Research On Control Strategy Of Peak Cut And Valley Filling In Microgrid Energy Storage System Based On Load Forecasting

With the development of China's economic level and the improvement of people's living standards,the peak-to-valley difference of the power grid has gradually increased,and in recent years,China has vigorously developed the randomness and reverse peaking characteristics of distributed energy represented by photovoltaics.The increasing peak pressure has brought great challenges to the safe and stable operation of the power grid.Microgrid,as a solution for connecting distributed power sources to the power grid,coordinates the originally dispersed energy sources and maximizes energy use.However,the distributed power system is greatly affected by the outside world.When the outside light is weak,the power generated by the microgrid is reduced.If it cannot be replenished,the microgrid will lose its balance.The emergence of energy storage system provides a new solution to the idea,using the energy storage system for peak and valley filling can effectively reduce the peak and valley difference between day and night,smoothing the load.Therefore,the main research content of this paper is the peak-and-valley control strategy of microgrid energy storage system based on load prediction.First,a brief analysis of the user's electricity consumption behavior is made to make its electricity consumption characteristics more obvious.On this basis,a load prediction method based on set empirical mode decomposition-fuzzy entropy and stacking integrated learning is proposed.The problem of modal aliasing in the process of decomposing feature extraction,this paper uses set empirical modal decomposition to improve the accuracy of the decomposition;the complexity of each component is calculated using fuzzy entropy,and the similar components of fuzzy entropy are combined to avoid components Excessive calculation complexity and error accumulation;the high-frequency components obtained by decomposition are predicted using LSTM,the low-frequency components are predicted using SVR,and stacking integrated learning is used to input the prediction results with weather data and humidity data into the fully connected layer to obtain the final The prediction results are compared with the prediction results of LSTM and BP neural network.Secondly,based on the daily load forecast curve,an optimized objective function,related constraints,and evaluation indicators of peak shaving for battery energy storage to participate in peak shaving of the power grid were established.Peak shaving was carried out using constant power and power difference control strategiesValley,and compared the effects of the two methods from four aspects: peak-valley difference,peak-valley coefficient,peak-valley difference and standard deviation.Finally,this chapter determines the structure of the microgrid and provides support for the load.It is determined that the DC microgrid used in this article contains two distributed power sources: photovoltaic cells and energy storage batteries.The working principles of the two distributed power sources are introduced,corresponding mathematical models are established,and the output characteristics of the photovoltaic cells are analyzed.Incremental method was used to track the maximum power point,and finally verified by simulation model.The basic structure of the energy storage system is determined.The energy storage battery system is connected to a bidirectional DC/AC converter through a bidirectional DC/DC circuit boost,and then connected to the power grid through a filter and a transformer.The control strategies of the bidirectional DC/DC converter and the bidirectional AC/DC converter are studied separately.The bidirectional AC/DC converter uses a double-loop control structure of voltage outer loop current inner loop,and the bidirectional DC/DC converter introduces virtual DC motor control.The working principle and control strategy are expounded,the small signal model is established to analyze the virtual DC motor's control parameters virtual inertia J and damping coefficient D,and the comparison and simulation verify that the introduction of virtual DC motor control compared to traditional The control has better ability to suppress voltage fluctuations while realizing voltage rise and fall.