Research On Smart Grid Power Generation Prediction And Dispatching Management Method For Renewable Energy Field

Inescapably,renewable energy must be vigorously advanced to address the current dearth of fossil energy and environmental contamination.Representatives of renewable energy,such as wind energy and solar energy are essential in the domain of new power generation and grids containing renewable energy.The output of these sources is contingent upon weather conditions,with meteorological factors such as temperature,wind direction,wind speed,and sunlight influencing the output.Thus,the output power of renewable energy generation has been altered.The integration of renewable energy generation into the grid is impeded by the instability of its generation due to weather sensitivity.To address these issues,it is essential to accurately forecast the output of renewable energy power generation and to implement sensible scheduling tactics to maximize its potential and ensure the grid’s stability.Although many scholars have paid attention to these problems,these approaches still have some shortcomings.For instance,in the field of renewable energy generation power forecasting,the existing algorithms mostly use a single model,which is difficult to deal with the non-linearity in the data relationship,and most of the existing algorithms only consider the short-term dependence between data,which affects the accuracy and stability of the model.In the power system containing renewable energy,the existing economic dispatch models ignore the cost of abandoning wind and solar,which is not beneficial to the consumption of renewable sources of energy.This thesis has devoted itself to the forecasting of wind power energy and the multi-objective scheduling problem of power grid systems with renewable energy,in order to address the issues previously mentioned.The main contributions are as follows:(1)In order to improve the prediction accuracy of wind power generation,an ensemble model is presented in this thesis.The model contains two sub-models,and the dynamic fusion mechanism is used to integrate the models,which compensates for the limited generalization capacity of a single model.It is proved by experiments that compared with other algorithms in existence,the algorithm in this thesis has obtained satisfying results in the assignment of wind power forecasting.(2)In order to capture the long-term dependencies between data,one of the submodels uses a gated recurrent unit with a long-term memory function as an encoderdecoder to extract feature information in time-series data,and the other sub-model adds self-attention to the temporal neural network,which enables the model to focus on those features that contribute more while realizing global timing relationship modeling.It is proved by experiments that compared with other algorithms in existence,the algorithm in this thesis can decrease the cumulative error and enhance the forecast accuracy.(3)For the dispatching problem of power grids containing renewable energy,this thesis not only considers wind power generation but also photovoltaic power generation.At the same time,reducing wind and light curtailment is one of the optimization goals,which promotes the consumption of renewable energy while minimizing operating costs.And use the deep deterministic policy gradient algorithm and environment interaction to get the optimal scheduling strategy.The final experimental results prove that the algorithm can adapt to the random fluctuations of renewable energy generation output and load.Compared with other algorithms,the algorithm has stronger adaptability and lower scheduling cost.