Research On Modeling,simulation And Optimal Operation Of Power Systems With High Dimensionality And Uncertainty

Developing renewable energy sources such as wind and photovoltaic power is of significant importance to safeguard our national energy security,upgrade energy industry and realize the goal of low-carbonization and clean production.Due to the inherent nature of uncertainty,intermittent and volatility,it is not easy to forecast the wind/PV power and decide the dispatching plan accurately.With the rapid growth of renewable energy integration,power systems will be faced with great challenge of uncertainties,where the situation gets more complicated as the input variable dimensions increase dramatically.Therefore,the future power grid will be gradually revealing two major features of high-dimension and uncertainty.At that time,traditional deterministic methods will be difficult to adapt to the power system operation and scheduling under such uncertain environment.On this background,this paper aims to address the aforementioned issues from the following three aspects: probabilistic modeling,static operation and optimal scheduling.Based on the theory of uncertainty analysis,the main research findings obtained are summarized as follows.(1)A generalized hybrid method for probabilistic load flow calculation is developed based on an improved Monte Carlo simulation approach.This method combines the characteristics of both simulation and analytical methods in traditional probabilistic load flow studies.According to the completeness degree of historical measurement data of new energy stations,we employ either the Gaussian mixture model or empirical distributions in probabilistic modeling.The correlated output samples of renewable energy are generated by uniform design sampling and Cholesky decomposition.Moreover,multi-linearization is applied to reduce the truncated error as well as time consumption.As for the input variables represented by Gaussian mixture models,Gaussian component combination method is introduced to optimize the solution of probabilistic power flow analysis.(2)A numerical method for probabilistic load flow with multiple correlated random variables is proposed.In order to circumvent the computation-intensive simulation in high dimensional uncertain systems,we combine the dimensional reduction integral method and Gaussian integration formulas to simplify the computation of multivariate statistical moments.Nataf transformation is also used to handle the correlation between input random variables.Finally,the probability distributions of output variables can be well reconstructed by C-type Gram-Charlier series expansion with the calculated moments.(3)A stochastic risk assessment approach is presented with the consideration of primary frequency regulation.The method is based on the outcomes of probabilistic load flow using analytical method,where the distributions of nodal voltage,branch power flow and system frequency can be obtained by just one calculation.The system frequency is taken as an additional state variable considering the static characteristic of frequency regulation.Then,synthesis risk indices that represent both the likelihood and consequences of uncertainties caused by renewable resources are quantitatively employed from both the perspective of component and system-level.Thereafter,the online static security assessment of power grid is realized with potential weak parts been identified.(4)A comprehensive way of modeling the short-term wind power forecast error is proposed.Firstly,we adopt a generalized mixed skewed model to fit the error distribution with biasness,long tail and multimodality according to the statistical characteristics of day-ahead forecast error for a single wind farm,which exhibits the merits of flexibility,adaptability and easy implementation.Then,stochastic dependence between real and forecasted wind power is captured using conditional probability model based on the Copula theory.This method can be further generalized to high dimensional cases of multiple wind farms under Pair-copula construction.Finally,the effectiveness of the proposed model is demonstrated in the application of shot-term probabilistic load flow and energy storage optimization.(5)To circumvent the problem of transmission congestion in the real-time dispatch,a graph-cut based dynamic reserve zoning approach is proposed and an energy and reserve co-optimization model is established with the proposed zonal reserves.The power system is represented by an undirected weighted graph in which the edge weights are defined based on the congestion risks of each line.The risk values are obtained by calculating the probability distribution of line flows considering the wind power uncertainty and line outage contingencies.Gomory-Hu algorithm is employed in solving the minimum cut problem to assure the resultant zones with minimum congestion risks.Thus,the intra-zonal congestions are mitigated and the utilization of allocated zonal reserves can be maximized.It is verified that the proposed method could effectively reduce wind curtailment and load shedding and thus,improve the economy and reliability of power grid operation.(6)An adaptive approach for quantifying transmission reserves is proposed by taking into account the risk of wind power uncertainty-induced congestion.A bi-level stochastic programming model is developed,in which predictive transmission reserves are incorporated in the upper level unit commitment model.The statistical features of ex-post line flows are calculated using new point estimate method in the real-time re-dispatch,which are then sent back to the upper level UC to adjust reserve requirements dynamically.Then,the bi-level framework is recast into a single level model based on the Karush-Kuhn-Tucher optimality conditions in the lower model,and is finally transformed into a mixed integer programming formulation for computational tractability.The proposed method is effective in alleviating transmission congestion,as well as improving wind accommodation and reserve deliverability.