Prediction And Risk Management Of Power Grid Icing Disaster Based On Big Data

With the global continuous low temperature,rain and snow,frost,fog and other weather disasters occur frequently,power grid disasters caused by bad weather are increasing.The hazards caused by transmission lines icing are more and more serious,which may cause flashover and even result in the fittings damage,transmission lines broken,rod and tower collapses and other serious accidents.The freezing disaster has become the common problem facing many of the countries' power grid.The United States,Canada,Russia,France,Iceland and Japan all have experienced severe ice accidents.As one of the countries seriously affected by extreme weather disasters,China is affected by the big climate and micro topography and micro meteorological conditions.Thus the frozen disasters occur frequently in our country,which poses a serious threat to the safe operation of the power transmission lines,causing huge economic losses to the society.Especially at the beginning of 2008,the freezing weather caused widespread damage to the power system in China.The power towers collapses,lines breaking and dancing,icing and flashover,and other disasters occurred in the power grid of Hunan,Jiangxi,Zhejiang,Anhui,Hubei and other provinces.With the great icing intensity,this disaster lasted for a long time,and affected a wide range of areas.So the extent of damage to the power grid facilities have reached the highest in history.Therefore,the researches of icing growth law,effective icing prediction models and icing risk assessment methods for power transmission lines,are of great scientific significance and application prospect to prevent freezing disasters and guarantee the safe operation of power system.Aiming at the problem of transmission lines icing,this paper identified the influence factors of icings,starting from the study of lines icing mechanism.And based on the preprocessing model of ice data with Spark large data running platform,a quantitative analysis model of influencing factors of transmission line is established.On this basis,combined with the Spark processing platform,the intelligent transmission line icing prediction model and icing risk evaluation method are proposed.Thus,the regional power grid icing disaster risk management system is constructed,and a number of empirical research are carried out.The main research results and innovations are as follows:(1)Based on the dynamic identification and measurement of influence factors for transmission line icing,this paper has established the influence factors database and comprehensive quantification model for the line icing in Spark platform.The study for the identification of factors are carried out from three aspects ofmeteorological,environmental factors and line parameters,and based on this research,the three aspects of the depth,relevance and hierarchy of the influencing factors are modeled and measured.In the construction of models,the depth coefficient is used to analyze the depth of influence,the Person correlation coefficient model and grey correlation degree method are used to bring out the correlation of influencing factors,and the analytic network process(ANP)model is used to quantify the weights of the factors.And then,by means of the weighted average of three aspects,the comprehensive measurement coefficient of influencing factors is formed.Furthermore,this paper has constructed the comprehensive quantification model based on infinite deep convolutional neural networks.Through the establishment and operation of Spark platform,and by the parallel processing,training and learning for the icing big data,the comprehensive measurement of the influencing factors is realized.Finally,the effectiveness and feasibility of the comprehensive measurement model of influencing factors are verified by the example calculation.(2)This paper has studied the intelligent prediction model of transmission line icing,and built a icing prediction method based on discrete wavelet consistent rate and QFA-W-LSSVM model on the Spark running platform.In the feature selection of discrete wavelet inconsistent rate,the icing data signal is decomposed and reconstructed by discrete wavelet transform.Then,through calculating inconsistency rate of high and low frequency decomposed signal,the best feature subset is obtained by comparison.In the QFA model,encoding and update for the fireworks individual location are fulfilled to improve the searching performance though the quantum encoding and quantum rotation gate.In W-LSSVM model,the LSSVM method is mainly improved by the horizontal weighting for input factors and vertical weighting for training sample to strengthen the learning and training ability.Place the QFA-W-LSSVM overlay prediction model on the Spark platform,and the model is validated from the icing data warehouse.The simulation results show that the proposed method is able to improve the prediction accuracy and effectiveness effectively,and it's feasible and effective.(3)The risk evaluation index system and evaluation method of transmission line icing risk are studied,and a risk evaluation model based on Dynamic Bayesian inference adaptive fuzzy inference system(DBN-ANFIS)is established in Spark platform.In this paper,five risk levels of transmission line icing state are divided,and the risk assessment index system is constructed based on the three aspects of micro-meteorological parameters,environmental factors and line parameters,respectively.In the icing risk assessment model,through the combination the icing timing sample information,the dynamic Bayesian inference constantly improves the previous experience and knowledge of the system for each layer of ANFIS model byregarding the experience as prior information.And the Bayesian inference can also constantly revise the previous results to improve the generalization ability of the algorithm.In order to realize the effective processing and identification for the risk assessment under the icing big data,the DBN-ANFIS risk evaluation model is placed on the Spark platform.And the test results show that the intelligent risk assessment based on DBN-ANFIS model on the Spark platform has good applicability and accuracy,and the algorithm is more stable and more fit.(4)The economic evaluation of transmission line icing disaster is studied,and the evaluation index system of the economic loss caused by ice breakage and the evaluation index system of the economic loss caused by icing blackout in regional area are put forward.In this paper,the AHP and entropy weight method are combined to get the weight model.Then,the fuzzy comprehensive evaluation method is also used to evaluate the line breakage economic loss and regional blackout economic loss which both caused by icing disaster.Lastly,two cases are applied to verify the effectiveness and feasibility of the proposed economic evaluation index system and evaluation method.The verification results show that the evaluation system and method are feasible and practical.(5)Based on the identification and quantification of influencing factors,icing forecast and icing state assessment in Spark platform,the paper has set up the risk management system of icing disaster in regional power grid based on the big data.In this paper,the icing risk management system was composed of four aspects: power grid icing disaster management organization,online early warning system,emergency response system and emergency response plan.In the organization of ice disaster management,the regional and local major emergency management organizations were mainly established.In the icing disaster early warning system,the power grid structure,icing monitoring and information collection system,communication system,data center,central processing unit,authorization system and operating system was respectively designed.In the icing disaster emergency support system,the communication and information security system,emergency team security system,emergency material equipment support system,technical resources guarantee system and other security systems was established respectively.In the icing emergency treatment plan,the IDNN model,the QFA-W-LSSVM model and the DBN-ANFIS model are carried out to make the warning of icing disaster in regional grid.According to the icing warning and response level,the process of early warning and the emergency response are described in detail.This paper has put forward the regional power grid icing disaster risk management system.And it could help to prevent power grid staffs to improve work efficiency and improve the safety and stability of power grid,and also provide the reference for the establishment ofregional power grid icing disaster emergency plan,which has wide applicability.