Sheng Wanxing

Study of chaos genetic algorithms and its application in neural networks

In this paper, based on the relationship between chaos and evolution of chaotic theory, chaos is combined with a genetic algorithm. A new hybrid strategy is proposed in order to overcome the weakness of the simple genetic algorithm (SGA). The hybrid chaos GA is applied in neural networks and simulations show that the chaos GA is better than SGA in speed of searching and precision of optimization.

Short-term load forecasting using artificial immune network

The precision of short-term load forecasting for electric power systems directly affects the economic benefit of power systems. However, since the relationship between load power and factors influencing load power is nonlinear, it is difficult to identify its nonlinearity by using conventional methods. A nonlinear short-term load forecasting model based on artificial immune network is presented in this paper. The design and learning of neural network system uses immune network regulation and immune programming, and artificial immune network is defined. This paper presents the method in power system short-term load forecasting, and it is compared with neural network method. The result indicates that this model is effective and has adaptive ability itself.

Substation automation high speed network communication platform based on MMS+TCP/IP+Ethernet

Ethernet, which is gradually being used in the process bus in the communication of automation devices, has become the focus of manufacturers and customers. In order to fulfill the requirements, IEC TC57 will publish the IEC61850 standard for substation automation network communication. All evidences show that network communication technology, based on manufacturing message specification (MMS), TCP/IP and Ethernet, is the trend of the next substation automation communication system. The above communication platform will have many benefits for manufacture and system integration. This paper firstly gives the basic principles of the network communication platform integration frame for substation automation, including software and hardware platform. It then studies the real-time performances and priority of Ethernet and the MMS, TCP/IP protocols used over Ethernet. This paper also discusses the real-time network communication models of client/server and peer-peer in substations and gives simulation results of event-driven data transfer.

Research on parallel algorithm of DC optimal power flow in large interconnection power grids

This paper investigated an optimal method of optimal power flow in large interconnected power grids. A decomposition collaborative model based on partial duality is analyzed, and a parallel algorithm based on DC optimal power flow model is presented in multi-region decomposition of interconnected power grids. The OPF computation of large power grid is decomposed into subproblems of multi regions, which is a quadratic programming problem used to solve a DC optimal power flow. The optimal convergence condition of multi-region is discussed. The interchange information among regions is export price and boundary nodal bus phase angle. The IEEE RTS-96 with two and three interconnected regions is studied to illustrate the effect of the proposed approach and to prove a great future in interconnected power system

Evaluation on the application mode of distributed generation

A distributed generation system (DGS) consists of a cluster of DG (photovoltaic, diesel engines, wind-power, fuel cells, micro-turbines), has many advantages in improving the quality of voltage in rural power grid and solving power supply in remote area or rural areas. The paper investigates a number of practical projects application modes and put forward several different DGS application modes, such as grid-connected DGS, grid-disconnected DGS, micro-grid system, and then analyses the characteristic of these DGs application modes.

The anomalous data identification study of reactive power optimization system based on big data

With the continuous development of smart grid and energy Internet, modern power system is gradually evolved into the one with funnel large amounts of data and calculation of large information systems, which shows the applicability and feasibility of the analysis technology of data mining. This paper puts forward a big data modeling method for the reactive power optimization based on the theory of the large dimensional random matrix. On the basis of it, large dimensional random matrix is disposed, applied with higher dimensional random matrix theory related to the characteristics of abnormal data detection, for judging the existence of abnormal data. If existed, this matrix is used in accordance with Pauta criterion identification to find the abnormal data. At the end of the article, it is verified by analysis examples of its effectiveness and applicability.

Condition evaluation index system of distribution equipments

In this paper, the condition-based maintenance index system for distribution equipments made by State Grid is supplemented by adding the maintenance strategy evaluation indexes, which will be helpful for the maintenance staff to optimize the choice of maintenance strategies. A risk factor is introduced in the security and economy indexes for maintenance strategy evaluation by building function between the equipment failure rate and equipment health state. So the evaluation result of different maintenance strategy, especially for preventive maintenance and Breakdown Maintenance, will be great different, and thus the decision makers with different risk preference are assisted to make their choices. Besides, A service restoration optimization program is implemented during the evaluation, so the outage load caused by maintenance will be divided into two sets: one is the load can be restored with the outage time equals to the switch operation time; another is the load cannot be restored with the outage time equals to the maintenance time. This improvement will make the corresponding evaluation indexes more accurate and hence can distinguish the slight difference of different maintenance strategy. Mathematical analysis shows that the established indexes can be calculated and can provide richer information to assist decision makers, therefore the established index system is practical and effective.

Condition evaluation of distribution equipment based on conjoint analysis

Conjoint analysis is a new evaluation method. It has many advantages, but it has been seldom applied in the field of power equipment condition evaluation. This paper applies conjoint analysis method in the condition evaluation of distribution equipment, and proposes corresponding solutions to the specific problems appeared in implementation process: by refining the rule of consistency check in conjoint analysis, an optimization model with the goal of minimizing the adjustment to the original scores without violating the logical consistency constraint has been built; in view of the problem that too few levels to do accurate evaluation, the concept of membership degree is introduced in the transition process of index scores to levels to improve the evaluation sensitivity to the minor changes of index, making the results more accurate; under the presume of making sure that survey cards have reasonable redundancy, the linear correlation cards are removed in orthogonal experiment to reduce the cards number to improve the efficiency of investigation. The example analysis results show that those proposed solutions are correct and effective.

Sparse Bayesian Learning Using Combined Kernels for Medium Term Load Forecasting

A pattern recognition method based on probabilistic forecasting, Sparse Bayesian Learning (SBL) model is applied for regression in medium term load forecasting. And for the kernel functions chosen, the paper utilizes linear combination principle to construct multiple combined kernel functions, the Gaussian kernel with polynomial kernel and tensor product spline kernel are collected. The parameters of these combined kernels are optimized by Particle Swarm Optimization (PSO). With the training and testing sample data from “2001 world-wide competition of electricity load forecasting”, the results show that all combined kernel models exhibit better accuracy than single kernel models. Besides, probabilistic forecasting results are also given based on the exclusive probability property of Sparse Bayesian Learning.