Copula and stacked LSTM network coupled park multivariate load joint prediction method coupled with

Abstract

The invention discloses a Copula and stacked LSTM network coupled park multivariate load joint prediction method. The method comprises the following steps: analyzing nonlinear correlation between cooling, heating and power multivariat loads as well as between the loads and other influence factors such as air temperature by utilizing a Copula theory, selecting input elements for load prediction, carrying out data division on the selected input elements to obtain a training set and a test set, and inputting the training set into a stacked LSTM deep neural network model; training the stacked LSTMdeep learning network model under a Keras environment deep learning framework, and storing weight information trained by the stacked LSTM deep learning network model; loading the trained stacked LSTMdeep neural network model to perform prediction simulation on the test set, and obtaining cold, heat and electric load data in a typical season; and using a mean absolute value error MAPE and a Taylor inequality coefficient TIC to evaluate and predict the cold, heat and electric load prediction results. The stacked LSTM deep learning network model is adopted to predict the load, and the park multivariate load can be effectively and accurately predicted.

Description

technical field [0001] The invention relates to the technical field of comprehensive energy load forecasting, in particular to a method for jointly forecasting multi-element loads in a park coupled with a Copula and a stacked LSTM network. Background technique [0002] With the proposal and development of emerging energy supply and demand systems such as distributed energy, integrated energy, and energy Internet, the dominant position of end users on the demand side, which has always been a passive recipient, has become increasingly prominent. Guided by the concept of "on-demand energy supply, sharing and integration", through in-depth mining of demand-side load characteristics and research on end-user energy consumption behavior, it is possible to understand users' personalized and differentiated service needs, promote interactive regulation of supply and demand, and improve Comprehensive energy efficiency of the system. Especially for regional integrated energy systems wi...

AI Technical Summary

Problems solved by technology

[0003] As a simple dynamic load forecasting method, although the load factor method originated in Japan can realize hourly forecasting of multiple loads, it is only suitable for rough grasp of load characteristics in the stage of energy system planning and design.

Images