Short-term load forecasting using neural networks

An artificial neural network (ANN) model for short-term load forecasting (STLF) is presented. The proposed model is capable of forecasting the next 24-hour load profile at one time, as opposed to the usual ‘next one hour’ ANN models. The inputs to the ANN are load profiles of the two previous days and daily maximum and minimum temperature forecasts. The network is trained to learn the next day's load profile. Testing of the model with one year of data from the Greek interconnected power system resulted in a 2.66% average absolute forecast error.     

基于自组织映射网络的峰值负荷预测方法

应用扩展自组织映射网络研究了电力系统峰值负荷预测问题。在传统的Kohonen自组织映射(SOM)网络的学习算法的基础上,为了提高电力系统峰值负荷预测的精度,进一步提出了一种扩展的自组织映射算法。在这个SOM网络中,除了权矩阵外,还有一个输入输出对的局部梯度(Jocobian)矩阵也被存储在神经元中。这样,在输出空间中梯度信息围绕输出权值产生了一个一阶扩展,便可得到一个输出的改进估计值。同时,提出了一个Jocobian矩阵的生成算法。最后采用纽约市的电力负荷数据为研究对象,证明了所提出方法的有效性。     

Stock price prediction using neural networks with RasID-GA

In general, neural networks are widely used in pattern recognition, system modeling and prediction, and can model complex nonlinear systems. In the previous work, we proposed a novel training algorithm, Adaptive Random Search with Intensification and Diversification combined with Genetic Algorithm (RasID-GA), for training the multibranch recurrent neural networks recently developed. In this paper, RasID-GA has been applied to predict stock market prices using the multibranch feed forward neural networks. We predicted the next day's closing stock price with several past closing stock prices. We used the stock prices of 20 brands for 720 days in order to evaluate the generalization ability of the proposed method. Copyright © 2009 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.     

Novel approach of modeling load duration curve for generation expansion planning based on Hill's function

A novel approach of modeling the load duration curve (LDC) based on Hill's function is proposed in this article. On the contrary to traditional models, the proposed model is completely an analytical one which can be determined by historic load data. This method is effective in calculating efficiency as well as controlling errors and it is quite simple in application because the model has only a few parameters, each of which has a definite economic or fiscal meaning. Based on the historic model, this method is easy and accurate in estimating the LDC model for a future year by changing the parameters of Hill's function, where only the peak load and the total demand in each year may be given. Results on the load data from IEEE reliability test system (IEEE‐RTS), PJM and Beijing Electric Power Corporation (BEPC) are presented to demonstrate the effectiveness of the proposed model. Numerical examples show that the modeling errors in both peak load and total demand, which are key indices for generation expansion planning and reliability evaluation, are less than 1%. The LDC model for a future year is also accurately estimated in these examples. © 2011 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.     

基于先验知识与DBM采样的类不平衡用电数据分类方法

智能电网建设过程中现有客户标签体系不够完善,针对海量用户用电数据的分类管理中带有标签的样本数据量小以及类不平衡分布的问题,提出了一种基于先验知识与深度玻尔兹曼机(DBM)采样的不平衡用电数据分类方法。首先,提取负荷曲线的特征,建立采样原则,利用先验知识和DBM对负荷曲线进行采样。然后,将采样数据通过极限学习机(ELM)网络进行训练。最后以爱尔兰用户用电数据为数据源,通过与原始非采样、随机过采样、合成少数类过采样技术(SMOTE)的对比性实验分析结果表明,所提出的基于先验知识与DBM采样的不平衡用电数据分类方法能够更好地对类不平衡用电数据集进行分类,实现用户用电行为的分析,有效支撑用户侧错峰避峰工作。     

Risk assessment and bi‐level optimization dispatch of virtual power plants considering renewable energy uncertainty

Aiming at the effective management of distributed energy resources (DERs ), we propose a bi‐level optimization dispatch model which consists of a virtual power plant (VPP ) and an independent system operator (ISO ) in order to achieve a coordinated dispatch between the upper level objective function of minimizing system's operation cost and the lower level objective function of maximizing the VPP 's net profit. However, the VPP 's dispatch strategy is variable due to the threat from unavoidable uncertainties including unit outage, load forecast deviation, and renewable energy forecasting error. In order to evaluate the influences of these risks on VPP 's dispatch strategy, a proper definition of expected energy not supplied (EENS ) and its relationship with VPP 's reserve capacity are provided. Also, chance‐constrained programming (CCP ) is adopted during the analysis. By including VPP 's risk cost into the dispatch model, VPP 's dispatch strategy is able to keep a balance between the potential risk and economic profit. In addition, a further discussion regarding the risk cost is conducted in view of renewable energy capacity proportion and the VPP capacity proportion. Based on the analysis results of the case study provided, the proposed model serves as a foundation for VPP 's economic and risk‐balanced dispatch.     

