Design of input vector for day-ahead price forecasting of electricity markets

In new deregulated electricity market, price forecasts have become a fundamental input to an energy company’s decision making and strategy development process. However, the exclusive characteristics of electricity price such as non-stationarity, non-linearity and time-varying volatile structure present a number of challenges for this task. In spite of all performed research on this area in the recent years, there is still essential need for more accurate and robust price forecast methods. Besides, there is a lack of efficient feature selection technique for designing the input vector of electricity price forecast. In this paper, a new two-stage feature selection algorithm composed of modified relief and mutual information (MI) techniques is proposed for this purpose. Moreover, cascaded neural network (CNN) is presented as forecast engine for electricity price prediction. The CNN is composed of cascaded forecasters where each forecaster is a neural network (NN). The proposed feature selection algorithm selects the best set of candidate inputs which is used by the CNN. The proposed method is examined on PJM, Spanish and Ontario electricity markets and compared with some of the most recent price forecast techniques.     

A new wind power prediction method based on ridgelet transforms,hybrid feature selection and closed-loop forecasting

To reduce network integration and boost energy trading, wind power forecasting can play an important role in power systems. Furthermore, the uncertain and nonconvex behavior of wind signals make its prediction complex. For this purpose, accurate prediction tools are needed. In this paper, a ridgelet transform is applied to a wind signal to decompose it into sub-signals. The output of ridgelet transform is considered as input of new feature selection to identify the best candidates to be used as the forecast engine input. Finally, a new hybrid closed loop forecast engine is proposed based on a neural network and an intelligent algorithm to predict the wind signal. The effectiveness of the proposed forecast model is extensively evaluated on a real-world electricity market through a comparison with well-known forecasting methods. The obtained numerical results demonstrate the validity of proposed method.     

基于卷积神经网络的鲁棒性基音检测方法

在语音信号中, 基音是一个重要参数, 且有重要用途. 然而, 检测噪声环境中语音的基音却是一项难度较大的工作. 由于卷积神经网络(Convolutional neural network, CNN)具有平移不变性, 能够很好地刻画语谱图中的谐波结构, 因此我们提出使用CNN来完成这项工作. 具体地, 我们使用CNN来选取候选基音, 再用动态规划方法(Dynamic programming, DP)进行基音追踪, 生成连续的基音轮廓. 实验表明, 与其他方法相比, 本文的方法具有明显的性能优势, 并且 对新的说话人和噪声有很好的泛化性能, 具有更好的鲁棒性.     

Solar energy forecasting based on hybrid neural network and improved metaheuristic algorithm

Prediction of solar power involves the knowledge of the sun , atmosphere and other parameters, and the scattering processes and the specifications of a solar energy plant that employs the sun's energy to generate solar power . This prediction result is essential for an efficient use of the solar power plant, the management of the electricity grid, and solar energy trading. However, because of nonlinear and nonstationary behavior of solar power time series, an efficient forecasting model is needed to predict it. Accordingly, in this paper, we propose a new forecast approach based on combination of a neural network with a metaheuristic algorithm as the hybrid forecasting engine. The metaheuristic algorithm optimizes the free parameters of the neural network. This approach also includes a 2‐stage feature selection filter based on the information‐theoretic criteria of mutual information and interaction gain, which filters out the ineffective input features. To demonstrate the effectiveness of the proposed forecast approach, it is implemented on a real‐world engineering test case. Obtained results illustrate the superiority of the proposed approach in comparison with other prediction methods.     

一种基于小波变换和ARIMA的短期电价混合预测模型

为在实时电价情况下预测未来24小时电价, 提出一种基于小波变换和差分自回归移动平均(ARIMA)的短期电价混合预测模型。该模型分别根据是否受到需求量影响使用ARIMA模型对多尺度小波变换分解后的时间序列进行预测。同时提出一种电价突变点发现和处理算法。使用澳大利亚新南威尔士州2012年真实数据验证表明, 相对ARIMA预测, 改进后的混合模型在不考虑需求量影响时预测精度更高; 电价突变点发现和处理算法能够准确处理电价异常点, 提高预测精度。     

基于极限学习机的股票价格预测

极限学习机（ Extreme Learning Machine , ELM）是一种新型的单馈层神经网络算法,克服了传统的误差反向传播方法需要多次迭代,算法的计算量和搜索空间大的缺点,只需要设置合适的隐含层节点个数,为输入权和隐含层偏差进行随机赋值,一次完成无需迭代。研究表明股票市场是一个非常复杂的非线性系统,需要用到人工智能理论、统计学理论和经济学理论。本文将极限学习机方法引入股票价格预测中,通过对比支持向量机（ Support Vector Machine , SVM）和误差反传神经网络（ Back Propagation Neural Network , BP神经网络）,分析极限学习机在股票价格预测中的可行性和优势。结果表明极限学习机预测精度高,并且在参数选择及训练速度上具有较明显的优势。     

