Search

Electricity sales affected by economic development, seasonal changes, holiday, temperature and other factors, is difficult to predict accurately. In this paper, based on X13 seasonal adjustment and factor regression, a forecasting method is proposed, which takes development trend, seasonal variation and random variation of electricity sales into consideration. Firstly, the historical electricity sales data is preprocessed to improve the data quality, and the X13 seasonal adjustment method is adopted to decompose the electricity sales data into three subsequences: trend, season and random items. Then combining the influential factors of each subsequence with the characteristics of the curves, each item is predicted and reconstructed to get the forecasting results respectively, where the trend item is predicted by several algorithms. Finally, analytic hierarchy process (AHP)-based comprehensive evaluation model is adopted to get the final forecasting results. The validation shows that the monthly average error of the whole industry is 1.11% from January to June 2016.

Grasping the effect of power supply net capacity on electricity sales has a great significance to analyze the trend of electricity sales accurately. Based on correlation analysis of historical data of the first, second, third industrial and residential electricity sales and power supply net capacity of power company, this paper presents a method of analyzing the effect of power supply net capacity on electricity sales by growth curve. The method eliminates the effect of seasonal factor and random factor of power supply net capacity and electricity sales by X13 seasonal adjustment, calculates the leading time of power supply net capacity ahead of electricity sales by dynamic time warping, calculates load operation ratio of power supply net capacity by growth curve. The method is used for quantitative analysis in studying electricity sales' changes caused by power supply net capacity. The experimental results show that the proposed method can be a good measure of the impact of power supply net capacity on electricity sales, the model can also be used to guide the analysis of the trend of electricity sales and provide effective data support for decision making of power companies.

In this paper, a short-term load forecasting method which considers the different load characteristics in different periods is proposed. Firstly, we use parallel K-Means algorithm to cluster the daily load curves with 96 points of electricity customer to obtain some date groups with different load characteristics. Then for each date group, we use the daily load curves in the group to build load forecasting model of every point by the random vector functional-link net. Finally, we find the similar historical day of the forecast day by dynamic time warping method and use the forecasting models of the date group which contains the similar date to forecast the customer's load in the forecast day. Empirical study shows that the method is suitable for the short-term load forecasting of massive customers and has satisfactory forecasting accuracy.

Due to the strong randomness and intermittency of photovoltaic (PV) power output, accurate PV power output forecast becomes more and more important for system reliability, meanwhile it can promote large-scale PV deployment. In this paper, a novel PV power output forecast model based upon weather type clustering and support vector machines (SVM) regression is proposed. Firstly, on the basis of calculated average historical PV power output of each weather type, expectation maximization (EM) algorithm is adopted to cluster weather types into some categories. Secondly, based on clustering results and weather information collected from authoritative meteorological administration, the input samples are selected to better reflect weather characteristics of the forecasting day. Finally, for certain weather type, a wavelet SVM regression approach is adopted to forecast PV power output. Extensive experimental results demonstrate that the proposed model for PV power output forecasting has a high forecasting accuracy.

Analyzing characteristics of customers' electricity consumption behavior is helpful to improve the management level of demand side energy efficiency. In this paper, a method for analyzing the customers' electricity consumption behavior based on the massive data is proposed. Firstly, clustering algorithm is used to cluster daily load curves of the main power grid in the past one year, then under the different date clusters, the electrical load characteristics of residents and non-residents are analyzed. Empirical study shows that the method provides effective support for intelligent electricity guidance of residents and orderly use of electricity of non-residents, which is conducive to peak load shifting and stable operation of the main power grid.

A scientific, objective and comprehensive evaluation of the works of the customer service staffs is essential for improving the customer service ability and qualities, and especially for improving the satisfactory rate. In this paper, we propose a scientific and objective method for the comprehensive evaluation of the staffs of State Grid Customer Service Center which consists of two steps. In the first step, the L1/2 sparse classification model is used to distinguish the indexes which will be used for the evaluation from all indexes. In the second step, the Analytic Hierarchy Process is used for the comprehensive evaluation of the staffs over the selected indexes. Extensive experimental study shows the priority of our proposed method.

The forecasting of the monthly electricity sales is a fundamental work of the marketing department of State Grid Corporation of China, which has been implemented using regression and time-series analysis based on historical electricity sales. In this paper, we first study the correlation between electricity sales of all trades and related influencing factors, e.g., weather, economy, holidays and events, using the Pearson correlation coefficient, and further divide all trades into several super trades using the EM clustering algorithm based on the correlation. Then, a electricity sales adjustment model is created for each super trade using the SVM regression algorithm, which can determine the adjustment quantity of the electricity sales based on anomalies of the influencing factors. Extensive experimental study shows that the proposed adjustment approach can greatly improve the accuracy with respect to forecasting the next 12 months electricity sales of the power companies of State Grid Corporation of China (SGCC).

Effective and efficient description of the features of different staffs is the key problem to achieve scientific and meticulous management in customer service centers, because the features are essential for decision-making works, e.g., the formulation of the scheduling strategy, the placement strategy and the performance evaluation standard. In this paper, by introducing the visual clustering algorithm, the electric customer service staffs are divided into four types, i.e., the quality type, the efficiency type, the medium type and the problem type, on the basis of the workload, work quality and work efficiency of the staffs. The clustering results are helpful for the formulation of the scheduling strategy, the placement strategy and the performance evaluation standard objectively. The visual clustering algorithm is now regularly applied to divide electric customer service staffs into different groups so that related strategies and standards may be tuned timely and properly. The successful application of our proposed method in State Grid Customer Service Center shows the effectiveness and the efficiency of the proposed visual clustering algorithm.