基于不同维度建模的城市电网电量预测方法研究

发布时间：2018-04-27 20:05

本文选题：城市电网 + 电量　；参考：《华南理工大学》2014年硕士论文

【摘要】：城市电网电量（供电量或售电量）预测是电力市场中的一项基本工作。建立可靠的预测方法，做好城市电网电量预测工作可以科学指导发电机的出力和变压器的经济运行以及电气设备检修的合理安排，也可以为供电企业的营销和线损管理提供决策支持，对指导电气设备检修、电网经济运行和推动电力市场的发展都具有十分重要的意义。在电量预测过程中，对季节性电量数据的预测存在三个亟需解决的问题。一、电量季节性数据具有波动性和趋势性两重趋势的非线性特征，单一预测模型难以准确描述这种非线性变化过程；二、电量季节性数据，既有自身的变化规律，又受到内部和外部多因素影响而呈现一定的不确定性；在数学建模过程中，如果不能同时引入相关变量来进行建模，将导致预测模型不能正确反应电量数据的真实变化过程，使预测结果的精度和可信度降低。三、在电量预测过程中，最优拟合模型不一定就是最优预测模型；以拟合精度选择预测模型的模型选择机制，如果舍弃其它拟合精度没那么高的预测模型，可能会遗失某些预测信息，，从而得不到正确的预测结果。针对上述三种问题，本文先提出了两类建模方法，一类为基于电量数据变化规律的单维度预测方法；一类为基于行业用电及相关因素的多维度预测方法。基于电量数据变化规律的单维度预测方法包含了4个预测模型，这4个预测模型利用电量自身的数据建模，用不同的方法从不同的角度反映了电量自身的变化规律；基于行业用电及相关因素的多维度预测方法包含2个预测模型，这2个预测模型在数学建模过程中引入了行业用电和经济维度，从电量内部（行业用电）和外部（经济因素）体现电量的变化特征，弥补了基于电量变化规律的单维度预测方法没有从其它维度反映出其它相关数据对电量影响的不足。再提出基于不同维度建模的城市电量预测方法，该方法运用方差—协方差优选组合法对两类不同维度的预测模型的所有预测信息进行最大化利用，实现预测结果的最优组合，提高了预测结果的精确度和可信度，为电量预测提供一种新思路。为了使得本文提出的方法更易使用和推广，利用MATLAB的GUI软件包开发了一套基于上述方法的预测软件。并利用该软件对广东省某城市电网进行实例分析，实例计算结果表明优选组合预测结果中既包含了体现供电量自身变化规律的结果，又包含了体现行业用电及经济因素对供电量影响的结果，预测精度大幅优于各单一模型。这说明，这种方法预测性能优越，大大提高了预测精度；开发的软件具有很高的实用价值。
[Abstract]:It is a basic work in the electricity market to predict the quantity of electricity (electricity supply or electricity sale) in urban power network. The establishment of reliable forecasting method and the work of electricity quantity prediction in urban power grid can scientifically guide generator output, economic operation of transformers and reasonable arrangement of electrical equipment maintenance. It can also provide decision support for marketing and line loss management of power supply enterprises. It is of great significance to guide the maintenance of electrical equipment, economic operation of power grid and promote the development of power market. In the process of electric quantity prediction, there are three problems that need to be solved in the prediction of seasonal electricity quantity data. First, the seasonal data of electricity quantity have the nonlinear characteristics of volatility and trend, so it is difficult for a single forecasting model to accurately describe the nonlinear change process; second, the seasonal data of electricity quantity has its own changing law. In the process of mathematical modeling, if the relevant variables can not be introduced to model at the same time, the prediction model will not correctly reflect the real change process of electricity data. The accuracy and reliability of the prediction results are reduced. Thirdly, in the process of electric quantity prediction, the optimal fitting model is not necessarily the optimal prediction model, and if the model selection mechanism of the prediction model is selected with the fitting accuracy, if the other prediction models with less fitting accuracy are abandoned, Some prediction information may be lost and the correct prediction results will not be obtained. In order to solve the above three problems, two kinds of modeling methods are proposed in this paper, one is a single-dimensional prediction method based on the law of change of electricity quantity data, the other is a multi-dimensional prediction method based on industry electricity consumption and related factors. The single dimensional forecasting method based on the change law of electricity quantity data includes four prediction models, which use the data of electricity itself to model, and reflect the change law of electricity quantity from different angles by different methods. The multi-dimensional forecasting method based on industry electricity consumption and related factors includes two forecasting models, which introduce the industry power consumption and economic dimensions in the process of mathematical modeling. The internal (industry) and external (economic factors) of electricity quantity reflect the characteristics of electricity quantity change, which makes up for the deficiency of the single dimension prediction method based on the law of electricity quantity change, which does not reflect the influence of other related data on electricity quantity from other dimensions. Then a method of city electricity forecasting based on different dimension modeling is proposed. The method uses variance-covariance optimal combination method to maximize the utilization of all prediction information of two kinds of different dimension prediction models, and realizes the optimal combination of prediction results. The accuracy and reliability of the prediction results are improved, and a new way of thinking is provided for the electric quantity prediction. In order to make the proposed method easier to use and popularize, a prediction software based on the above method is developed by using the GUI software package of MATLAB. The software is used to analyze an urban power network in Guangdong Province. The result of example calculation shows that the forecasting result of optimal combination includes the result that reflects the law of the change of electricity supply itself. It also includes the results of reflecting the influence of industry electricity consumption and economic factors on power supply, and the prediction accuracy is much better than that of each single model. This shows that this method has superior prediction performance and greatly improves the prediction accuracy, and the software developed has a high practical value.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM727.2

【参考文献】