基于支持向量机的抽水蓄能电站直接厂用电建模研究

发布时间：2018-02-24 13:33

本文关键词： 抽水蓄能电站厂用电量模型南方电网支持向量机模糊 K-means聚类粒子群算法　出处：《华南理工大学》2014年硕士论文　论文类型：学位论文

【摘要】：厂用电率作为衡量抽水蓄能电站经济运行和节能降耗的重要技术经济指标，直接影响抽水蓄能电站经营运行效益与节能减排成效。目前国内外针对抽水蓄能电站厂用电方面的深入研究比较少，且大部分都只集中在综合厂用电量/率的整体研究上。因此，对抽水蓄能电站的厂用电情况进行细化分析，建立科学的厂用电计算模型，具有重要意义。研究成果能为抽水蓄能电站的厂用电分析与管理考核工作提供科学依据，有利于促进电站厂用电的精细化管理、提高节能管理水平与经营运行效益。本文分析研究了抽水蓄能电站的厂用电构成与管理、考核指标及影响因素，提出了将直接厂用电率代替综合厂用电率作为厂用电考核指标，并阐述其合理性。本文在研究了支持向量机、K-means聚类及粒子群算法的基本理论与应用的基础上，提出了一种基于模糊K-means聚类的样本优化与粒子群的参数优化的支持向量机回归模型，以实现提高抽水蓄能电站厂用电模型精度，全年整体计算误差控制在5%以内的目标。本文所研究的抽水蓄能电站厂用电计算模型分为两部分。一是总体计算模型，基于往年历史数据，运用支持向量机法构建直接厂用电量与发电量、抽水电量、气温、水文等关键影响因素之间的回归模型，模型具有对总体直接厂用电进行拟合计算及预测的功能；二是模块化计算模型，将厂用电按系统按模块进行划分，，在对各设备用电量与发电量、抽水电量、气温、水文、运行规律、出力曲线等因素之间的关系进行定性分析，对设备电气参数、机组启停工况及运行时间等数据进行定量计算的基础上，计算各模块用电量，并同时运用支持向量机法构建模块化拟合模型，该模型具有对各模块直接厂用电进行计算、拟合及预测的功能。两部分模型相互结合与验证，可实现对南方电网调峰调频发电公司抽水蓄能电站本年度厂用电量进行验算分析与考核、下一年度厂用电量进行预测的功能。本文最后建立了惠州抽水蓄能电站直接厂用电计算模型并开发了配套应用软件，最终计算结果验证了模型的准确性和适用性，精度满足《南方电网“十二五”节能减排规划》中明确要求“调峰调频发电公司控制其所管辖电厂的厂用电率和标准值误差在5%以内”的节能减排工作目标。
[Abstract]:As an important technical and economic index to measure the economic operation, energy saving and consumption reduction of pumped storage power station, It has a direct impact on the efficiency of operation and energy saving and emission reduction of pumped storage power plants. At present, there are few in-depth researches on the power supply of pumped storage power plants at home and abroad. And most of them only focus on the overall study of the comprehensive power consumption / rate. Therefore, the detailed analysis of the situation of the pumped storage power station is carried out, and the scientific calculation model of the power consumption is established. The research results can provide a scientific basis for the analysis and management of the power consumption of pumped storage power stations, promote the fine management of the power plants, and improve the level of energy saving management and operation efficiency. In this paper, the composition and management of auxiliary power, the assessment index and the influencing factors of pumped storage power station are analyzed and studied, and it is put forward that the direct power consumption rate should replace the comprehensive service power consumption rate as the test index. This paper studies the basic theory and application of support vector machine (SVM) K-means clustering and particle swarm optimization (PSO). A support vector machine regression model based on fuzzy K-means clustering for sample optimization and particle swarm optimization is proposed to improve the precision of the model and control the overall calculation error within 5%. The model is divided into two parts. One is the overall calculation model. Based on the historical data of previous years, the support vector machine method is used to construct the direct power consumption and generation, pumping capacity, temperature. The regression model between the key influencing factors such as hydrology has the function of fitting and calculating and forecasting the total direct auxiliary power, the second is the modular calculation model, which divides the service power into modules according to the system. Based on the qualitative analysis of the relationship between each equipment's electricity consumption and electricity generation, pumping capacity, air temperature, hydrology, operation law, output curve, and so on, the electrical parameters of the equipment are analyzed. On the basis of quantitative calculation of unit starting and stopping condition and running time, the electricity consumption of each module is calculated. At the same time, the support vector machine method is used to construct the modular fitting model, which can calculate the direct power consumption of each module. The function of fitting and forecasting. By combining and verifying the two models, the function of checking, analyzing and checking the power consumption of pumped storage power station of Southern Power Grid peak-shaving and frequency modulation generation company in this year and forecasting power consumption in the next year can be realized. In the end, the calculation model of direct power supply for Huizhou Pumped-storage Power Station is established, and the corresponding application software is developed. The result of calculation verifies the accuracy and applicability of the model. The precision meets the requirements of the "12th Five-Year Plan for Energy Saving and Emission reduction of Southern Power Grid", which explicitly requires that "peak-shaving and frequency modulation power generation company control the rate of power consumption and the error of standard values of the power plant under its jurisdiction within 5%".
【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】：TV743

【参考文献】