Family Load Forecasting Method Based On Markov Chains

With the development of the smart grid technology has put forward new requirements such as informationization and interaction between supply and consumption side.Therefore,it is of a great significance to realize the load forecasting of family load.It is much more difficult to do load forecasting on the consumers' layer than the system's layer.Meanwhile,nowadays,there is little research concerns load forecasting method towards a certain devices.Based on all above,this paper proposes a bottom-up family load forecasting method.The model allows the users to predict the consumption of different types of devices and the family load in different time sections.The open accessed data of the REFIT project hold by Loughborough University in U.K and the experimental data provided by College of Engineering in ANGERS(ESEO)are used in this paper.The main work of this paper shows as followings:1)A decomposition method of different devices is proposed based on the analysis to these data.This paper analyzes the operation characteristics of different electrical equipment in different homes to divide the devices into two types: single-state and complex-states.For a single-state device,it can be regarded as an ON/OFF two-state system directly.For a complex state device,it is disassembled into a series of ON/OFF two-state subsystems according to its operation characteristics.The linear superposition of these subsystems gives a complete device system.Markov chain model is used to predict the time sequence of its operation state for each subsystems.Besides,a Mont Carlo simulation will be introduced into the forecasting process to obtain a stable result.The daily average power and hourly average power are chosen as the evaluation index.2)Based on the Mont Carlo Markov Chains(MCMC)model,a bottom-up load forecasting model for the family load will be proposed next.In this model,the parameters of MCMC will be get according to the similar day's consumption information.This model includes three lays: subsystems of the devices load,load of devices in different types,and family load.The average load of each layer will be simulated until the family load is obtained.3)At the end of this paper,a Hidden Markov Model(HMM)will be introduced to predict the operation sates of the fixed-frequency temperature-related equipment,such as fridge,freezer,and air-conditioner,etc.The simulation results of different models will be compared in this part,the adaptive occasion of each model will be illustrated as well.The error analysis process is introduced in this paper to illustrate the effect of different type of models and it shows that MCMC model is more suitable for overall load prediction of a certain devices with errors between 2%?8%.And HMM is more applicable to do state identification for temperature related devices with an accuracy of 70%.