| The energy industry is our country’s pillar industry and drives the country’s prosperity,social development,and people’s happiness.The power industry occupies a leading position,and the deepening reform of the power system has a pivotal impact on the development of the country and society.In 2019,the State Grid Corporation of China proposed a "three-type,two-network" world-class energy Internet enterprise construction requirement in response to the "two sessions" work deployment,which guides a new direction for the transformation and upgrading of the power system.This article mainly focuses on the key issues of the ubiquitous power Internet of Things in the "two networks",including the difficulty of data aggregation at the bottom of the IoT,the inconsistent modeling of energy systems,and the information barriers of the power entities.We propose targeted solutions at the physical layer,platform layer,and application layer,while organically fusing traditional structured data processing technologies and neural network algorithms to form a hybrid model supplemented by experimental verification,and proposes a data fusion technology that comprehensively considers data collection,data preprocessing,and hybrid modeling.At last,we can combine the characteristics of the power industry and specific business needs,effectively extract the value information from the massive power Internet of Things data,and realize downsizing and increasing efficiency and value sharing.Aiming at the problem of effective fusion of heterogeneous data in the ubiquitous power Internet of Things environment,this paper integrates traditional sequential data processing technology and neural network technology,and presents the effect in the form of energy consumption prediction.First,we construct a data acquisition scheme based on software-defined intelligent terminal technology to obtain raw data,use K-means to construct exogenous variables,and then perform stationarity verification and smoothing processing of experimental samples,and construct the underlying S ARIMAX based on periodic trends and do fitting.Next we import the fitted data set into the LSTM model for secondary fitting,construct a SARIMAX-LSTM two-layer model.Finally we verify the effect of the mixed model through the original energy consumption data,weather data and holiday data.Experimental analysis shows that the SARIMAX-LSTM hybrid model can integrate weather factors,holiday factors,and seasonal factors to make high-precision predictions of energy consumption,and realize the effective integration and value mining of heterogeneous data. |
