Investigation Into Load Prediction And Wireless Sensor Network Based Monitoring In Smart Grids

With the development of mobile Internet technology, the intelligentized management has been the inevitable direction for industrial transformation and upgrading. The electric power is a basic industry of the national economy. Wireless sensor network monitoring and load prediction of the electric power can not only improve the quality of the public service for the whole society but also effectively contribute to the development of the national productivity. Therefore, the research focuses on the wireless sensor network monitoring and load prediction.(1) The dissertation discusses the value and roles of smart grid technologies taking place in the operation and maintenance, marketing, business operation withinthe electric power industry, andputs forward the architecture model of the data flow of the smart grid technologies from the angle of data flow (information application layer), and discusses the functional logics, technical methods, application key points and difficulties of various layers. The dissertation discusses the effect of WSN node energy consumption, network business functions, deployment application environment, and protocol strategy on the continuous stable operation of WSN. The dissertation analyses and discusses the environmental factors (date, season, and event) that influence electric charge data and their characteristics (regularity, randomness, predictability, hysteresis, and relevance) from the perspective of data prediction for smart power grids.(2) Based on the AODV routing protocol and IEEE 802.15.4 MAC protocol, the dissertation puts forward a topological structure and networking strategy applicable to sensor monitoring of the long-distance electric transmission lines (M2-TSLC, Modified MAC protocol & TSLC routing protocol), proposes further optimization of the WSN network using the cross-layer design method, and evaluate the quality of the optimized WSN network in combination with the four indexes of surplus energy consumption, throughput, end-to-end delay, and delivery rate. The method not only effectively saves the WSN energy consumption but also has good communication performance.(3) With the electrical load of the power grid being the research object, the dissertation involves testing of the autoregressive-moving-average (ARMA) model and grey prediction (GM) model, and puts forward the grey power load based on invasive weed optimization (IWO) specific to error accumulation and low accuracy present in the process for solving a and b with the least square method in the original grey prediction model. The error analysis dimension extends from absolute error and relative error to mean absolute error, average absolute percentage error, mean square error, and root-mean-square error for more comprehensive comparisons and analyses of the three prediction algorithms (ARMA, GM, and IWO-GM).(4) The dissertation also discusses the dynamic threshold computing method in combination with the unilateral threshold method and bilateral threshold method (historical analogy method and mean variance method), expounds the different memory patterns (immediate memory, short-term memory, and long-term memory) of the memory curve (The Ebbinghaus Forgetting Curve) with mathematical models, combines the physical deformation process of the spring with the memory curve idea, analyses and induces the action of the mechanics upon the memory spring and applies the result to the dynamic threshold analysis, and puts forward an elastic dynamic threshold power load warning method with memory characteristics (MC-EDT, Memory Curve-Elastic Dynamic Threshold). In consideration of the relatively fewer abnormal values of the electric power load data, the research uses the equally-weighted early warning rate as an approach for evaluation of the dynamic threshold. The authors test, verify, analyze, and discuss the related parameter regularities and the optimal configuration for the MC-EDT algorithm. The MC-EDT algorithm considerably improves the accuracy and precision of early warning for dynamic threshold compared with the mean variance method.In conclusion, with the data on the smart grid being the object, the research discusses the data for smart grid technologies extracting from information resources database as well as its information system architecture. Besides, the thesis puts forward and proposes the M2-TSLC energy conservation protocolmonitoring from smart grid power WSN data energy conservation, and electric power load IWO-GM prediction algorithm, electric power MC-EDT dynamic threshold onpower load warning strategy, and thus providing a new approach for scientific decision-making and information value transformation for the smart griddata.