeration time and the energy consuming according to different DO curves were also studied. DO can be controlled in real time by the fuzzy logical to change the frequency of motor. Then we can adjust the concentration of DO according to different DO curves, and get the obvious advantages in shortening time of wastewater treatment and reducing energy consumed

A Soft Sensing Method Based on Process Neural Network

A new method of soft sensing based on process neural network (PNN) is proposed in this paper. PNN is an extent of traditional neural network, and it is a new configuration of artificial neural network put forward in recent years. The thesis discuss some modified algorithms for raising training speed of PNN, these algorithms are based on function orthogonal basis expansion which exist low-speed convergence in network training. An improved algorithm for BP network based on function orthogonal basis expansion in process neural network for soft sensing is researched. After increasing the normalizing rule on original algorithm, and introducing function momentum adjustment item and learning rate automatically adjustment method for network weight function, which has means of zero and standard deviations of one, the training time of learning algorithm for process neural network is reduced, and a good effect is represented by simulation in wastewater treatment system

Review of FRT requirements for integration of wind energy in China and Europe

A research on differences of latest Grid Code requirements for wind farm integration between China and selected EU countries are conducted. The similarities and differences of these Grid Codes in the aspects of low voltage ride-through ability, transient fault ride-through requirements, negative phase sequence requirements and reactive power control requirements are compared. The main causes of these differences are investigated and the author tried to predict the future trend of regulations.

Short-term predicting model for water bloom based on Elman neural network

This paper addresses the problem of predicting water bloom in short-term period. Important factors of water bloom are studied. A short-term predicting model of Elman neural network is presented according to the characteristic of time accumulation. The algorithm of Elman is first improved, and then the predicting model is trained, tested and compared with BP model. Experimental results show that: The short-term change of chlorophyll could be predicted better by Elman predicting model, which is accurate and extensive. This model is proven to be useful to predict water bloom in short-term period.

Panoramic image matching based on graph cuts algorithm

This paper proposes a method based on graph cut algorithm to match panoramic images. Label is denoted by disparity and the energy function is established, then the problem of matching can be converted into that of energy function minimization. A graph is constructed such that the energies can be related to the capacities of the cuts of the graph. The minimal energy is obtained by the network-flow theory, and hence the disparity data has been gotten. Consequently, we obtain the dense depth image. The experimental result shows that compiling with the other existing matching algorithm, the graph cuts algorithm can get better depth image, the high accuracy in the process to match the panoramic images reveal that it can be widely used in 3d indoor scene reconstruction.

Research on DO control based on Fuzzy Smith-PID controller for sewage disposal

Considering the characteristics of the time-delay, inertia and time-varying of the dissolved oxygen (DO) control in the sewage disposal process, a Fuzzy Smith-PID control model is proposed based on the analysis of the regular PID and Fuzzy control in this paper, which integrates the advantages of the high robustness of the Fuzzy controller, the compensation of the time delay problem of Smith Estimation Machine and the high accuracy of PID controller. Simulation results show that this complex strategy has good robustness and stability in both of the model matching and mismatching situation, and provides an effective way to solve the problem of the time-delay, inertia and time-varying for sewage disposal system.

An Intelligent System on Water Bloom Remote Monitor and Predication

According to lagging state of water quality monitor and the problems of the difficult of water bloom prediction, water bloom remote monitor and prediction system based on BP neural network are proposed in this paper. This system can realize the automatic real-time monitor for the change of water quality and via the prediction by neural network for water bloom; it can provide a kind of efficient and practical system for water environment control.

Compensation and Simulation of Restrictive Memory Least Square Method for BOD Soft Measurement Mechanism Model

In order to overcome the difficulty of on-line measurement for output water quality such as BOD (Biochemical Oxygen Demand) in the waste water process, a second-order soft measurement model based on Lawrence—McCarty formula is established. And to improve the precision of model formula, the restrictive memory least square method is used to identify the system parameter and then the identified model is utilized to compensate the measurement error of the Lawrence—McCarty formula. The established model and its compensating accuracy are tested by laboratory measurements, and the results show that the compensated mixed model is superior to the original soft measurement formula. This combined model can be used in different water qualities and provide parameters on-line measurement for waste water process.

Remote control for robot manipulators with neural network compensation

A new remote time-delay feedback controller is presented for a class of robot manipulator systems with unknown nonlinear dynamics and communication time-delay. The proposed control scheme consists of a local neural network (NN) compensation and a time-delay feedback controller. A NN-based identification is first employed to identify the robot manipulator system. Local linearization compensation is then used to cancel the unknown nonlinearity of the system. The filter technology is utilized to obtain the time-delay-free inverse model of the linearized system and a desired reference model is used to constitute the feedback controller, which can achieve the tracking performance. Rigorous stability analysis and simulation are both included to verify the effectiveness of the proposed scheme.

A self-localization method for stereo camera on the mobile robot

In order to determine the position of stereo camera on the mobile robot, whose motion center is unknown, a simple and effective method is proposed. No calibration equipment is needed. By extracting natural scene features based on non-maximal eigenvalue suppress and Harris corner detector, and matching them using grey level correlation and scene constrains, the space transformation relation between robot motion coordinate and stereo vision coordinate is calculated analytically. Accuracy and error analysis and experiments show that the total error is less than 1 cm and the self-localization process is rapid, so it can satisfy the requirements of objects tracking, 3D environment reconstruction and mapping via stereo vision.