Research On The Method Of The Level Assessment Of Oil Spill Pollution On Bohai Offshore Platform

The oil pollution is considered one of the pollutants that are larger harmful to the marine environment; Assessment index system of oil spill for the offshore oil platform was established by analyzing the influenced factors related to the degree of oil pollution in the paper. Analytical Hierarchy process (AHP) was used to calculate the weights of evaluation indexes;and the membership function and the membership degree tables were established severally; and a second order fuzzy comprehensive evaluation model for the degree of oil pollution was set up finally. Every assessment index was divided into several grades and the Rand function was used to generate enough samples.A more reasonable network structure was established and a BP neural network model of the degree of oil pollution was finally set up.The two models provided the new method for judging the pollution level of oil spill and also could provide certain technical support for the emergency decision-making department to take the emergency measure.Assessment index system of oil pollution for the offshore oil platform was established by analyzing the influenced elements related to the degree of oil pollution. The system includes five aspects which are oil spillage,oil properties,oil spill position,meteorological factors and hydrological factors;and it is composed of static index and dynamic index:toxicity,persistence,flammability,viscosity of crude oil; sensitive resources,the distance offshore,coastline types;oil spillage,wind speed,visibility,tidal speed,tidal direction,surface water temperature,wave height.Combine triangular and trapezoidal curve to establish membership function for oil spillage and sensitive resources, and then theirs membership degree could be determined. The left indexes are toxicity,persistence,flammability,viscosity of crude oil,the distance offshore,coastline types,wind speed,visibility,tidal speed,tidal direction,surface water temperature and wave height, whose membership degree could be determined by structuring membership tables. The weights of the first and the second grade indexes were calculated by Analytical Hierarchy process.The weight of oil spillage,oil properties,oil spill position,meteorological element and hydrological element is 0.39,0.10,0.12,0.20,0.19 respectively. The weight of toxicity,persistence,flammability,viscosity of crude oil is respectively 0.34,0.29,0.20,0.17;and the weight of sensitive resources,the distance offshore,coastline types,wind speed,visibility,tidal speed,tidal direction,surface water temperature and wave height is 0.41,0.37,0.22,0.83,0.17,0.42,0.13,0.10,0.35.On the basis of membership degree and weights of evaluation indexes, the second order fuzzy comprehensive evaluation model for the degree of oil pollution could be set up. The model provided a new method for judging the pollution level of oil spill scientifically, and it avoided one-sidedness of subjective judgment depended on professional.Due to lack of objective practical samples, every assessment index was divided into five grades;Rand function was used to generate 2300 samples within each classification standard and 100 samples as test samples were extracted stochastically. Set 6 neurons in the hidden layer,epochs=3000,goal=0.00001 and trained the BP neural network respectively with six used commonly improved BP algorithm. Contrasted training results of the network, and then selected L-M algorithm whose performance was the best. The scope of number of neurons in the hidden layer was determined by empirical formula, and then the best appropriate quantity of neuron in the hidden layer was 8 through'cut and try method'.According to input samples,output samples,the number of network layer,the number of neuron of each layer,algorithm,transfer function,performance function and so on, The BP neural network was generated by newff function, and the program was programmed in MATLAB6.5 interface.Trained the BP neural network, when the training error was smaller than the goal error, the training stopped.Test samples were simulated to measure the performance of the network, and the result proved that the BP neural network had very strong generalization capability.Oil spill accident was evaluated by the fuzzy comprehensive evaluation model and the BP neural network model,and the judgment results were consistent, and the level of oil pollution could be assessed correctly by means of the two kinds of model. Once the marine petroleum platform occurred overflows accident, the level of oil pollution could be evaluated by online monitor system and the foundation database parameters.The level of oil pollution was mastered promptly, and this provided the reliable basis for taking the emergency measure and had the strong usability.