1 Trend Analysis
Trend Analysis Trend Curve Analysis, Curve Fit or Curve Regression, which is the most research, and the most popular quantitative prediction method. It is a curve based on known historical data, so that this curve reflects the growth trend of the load itself, and then estimates the load predictive value of the time according to this growth trend curve. Common trend models have linear trend model, polynomial trend model, linear trend model, logarithmological model, power function trend model, index trend model, logistic model, Gonpertz model, etc., seek trends The process of the model is relatively simple. This method itself is a determined extrapolation, in processing historical data, fitting curve, and obtains the random error. The curve of trend analysis is used, and its accuracy is in principle consistent with the fitting range. In many cases, choose the appropriate trend curve, it is indeed a better prediction result. However, the results given by different models will be large, the key to use is based on regional development, and choose the appropriate model. Analyze Zhuhai City's electricity-electricity history data since 1995, found that there is a more obvious binary growth trend, the model curve is Y = 0.229565x2-914.8523x 911472.65, using the model curve to get electricity from 2005 to 2010 The amount level is 5.278 billion kWh and 8.508 billion kWh, respectively. The fit curve is shown in Figure 1.
2 regression analysis
Regression Analysis (Also known as statistical analysis) is also a quantitative prediction method currently widely used. Its task is to determine the relationship between predictive values and influencing factors. The power load regression analysis is based on statistical analysis of the influence factor value (such as the total national product gross domestic product, the total output value, the population, the climate, etc.) and the electricity, and the function relationship between electricity consumption and influencing factors. Thus, it is predicted. However, due to the regression analysis, what expression of which is used for the factor and the factor system is sometimes speculative, and it affects the diversity of electronics and certain factors, so that the regression analysis in some cases restricted.
Analysis of the electricity consumption of Zhuhai calendar and domestic product GDP, population POPU and other data, seeking return equation: y = -3.9848 0.0727GDP 0.10307Popu, using this model to predict 2005 and 2010 electricity The amount level is 4,711 billion kWh and 7.098 billion kWh, respectively.
The regression analysis prediction method is to analyze the analysis of historical data, explore economic and social factors and power load in contact and development changes, and calculate the future according to the prediction of the economic and social development in the planning period. Load. It can be seen that this method does not only depends on the accuracy of the model, but also the accuracy of the influence factor itself.
3 index smooth
Trend analysis and regression analysis are based on the actual value of the time series, and the model is used to perform prediction calculations. The exponential smoothing method is to directly predict the future value of the time series with an index weighted combination of previous historical data.
Figure 1 Fit graph
Among them, the attenuation factor 0 <α <1, reflects the "Touring Tour", that is, the basic principles affecting the prediction of prediction and long-term data. When α is, the faster the weighting coefficient of the near future data is larger, and it is emphasized that the role of new data. For example, when α = 0.9, each weighting coefficient is 0.9, 0.09, 0.009, respectively. In extreme cases, α = 1, the past data has no effect on the forecast.
For power system load forecasting, it is important that the closer of the curve should be more accurate, and the data for more than a long time is unnecessarily fit. Similar inertial effects.
It can be seen from the example of Zhuhai City, and the prediction effect is better. Example calculations show that the method can better simulate the actual and prediction of Zhuhai. However, it is not suitable for predictions in too long.
4 single-consuming method
The single consumption is based on the economic value created by electricity per unit of electricity per unit, from the forecasting economic indicators, plus residents' live power, and constitutes power. When predicted, by statistically analyzed the power consumption of the past unit output value, combined with the industrial structure adjustment, it is found that the integrated unit of the second and third industries in the planning period, then according to the national economy and social development. The planned indicators are predicted by single consumption. The single consumption requires a lot of meticulous statistics, analysis work, and the recent prediction effect is better. However, under market economy conditions, future industrial unit consumption and economic development indicators have uncertainty, and it is difficult to determine the accuracy of medium-term forecasting.
5 gray model method
The gray system theory is the viewpoints and methods of anti-blurring to extend into complex large systems, combining automatic control and operational training, research is widely existing in the objective world with gray sexual problems. Some information is known and unknown systems called gray systems.
The first-order gray model was used to predict the electricity consumption of Zhuhai City. In 2005, the whole society's electricity is predicted that its results should be satisfactory. Through the different processing methods of the original data, it is predicted that the 2005 full community is about 5 billion kWh, which is quite close to the results of other common methods forecast. In this 6 programs, the programs 3 were inspected in cases, and the rest were excellent. However, the results obtained using long data columns are not excellent, and the data is too long, the system is interfering, and the unstable factors are large, but the accuracy of the model is reduced, and the credibility of the forecast results is reduced.
6 load density method
The load density is generally represented in KW / KM2. Different regions, different functional areas, load density is different. Using load density method, it is generally divided into several functional districts, such as business districts, industrial areas, residential areas, cultural and educational districts, and then according to regional economic development planning, population planning, residential income levels, etc., refer to the region Or use the level of electricity at home and abroad, select a suitable load density indicator, the application of the functional zone and the electricity load of the entire prediction zone. The calculation formula is A = SD, where S is the land area, and D is the electrical density. This method is mainly suitable for urban areas where land plan is more clear, and we use this method when we predict the predictive load in Zhuhai City distribution network.
7 Elastic Coefficient Method
The power elastic coefficient is a macro indicator that reflects the relationship between the annual average growth rate of power consumption and the annual average growth rate of the national economy. The power elastic coefficient can be represented by the following formula:
E = KY / KX
E- is the power elasticity coefficient
KY - average growth rate for electric consumer year
KX- is the average annual growth rate of the national economy
Under the market economy, the power elasticity has become ugly, and with the rapid development of science and technology, the management of power and electricity demand, the relationship between electricity and economy has changed dramatically, and the changes in electricity demand and economic development are serious. Disorder, making the elastic coefficients difficult to touched, using an elastic coefficient method to predict the effect of power demand to obtain satisfactory, should gradually fade.
8 analysis and comparison
(1) From the application conditions, regression analysis and trend analysis is committed to the research and description of statistical laws, suitable for large samples, and in the past, now and future development models; index smoothization is the use of inertial principles to growth trends Push, realize the prediction principle of "close to the remote"; the output value unit consumption method generally determines the single-consuming indicator based on the changes in the analysis of the factors affecting the output value, and then according to the national economic and social development plan The indicator predicts the power demand; the gray model method is to seek regularity through the sorting of raw data, which is suitable for analysis and prediction under the conditions of poverty.
(2) From the form of data used, the gray system theory is to model the generation of several sequences. The regression analysis method, the trend analysis method is the model of raw data. The index smoothization is to directly predict future values by exponential weighted combinations of raw data.
(3) From the calculation complexity, relatively simple is the regression analysis method and trend analysis method.
