Modeling Of Wind Power Prediction Error Distribution And Research On Wind-Fire Joint Dispatch Considering Uncertainty

With the increasing exhaustion of traditional energy and the proposal of dual carbon targets,the rapid development of new energy power generation technology.As a relatively mature new energy power generation technology,wind power has received widespread attention;However,the inherent volatility,randomness,intermittency and nonlinearity of wind power make it very difficult to predict wind power deterministically.Compared with wind power certainty prediction,wind power probability prediction has greater application value in power system risk assessment,decision-making and grid optimization and dispatching.At present,the capacity of wind turbine assembly machines in the power system is increasing,reducing the wind curtailment rate of wind power generation and improving the absorption capacity of wind power generation is becoming more and more important,so it is more and more important to make appropriate probability prediction of wind power output,and on this basis,it is more and more important to propose a grid economic dispatching model that considers the uncertainty of wind power.To this end,a variable modal decomposition(VMD)-extreme learning machine(ELM)/autoregressive integrated moving average(ARIMA)wind power certainty prediction method is proposed.First of all,the historical time sequence of the wind power is used to decompose,and then the decomposed series of the intrinsic mode function(IMF)is established the ELM predictive model,the residual is established the ARIMA predictive model,and then the reconstruction obtains the wind power determination predictive value.The results show that after the use of VMD to decompose the historical time sequence of wind power,different predictive models are established for different components,compared with the results obtained by establishing the same prediction model for different components,the mean absolute error(MAE)is about 5.9 lower,the root mean square error(RMSE)is about3.7 lower,and the relative(R)is about 0.38 higher.Then,a wind power predictive error density fixing model with an adaptive diffusion kernel density estimation(ADKDE)is proposed.First,the wind power predictive error is obtained according to the wind power certainty prediction,and an ADKDE model is established to fit the wind power predictive error distribution;then under different predictive methods,different installation capacity,and different sampling periods,analyzing the fitting effect of ADKDE to wind power predictive error is compared with other comparative models proposed in this paper based on the R,MAE,RMSEand relative entropy.Based on the ADKDE fitting model,the wind power interval prediction is carried out,that is,according to the results of wind power certainty prediction,an ADKDE model is established for the wind power predictive error.Finally,according to the given confidence level(CL)and the cumulative distribution function corresponding to the predictive error of wind power,and the upper and lower limits of wind power interval prediction are obtained by combining the predicted value of wind power.The results show that the proposed ADKDE model is used to predict the wind power interval,and the wind power interval coverage is higher and the bandwidth is narrower.Finally,an adaptive diffusion Gaussian kernel density estimation fitted wind power predictive error is proposed to a wind-fire joint optimization scheduling model.First of all,the iterative self organizing data analysis techniques algorithm(ISODATA)is used to segment the wind power predictive value and its corresponding predictive error,and then adopts an adaptive diffusion Gaussian kernel density estimation to fitting the segmented wind power predictive error.Opportunity constraint planning is used to consider the positive and negative rotational standby capacity probability constraints of wind power and load forecast error in the constraint as a whole,and finally transform them into a deterministic constraint of net load forecast error.Finally,the GUROBI solver is used to solve the three scenarios,different wind power predictive error distribution models and different CL.The results show that compared with the comparison model,the proposed model reduces the cost of rotating spare capacity is reduced by 6.71%,the environmental cost with carbon emissions is reduced by 20.4%,and the total power generation cost is reduced by 2.98%.