Peak load forecasting using multiple-year data with trend data processing techniques

This paper presents a regression-based daily peak load forecasting method using multiple-year data with trend cancellation and trend estimation techniques. Daily peak load heavily depends on daytime temperature and is influenced by the other weather factors such as humidity. Since the characteristic of the load is varying, peak loads just before a forecasting day are more significant for the forecasting. The regression model can represent relationships between these weather factors and peak loads. However, the forecasting model is sometimes not adequate for precise load forecasting. The regression model is well matched with the late data, but the model causes large forecasting errors in transitional seasons because of seasonal change of load characteristics. In order to forecast precisely through a year, a method of using seasonal or whole year data from past years is proposed. In this paper, two kinds of trend data processing techniques are described. The first is trend cancellation. The second is trend estimation. The trend cancellation technique removes annual load growth by means of division or subtraction processes with morning load on the forecasting day. The trend estimation technique estimates the trend between the forecasting year's load and the past year's load by using the variable transformation techniques. The performance of both techniques, verified with simulations on actual load data, is also described. © 1998 Scripta Technica, Electr Eng Jpn, 124(1): 7–16, 1998     

A novel quasi‐Z‐source indirect matrix converter

A new type of three‐phase quasi‐Z‐source indirect matrix converter (QZS‐IMC) is proposed in this paper. It uses a unique impedance network for achieving voltage‐boost capability and making the input current in continuous conduction mode (CCM) to eliminate the input filter. The complete modulation strategy is proposed to operate the QZS‐IMC. Meanwhile, a closed‐loop DC‐link peak voltage control strategy is proposed, and the DC‐link peak voltage is estimated by measuring both the input and capacitor voltages. With this proposed technique, a high‐performance output voltage control can be achieved with an excellent transient performance even if there are input voltage and load current variations. The controller is designed by using the small‐signal model. Vector control scheme of the induction motor is combined with the QZS‐IMC to achieve the motor drive. A QZS‐IMC prototype is built in laboratory, and experimental results verify the operating principle and theoretical analysis of the proposed converter. The simulation tests of QZS‐IMC based inductor motor drive are carried out to validate the proposed converter's application in motor drive. Copyright © 2013 John Wiley & Sons, Ltd.     

实时电价下含V2G功能的电动汽车理性充放电模型及其分析

为了利用实时电价实现电动汽车理性充电,以电动汽车运营收益最大化为目标,以满足电动汽车动力电池充放电容量及电动汽车行程需求为约束条件,构造了一个电动汽车充放电收益最大化模型,该模型较好地表示电动汽车充放电决策。以美国家庭出行调查为依据,根据用户充出行规律,采用蒙特卡洛模拟法模拟用户行程需求,对电动汽车充放电运行的经济效益进行仿真计算和分析。研究结果表明,通过响应电网实时电价,理性充放电模型可显著提高电动汽车的经济效益。同时,由于夜间电价相对便宜而白天相对较高的电价激励,电动汽车多在配电系统负载率较低时充电,在系统峰荷附近反向放电,从而起到削峰填谷的效应。     

Risk quantification for ANN based short‐term load forecasting

A new risk assessment method for short‐term load forecasting is proposed. The proposed method makes use of an artificial neural network (ANN) to forecast one‐step‐ahead daily maximum loads and evaluate uncertainty of load forecasting. With ANN as the model, the radial basis function (RBF) network is employed to forecast loads due to its good performance. Sufficient realistic pseudo‐scenarios are required to carry out quantitative risk analysis. The multivariate normal distribution with the correlation between input variables is used to give more realistic results to ANN. In addition, the method of moment matching is used to improve the accuracy of the multivariate normal distribution. The peak over threshold (POT) approach is used to evaluate risk that exceeds the upper bounds of generation capacity. The proposed method is successfully applied to real data of daily maximum load forecasting. © 2008 Wiley Periodicals, Inc. Electr Eng Jpn, 166(2): 54– 62, 2009; Published online in Wiley InterScience ( www.interscience.wiley.com ). DOI 10.1002/eej.20464     

