Research And Application Of Smart Home Power Prediction System Based On Machine Learning

This article takes the power prediction of smart home equipment as an entry point,analyzes the current technology of machine learning in load prediction and power prediction in smart home,and takes the analysis of load changes combined with environmental data as the design requirement,with the goal of providing more accurate load forecasting data to the smart grid from the household electricity consumption,it is proposed to build a smart home power forecasting system.In order to realize the prediction of equipment operating power,this paper establishes a power prediction system based on LSTM network.First,select features from the historical operating power of the smart new wind machine and the environmental data during its operation,use the average absolute error as the loss function,and use the Adam optimizer to build a high-availability power prediction system.In order to solve the problem that the trained model is difficult to be applied to the actual life environment,this article made the following specific work to realize the application of the model:(1)The energy measurement terminal composed of the device data collection terminal and the device under test measures the power consumed by the device under test and its environment in real time through sensors and Python scripts.The energy metering device converts the real-time measured data into JSON format,and publishes the formatted data to the smart home cloud platform through the MQTT protocol.The cloud platform distributes the data according to the subscription status of the relevant client.And execute the control instructions issued by the smart home cloud platform,and guarantee the execution rate of the instructions.(2)Establish a smart home cloud platform and deploy it to an Alibaba Cloud server to provide external services.The cloud platform provides services such as device connection,data management,and API.(3)The user terminal is realized by We Chat Mini Program and webpages.The We Chat Mini Program supports users to view the measured data and predicted data of the device under test in real time,and supports the user to control the device under test in real time.The user can view the detailed historical data of the tested device on a webpage through a large-screen device(such as a PC).(4)In order to improve the applicable scope of the model,the prediction model that has been trained on the intelligent new wind turbine data set is transferred and learned on the device under test.The model will combine the previous learning knowledge to learn the characteristic data of the device under test and realize the prediction of the active power of the device under test.The power prediction model designed in this paper not only improves the accuracy of power prediction,but also through the combination of the power prediction system and the smart home system,users can fully understand the electricity consumption of home devices.Users can cultivate good electricity consumption habits through understanding of household energy consumption,so as to achieve the purpose of energy saving.In order to improve the user experience,the cloud platform designed in this paper also supports Ali Genie's voice control of the device data collection terminal.