Research On Energy Supply-Demand Forecast Of Local-area Integrated Energy System

A local-area integrated energy system is a system that includes various types of energy such as electricity,heat,and cold.On the one hand,it can realize the coordination and complementarity of various types of energy including electric energy,heat energy,and cold energy,realize the cascade utilization of different energy,and improve the efficiency of energy.On the other hand,it can realize the interconnection between different energy sources,expands the electricity demand of the society,thereby promote the consumption of renewable energy represented by photovoltaic and wind energy,and alleviate the phenomenon of"abandoning the wind" and "abandoning the light" to some extent.The construction of a local-area integrated energy system is of great significance for promoting the transformation to clean energy and increasing the proportion of renewable energy.Accurate energy demand forecasting is a prerequisite for the planning and design,operation design and energy management of local-area integrated energy systems,and has important practical value.This thesis aims at short-term prediction in local-area integrated energy system,and is divided into demand-side forecasting and supply-side forecasting.Demand-side forecasting refers to the prediction of multiple loads with electricity,heat and cold.Firstly,a detailed analysis of the coupling relationship between single load and multiple loads from multiple time scales of monthly,weekly and hourly levels.Secondly,the K-MEANS algorithm is used to cluster energy-consuming patterns in multi-load scenarios,and the typical daily load curve is extracted.Introce Copula theory to quantitatively analyze the correlation of electric-heat and eletric-cool in different typical days.Then,by plotting the scatterplot of electricity-heat and electricity-cool in typical scenarios,the functional relationship between the loads is fitted,which provides a basis for the extraction of higher-order feature variables.Finally,the extreme learning machine model optimized by the bat algorithm is used to predict different loads.The comparative analysis of the examples shows that the proposed method can effectively improve the prediction accuracy of multiple loads.Supply-side forecasting takes the photovoltaic power forecast as an example to make a short-term forecast of photovoltaic power generation in the local-area integrated energy system.Firstly,based on the Copula theory,the correlation measurement of the factors affecting the photovoltaic power generation is carried out.Secondly,a new method for selecting similar days is proposed for the characteristics of photovoltaic power that is highly correlated with irradiation intensity,that is,selecting the irradiation intensity with the highest nonlinear correlation as the basis for selecting similar days,using K-means clustering algorithm cluster the irradiance intensity sequence of the samples,calculate the center-of-gravity coordinate matrix of each class,then select the sample with the smallest total absolute value distance by calculating the absolute value distance between the daily irradiance to be predicted and the center-of-gravity coordinates of each class.Classes are used as samples for training prediction models.Finally,a random forest model is used for prediction,and a grid search algorithm is used to optimize the number of decision trees and split attribute values in the random forest model,thereby further improving the short-term prediction effect of photovoltaic power generation.