Research On Risk Analysis And Control For Power Communication Networks

With the development of Smart Grid, the effect of the information and communication system is becoming more evident. Power communication networks (PCNs) are the key components of the information and communication system, and any operational security event of PCNs will influence the operational safety, stability and economy of electric power system. Therefore, the operational risk research of PCN has been a particularly important subject. Network risk analysis can provide the important theoretical basis for network optimization and design, and network risk control can decrease network risk effectively. Therefore, network risk analysis and control can improve the network security effectively. Based on the analysis of existing related works, research on PCN vulnerability analysis, risk evaluation and routing-based risk control is carried out in this paper.As the asset value metrics of electric power services, the service importance value is the basis of PCN risk analysis and control. Existing calculation methods of electric power service importance value (EPSIV) are based on expert grading which lead to uncertain results and influence the accuracy of all research relating to EPSIV. A new method of calculating EPSIV, called feature index method (FIM), is proposed in this paper, which is based on objective elements instead of subjective elements in order to eliminate the uncertainty in results. A service layer network model is created using the EPSIV calculated by FIM, and based on the model, PCN vulnerability is analyzed under different link failure modes. According to the analysis results, service paths are adjusted which reduced the network vulnerability.There are some problems in existing PCN risk evaluation methods, e.g. scattered indicator, unclear results, and lack of integrity. In order to solve the problems, a novel multi-layer risk evaluation method combined with complex network theory is proposed in this paper, which defines edge betweenness, the quantity of EPSIV, and the traffic carried on an edge as the edge importance values on physical topology layer, service layer, and transport layer respectively. Then, a new PCN risk evaluation indicator, the edge cross-layer entropy, is presented and the effectiveness of the indicator is proved by simulation results.Based on the above vulnerability analysis and risk evaluation, routing optimization is chosen as the PCN risk control method. In order to solve problems such as less network layer being considered and insufficient applicability of routing models, a cross-layer risk-aware routing model is proposed, which takes into account the influence of routing on PCN service layer, transport layer, and physical topology layer. A novel chaotic immune multi-constraint routing algorithm (CIMCRA) is presented to control the network dynamic risk by finding paths for dynamic arriving service requests. In this algorithm, dynamic vaccination and free mutation is combined to raise convergence seed and ensure global optimization. Taking account of the service requests that have strict time delay constraints, a hybrid routing strategy is proposed based on CIMCRA, and the effect of network risk control using the hybrid routing strategy is analyzed under different attack modes. For improving path code and decode efficiency, a path generation method based on dynamic adjacency matrix is presented, which codes paths with natural numbers and can directly obtain available path solutions while having a simple decode procedure. The path generation method can improve the efficiency of all evolution algorithms used in routing optimization.During the PCN design process, the designer needs to assign routes to all service requests simultaneously according to network topology and the static service matrix. Furthermore, multiple routing schemes are needed in order to make appropriate choices. A chaotic immune multi-objective routing algorithm (CIMORA) is proposed in this paper, which codes paths with fixed length by assigning random weight to every edge, and decodes paths by Dijkstra algorithm to realize the route assignment of all service requests. In this algorithm, a biogeography-based optimization method and a dynamic clonal method are introduced to increase the convergence speed, and a clustering algorithm is also introduced to enhance the ability of searching optimal solutions. The performance of CIMORA is proved superior by comparison with two other existing multi-objective optimization algorithms.In traditional evolutionary algorithms, the output of a single run may be optimal, suboptimal or even poor when population size and the number of iterations are fixed because of the random number based crossover and mutation strategy. This uncertainty is not acceptable in PCN. In order to solve the problem and ensure the certainty of the outputs of CIMCRA and CIMORA, chaotic sequence values are employed instead of random numbers in all operators using random numbers. Therefore, chaotic sequences with a high degree of randomness are needed to maintain the search capability of the algorithms, but the degree of randomness of chaotic sequences generated by traditional chaotic systems is not high enough. A new five-dimensional hyperchaotic system is proposed in this paper, and its existence and complex nonlinear dynamics behavior are proved by chaotic attractor, Lyapunov index, etc. Then, the randomness of the sequences generated by the five-dimensional hyperchaotic system mapping is tested resulting in a higher degree of randomness than other traditional pseudorandom sequences, ensuring that CIMCRA and CIMORA can produce highly certain results and suggesting that they have good search performance.