| In some remote areas,the use of abundant local energy sources such as hydro,solar and wind energy to build a microgrid is an important way to solve local power problems.The construction of the microgrid can not only meet the local load demand well,but also uses green energy to effectively alleviate the traditional energy crisis and environmental problems.Establishing a hydropower-wind-photovoltaic microgrid scheduling optimization model can not only make the microgrid have better economic and environmental benefits,but also has important significance for the safe and stable operation of the microgrid.However,because microgrids mainly have a large amount of distributed energy sources,among which wind and photovoltaic power generation are greatly affected by the local environment.Improving the prediction accuracy of wind and photovoltaic power generation will also improve more accurate data for the safety and stability of microgrids.This paper focuses on the power prediction,load prediction,and optimal scheduling of microgrids.The main work is as follows:(1)This paper proposed a similar day selection method considering distance similarity and trend similarity based on the influence of weather factors and it is applied to the power prediction of photovoltaic power generation.The cuckoo algorithm is introduced into the searching process of Elman neural network to improve the global searching,performance,structure and parameters of Elman neural network.The paper proposed a prediction model of cuckoo algorithm optimized Elman neural network.The results show that the photovoltaic power prediction model of cuckoo optimized Elman neural network in sunny,cloudy and rainy days is better than that of traditional Elman neural network.The error of the proposed model is reduced by 4%,4.44% and 2.72% respectively,and compared with the traditional Elman wind power forecasting model,the error of the power forecasting model optimized by cuckoo is less than 0.29%.(2)Considering that the error of load forecasting model is a Markov chain,the paper proposed an improved long-term and short-term memory network load forecasting model base on the deep learning theory in order to improve the accuracy of the predicted value.The results show that compared the error of the the long-term load forecasting model,the error of improved load forecasting model is less than 1.4428%.(3)It is inconvenient for microgrid scheduling due to the error between wind power prediction and photovoltaic power prediction.In this paper,the mathematical programming model of microgrid optimal scheduling is established.The uncertain constraints in the modelare dealt by chance constraint method,which is transformed into deterministic constraints.The improved sequential quadratic programming method is used to solve the planning problem of microgrid optimal scheduling.The results show that with the decrease of confidence,the total cost decreases from 2565.5 yuan to 2079.3 yuan.(4)The fluctuation of load and renewable energy power generation in a single period will lead to a large degree of imbalance in source and load consumption.Therefore,a two-stage dispatching model of the hydro-wind-solar microgrid is established in consideration of the source-load dissipation problem.Based on the shortcomings of the original particle swarm algorithm,a particle swarm quadratic sequence algorithm with expanded search was adopted as the solution method of the programming problem.Comparing the indicators of the two-level dispatch and the first-level dispatch,the load fluctuation rate of the two-level dispatch during the wet season has decreased by 14.22%,the total cost has decreased by763.52 yuan,the abandoned power has decreased by 133.1k W,and the load fluctuation rate of the dry season has decreased by 6.89 %,The total cost was reduced by 691.49 yuan.It can be seen that the two-stage dispatching model can make the microgrid economically better and the renewable energy utilization rate higher. |
PDF Full Text Request