Estimation Of Wind Turbine Parameters Using Adaptive Neuro-Fuzzy Algorithm

Renewable and sustainable energy resources,especially wind energy has received considerable attention for power generation.Wind energy has become a best alternate of traditional fossil fuel because of its better efficiency,cost and reliability.The energy produced by wind is clean with no emission of greenhouse gasses which helps in reducing global warming and environmental pollution.Therefore,advanced control techniques have been applied to improve its performance.This thesis investigates the implementation of hybrid intelligent learning based adaptive neuro-fuzzy algorithm to estimate different key parameters of variable-speed wind turbine system.The research mainly includes the following topics:(1)The probabilistic wind speed distribution and evaluation of wind energy potential are very important factors while selecting a suitable site for wind turbine installation.Wind farm designers use Weibull wind speed probability distribution function(PDF)to analyze the wind speed characteristics and variations at a specific site.This thesis proposes a hybrid intelligent learning based adaptive neuro-fuzzy inference system(ANFIS)to accurately estimate the Weibull wind speed PDF.The artificial neural network(ANN)trains the parameters of fuzzy membership functions(MFs)using hybrid optimization method.A number of experiments are conducted for different type of input-output MFs and it is observed that best results are obtained by assigning Gaussian MFs to the input variable and linear MFs to the output variable.The results of proposed neuro-fuzzy system are compared with five well-known numerical methods.The results indicate that the ANFIS out performs all numerical methods and provides the best fit of measured Weibull distribution curve.The Weibull parameters are further utilized to calculate the wind power density.The considered wind resources are observed to have moderate wind energy potential.Four small scale wind turbines with rated power 50 kW,100 kW,150 kW and 250 kW are taken into consideration to select the most efficient and economically viable wind turbine for available wind resources.The average electrical power,the annual energy produced and the capacity factor confirm the economic viability of 50 kW and 100 kW wind turbines.(2)The precise measurement of wind speed is a crucial task and has huge impact on wind turbine output power,safety and control performance.The conventional wind speed sensors measure wind speed only at single point which does not reflect the effect of wind speed on whole wind turbine rotor.This thesis proposes an adaptive neuro-fuzzy algorithm for online estimation of effective wind speed from instantaneous values of wind turbine tip speed ratio(TSR),rotor speed and mechanical power.The system is trained using least squares estimator method and back propagation gradient descent method.It is observed that bell-shaped fuzzy MFs provide the best results.The estimated value of effective wind speed is further utilized to design the optimal rotor speed estimator for maximum power point tracking(MPPT)of variable-speed wind turbine(VSWT).Both estimators are implemented in MATLAB and their performance is investigated for national renewable energy laboratory(NREL)offshore 5 MW baseline wind turbine.The simulation results show the effectiveness of proposed method.The proposed scheme is computationally intelligent,easy to implement and more reliable for fast estimation of effective wind speed and optimal rotor speed.(3)Power coefficient is the measure of wind turbine efficiency.The efficiency of a wind turbine varies with operating TSR.Power coefficient is a nonlinear function of operating TSR and pitch angle.Power coefficient reaches to its maximum value at optimal value of TSR.When wind speed is below the rated value,the VSWT is operated at its maximum power coefficient for maximum power extraction.Therefore,accurate estimation of wind turbine power coefficient and TSR is very important to optimize its operation.This thesis proposes an adaptive neuro-fuzzy algorithm to accurately estimate the real time value of power coefficient and TSR for NREL offshore 5 MW baseline wind turbine.The least squares algorithm is used to train the consequent parameters in forward pass and back propagation gradient decent algorithm is used to train the premise parameters in backward pass.During the training process,the ANN adjusts the shape of MFs by analyzing training data set and automatically generates the decision making fuzzy rules.The performance of proposed TSR estimator is compared with conventional multilayer perceptron feed-forward neural network(MLPFFNN).The results verify the effectiveness of proposed method.