Spiking Neural P Systems And Their Applications In Fault Diagnosis Of Electric Power Systems

When faults occur in electric power systems, a large amount of alarm information will pour into the control center. The abundant information provides various references for fault diagnosis. Howerver, it also makes supervisors too busy to handle timely. Meanwhile, due to the influence of factors such as technology, equipment and environment, the alarm information is likely to exsit uncertainties such as information error, loss and distortion. Thus,computers and artificial intelligence technology are needed when analyzing and processing these alarm information in fault diagnosis of electric power systems. So, information processing ability of fault diagnosis methods is very important. Spiking neural P systems(SNPS) is a special type of neural-like P systems inspired by communication mechanisms between biological neurons. Its computing models are dynamic, discrete, distributed and parallel ones with powerful computing ability and information processing capacity,which are suitable for solving fault diagnosis problems of electric power systems. Therefore, from the way of model abstraction, algorithm design and practical examples and based on the framework and theory of membrane computing and SNPS, this paper focuses on improving the information processing and self-adaption ability of SNPS, and proposes SNPS application models to handle fault diagnosis in electric power systems.Firstly, based on the demands of knowledge representation and knowledge reasoning in fault diagnosis of electric power systems and considering the function and structures of actual biological networks, this paper designs SNPS application models from the way of real number fuzzy reasoning and proposes two kinds of fuzzy reasoning spiking neural P systems with real numbers (rFRSNPS), in which parameters such as output weights (synaptic weights), firing thresholds and information prejudging parameters are considered and definitions of neurons, synapses, spikes and firing rules are redefined. The proposed rFRSNPSs called weighted fuzzy reasoning spiking neural P systems with real numbers(rWFRSNPS) and weighted prejudging fuzzy reasoning spiking neural P systems with real numbers (rWPFRSNPS), respectively. Then, reasoning algorithms of rWFRSNPSs and rWPFRSNPSs are designed and fuzzy production rules are modeled based on them,respectively. If the truth values of input neurons and rule neurons are given, then the proposed algorithms can automatically reason out the truth values of other proposition neurons in an rFRSNPS to fulfill the knowledge representation and reasoning. The intuitive graphical modeling process and simple matrix reasoning computing are quite adaptive to complex knowledge representation and reasoning. On a background of transmission network fault diagnosis, this paper designs layered diagnosis models based on rWPFRSNPS for main sections in power systems and proposes a fault diagnosis method based on rWPFRSNPS in which a temporal information processing method based on cause-effect networks is considered. These diagnosis methods result in a decrease of matrix dimension, a release of calculation burden and an improvement of model adaptive ability.In order to improve the ability of SNPS in processing uncertain and imprecise information and considering the actual existence situation and real communication process of knowledge and information, this paper designs SNPS application models from the way of fuzzy number fuzzy reasoning and proposes fuzzy reasoning spiking neural P systems with trapezoidal fuzzy numbers (tFRSNPS), in which trapezoidal fuzzy numbers are considered and definitions of neurons, spikes, ways of pulse accumulation, firing rules and firing conditions are redefined. Then, a reasoning algorithm is designed and fuzzy production rules are modeled based on tFRSNPSs. This paper also proposes a fault diagnosis method based on tFRSNPS in which fault fuzzy production rule sets for main sections in power systems are given. The introduction of trapezoidal fuzzy numbers guarantees that SNPS has a strong ability to express and reason fuzzy knowledge without prejudging the inputs and decreases the fault diagnosis complexity.Optimization spiking neural P systems (OSNPS), independent of evolutionary operators in evolutionary computation to achieve individual evolution, is a kind of algorithm with superior performances. This paper employs binary strings to reprensent chromosomes(individuals) and makes the first attempt to solve fault diagnosis of power systems by using the idea of optimization methods in the framework of MC and proposes a fault diagnosis method based on OSNPS.To verify the practical application performation of SNPS, this paper discusses the applications of rWFRSNPS, rWPFRSNPS, tFRSNPS and OSNPS in fault diagnosis of transimission networks,where the 14 bus power system and 220kV local power system are used to test their feasibility and effectiveness. The experimental results on several cases show that both the proposed three kinds of fuzzy reasoning spiking neural P systems and OSNPS produce prospective diagnosis results with different characteristics.