The load characteristics classification and synthesis of substations in large area power grid

It is very difficult to build load model for every substation since there are numerous substations in large area power grid. A practical method is to classify the substations into several classes, pick out a typical substation from each class and build its load model, then generalize it to other substations of the same class. In this paper, a new method based on SOM (Self-Organization Map) neural network is presented for load characteristics classification and synthesis of substations in large area power grid. SOM neural network is a clustering method with self-organizing characteristics and mapping capability that can classify different input patterns automatically. Besides, the trained SOM neural network can discriminate the new input pattern conveniently without retraining. Therefore, the new substations can be discriminated with the existing classification result unchanged. The effectiveness of the proposed method is verified by a simulation of 183 220 kV substations in Shandong power grid using MATLAB Neural Network Toolbox. At first, the load composition rate in each substation is chosen as the feature vector, then SOM neural network is introduced to the classification and synthesis of the load characteristics of substations. At last, the synthetic load characteristic of each class is derived from the cluster center. The result is satisfactory since the method not only decreases the randomness and subjectivity of the load characteristic classification and synthesis of substations, but also improves the effectiveness and efficiency of load modeling. The method offers a new way for practical load modeling.     

基于自组织映射神经网络的电力用户负荷曲线聚类

电力用户负荷曲线的聚类是形成合理电价体系和实施负荷管理措施的基础。文中基于自组织映射（SOM）神经网络进行低压终端用户的负荷曲线聚类研究。首先定义并提取功率曲线、分时功率、功率频谱3类向量,分别作为SOM神经网络的输入进行可视化聚类。采用相对量化误差和拓扑误差2个指标表征聚类质量,选取聚类结果最好的SOM输出层结合 k均值法进行用户负荷曲线划分。根据Davies指标将所研究的131条曲线划分为8类,对每类曲线进行描述。最后进行新用户的识别,结果表明聚类方法有效、可靠。     

Cascading failures of power grids caused by line breakdown

In recent years, several large blackouts have drawn much attention to security problems in electric power transmission systems all over the world, which were triggered by initial minor disturbances and caused by cascading failures. Many models for cascading failures in power grids based on node attacks have been proposed, where only one kind of load is involved. In real power systems, however, strict measures are normally taken to protect the nodes, i.e. the power stations, so that failures at the nodes can hardly occur. It is much easier for transmission lines to break down, which may lead to cascading failures in entire power systems. Hence, a new model involving active and reactive loads and considering transmission line failures instead of node failures is proposed to understand cascading failures of power grids. When a transmission line breaks down, its load is redistributed to its neighbouring lines according to their respective capacities. It is well known that a power grid's topology and connectivity play a key role in determining its dynamic behaviour. Therefore, cascading failures of power grids with typical topologies, i.e. random, small‐world and scale‐free networks, are investigated using the model proposed and considering transmission line breakdowns. Finally, the model's effectiveness is validated employing real data of the European power grid. Copyright © 2014 John Wiley & Sons, Ltd.     

基于深度学习分位数回归模型的充电桩负荷预测

电动汽车的迅速发展将使充电桩负荷对电网造成影响,为此提出了使用深度学习分位数回归的充电桩负荷预测方法。该方法首先根据历史数据采用Adam随机梯度下降法训练出不同分位数条件下的LSTM神经网络参数估计,然后预测未来96 h内各分位数条件下的结果,再用核密度估计做出同一时刻结果的概率密度函数,最终得到负荷概率密度预测。根据实际充电桩负荷结果表明,提出的概率密度预测方法能较为精准地覆盖真实值,相比于BP神经网络分位数回归有着更高的精确度和参考价值。     

考虑快慢充负荷特性的电动汽车调峰定价策略

随着分布式光伏大规模发电的广泛应用,净负荷“鸭型”曲线特征明显,电动汽车白天充电无法充分利用新能源,夜间充电使原有负荷峰值叠加。为避免净负荷“峰上加峰”现象,文中以减小净负荷峰谷差为目标,实现充电负荷转移。首先,基于快慢充行为特征的统计数据,采用蒙特卡洛法模拟用户充电行为,实现未来充电负荷分布的预测,并根据慢充的入网特性以及快充的延迟充电特性建立快慢充负荷约束。然后,基于梯度下降法对负荷转移率进行计算,并引入用户消费心理学构建充电负荷价格响应模型。最后,对电网调峰的经济性进行分析以限定电价变动约束,以净负荷峰谷差最小为目标构建充电引导模型,并利用深度强化学习对其求解。仿真结果表明,所建模型和求解策略能有效引导充电负荷避开净负荷的峰期,并确定合理电价,减小电网的峰谷差。     

Economical load dispatching method considering power plant dynamics

To reduce boiler fuel cost, a new ELD (economic load dispatching) method based on a dynamic fuel cost model, which is more accurate than the conventional quadratic model, is proposed. First, an ARMA (autoregressive moving average) model, which is constant‐coefficient linear digital filter, is applied in order to supplement the quadratic model. We call this the supplemented model ARMA‐model‐supplemented quadratic model. By using this model, the model deviations from actual data have been reduced. Second, based on the ARMA‐model‐supplemented quadratic model, we formulated the ELD problem as a quadratic programming problem, where the objective function is the summation of all units' fuel costs over multiple time points and the constraints are the supply–demand balances, the upper and lower generation limits, and the ramp rate limits. This problem can be solved by the standard quadratic programming technique. We call this new ELD method BEST (boiler‐dynamics‐based economical load dispatching) method. Then, in order to make the problem size smaller, we propose a scheme to ignore all time points except for those corresponding to the peak, the bottom, and the steepest point on the forecasted load curve. We call this scheme the sample scheme. Finally, the BEST method with the sample scheme is evaluated by numerical simulations on the Kansai Electric Power Co. system and it is shown that the proposed method can reduce the calculation time without compromising the fuel cost. © 1999 Scripta Technica, Electr Eng Jpn, 127(1): 39–46, 1999     

一种光伏系统短期功率预测模型

为提高光伏系统发电功率预测精度,优化系统的发电计划,减少电力系统运行成本,进而为系统调度和实时运行控制提供依据以有效减轻光伏发电系统接入对电网的影响,建立一种基于三层神经网络和功率波动特性的短期光伏出力预测模型。利用气象局已发布的日类型和温度信息挑选与预测日最相关的相似日,基于神经网络用相似日历史太阳辐照、温度、输出功率建立光伏系统出力初步预测模型;以预测日天气预报信息作为神经网络的输入获得预测日的功率预测值;基于由光伏系统相似日历史出力数据统计分析得到的波动量统计规律对初步预测结果加以修正,建立了具有较高精度的光伏系统出力预测模型。仿真结果表明该方法建立的预测模型具有较高精度,能够为调度运行人员提供决策辅助。     

荷源联合调峰运行方案的电力节能评估研究

针对当前大规模风电远距离外送造成输电网网损激增和调峰能力不足的矛盾,提出将高载能负荷参与常规电源调峰,在风电就地消纳的同时实现电力节能。在深入分析高载能负荷的调节特性和对电网调峰影响的基础上,从电网、电源和高载能负荷三个方面提出了表征电力节能的指标并进行量化。最后基于综合模糊评价法对荷源联合调峰运行方案进行电力节能综合评估。仿真结果表明,该指标体系及方法在荷源联合调峰运行方案电力节能评估方面是有效可行的。     

计及风电—负荷耦合关系的含大规模风电系统调峰运行优化

随机性风电出力的准确建模是含大规模风电系统调峰运行优化的关键环节之一。针对中国系统运行的实际特点,提出计及风电出力与日负荷时序耦合关系的交互特性评价指标,并据此建立基于调峰问题驱动的风电出力模型。该模型以关键场景、聚类场景及其概率分布表征风电接入对系统运行和调峰需求的综合影响。基于该风电出力模型,建立了含大规模风电系统的调峰运行优化模型。通过多场景下的运行优化,获得各项运行指标的期望值,实现对系统调峰运行特性的综合分析。对中国西部某省进行了算例分析,验证了所提方法的有效性和实用性。     

融合历史数据和实时影响因素的精细化负荷预测

随着智能电网技术的飞速发展,对负荷预测的精度提出了越来越高的要求。融合负荷、天气等多源数据,提出了一种基于数据融合的支持向量机精细化负荷预测方法。首先对负荷历史数据进行聚类分析,将运行日分成六类。然后将负荷数据和温度、湿度等天气数据进行融合,针对六类聚类结果分别建立基于数据融合的支持向量机精细化负荷预测模型,并对模型参数进行全局优化。采用不同的预测模型对浙江省某地级市2013年的负荷进行预测,结果表明所提出的负荷预测方法的预测精度明显高于传统的负荷预测方法的预测精度。