Application of a new hybrid neuro-evolutionary system for day-ahead price forecasting of electricity markets

In this paper, a new forecast strategy is proposed for day-ahead prediction of electricity prices, which are so valuable for both producers and consumers in the new competitive electric power markets. However, electricity price has a nonlinear, volatile and time dependent behavior owning many outliers. Our forecast strategy is composed of a preprocessor and a Hybrid Neuro-Evolutionary System (HNES). Preprocessor selects the input features of the HNES according to MRMR (Maximum Relevance Minimum Redundancy) principal. The HNES is composed of three Neural Networks (NN) and Evolutionary Algorithms (EA) in a cascaded structure with a new data flow among its building blocks. The effectiveness of the whole proposed method is demonstrated by means of real data of the PJM and Spanish electricity markets. Also, the proposed price forecast strategy is compared with some of the most recent techniques in the area.     

A soft computing system for day-ahead electricity price forecasting

Hourly energy prices in a competitive electricity market are volatile. Forecast of energy price is key information to help producers and purchasers involved in electricity market to prepare their corresponding bidding strategies so as to maximize their profits. It is difficult to forecast all the hourly prices with only one model for different behaviors of different hourly prices. Neither will it get excellent results with 24 different models to forecast the 24 hourly prices respectively, for there are always not sufficient data to train the models, especially the peak price in summer. This paper proposes a novel technique to forecast day-ahead electricity prices based on Self-Organizing Map neural network (SOM) and Support Vector Machine (SVM) models. SOM is used to cluster the data automatically according to their similarity to resolve the problem of insufficient training data. SVM models for regression are built on the categories clustered by SOM separately. Parameters of the SVM models are chosen by Particle Swarm Optimization (PSO) algorithm automatically to avoid the arbitrary parameters decision of the tester, improving the forecasting accuracy. The comparison suggests that SOM–SVM–PSO has considerable value in forecasting day-ahead price in Pennsylvania–New Jersey–Maryland (PJM) market, especially for summer peak prices.     

Modified genetic algorithm-based feature selection combined with pre-trained deep neural network for demand forecasting in outpatient department

A well-performed demand forecasting can provide outpatient department (OPD) managers with essential information for staff scheduling and rostering, considering the non-reservation policy of OPD in China. Based on the results reported by relevant studies, most approaches have focused on forecasting the overall amount of patient flow and ignored the demand for other key resources in OPD or similar department. Moreover, few studies have conducted feature selection before training a forecast model, which is a significant pre-processing operation of data mining and widely applied for knowledge discovery in expert and intelligent system. This study develops a novel hybrid methodology to forecast the patients’ demand for different key resources in OPD, by combining a new feature selection method and a deep learning approach. A modified version of genetic algorithm (MGA) is proposed for feature selection. The key operators of normal genetic algorithm are redesigned to utilize useful information provided by filter-based feature selection and feature combinations. A feedforward deep neural network is introduced as the forecast model, and the initial parameter set is generated from a stacked autoencoder-based pre-training process to overcome the optimization challenges in constructing deep architectures. In order to evaluate the performance of our methodology, it is applied to an OPD located at Northeast China. The results are compared with those obtained from combinations of other feature selection methods and demand forecasting models. Compared with GA and PCA, MGA improves the quality and efficiency of feature selection, with less selected features to get higher forecast accuracy. Pre-trained DNN optimally strengthens the advantage of MGA, compared with MLR, ARIMAX and SANN. The combination of MGA and pre-trained DNN possesses strongest predictive power among all involved combinations. Furthermore, the results of proposed methodology are crucial prerequisites for staff scheduling and resource allocation in OPD. Elite features obtained by MGA can provide practical insights on potential association between manifold feature combinations and demand variance.     

基于局部子空间的多标记特征选择算法*

在已有的特征选择算法中,常用策略是通过相关准则选择与标记集合相关性较强的特征,然而该策略不一定是最优选择,因为与标记集合相关性较弱的特征可能是决定某些类别标记的关键特征.基于这一假设,文中提出基于局部子空间的多标记特征选择算法.该算法首先利用特征与标记集合之间的互信息得到一个重要度由高到低的特征序列,然后将新的特征排序空间划分为几个局部子空间,并在每个子空间设置采样比例以选择冗余性较小的特征,最后融合各子空间的特征子集,得到一组合理的特征子集.在6个数据集和4个评价指标上的实验表明,文中算法优于一些通用的多标记特征选择算法.     

Convolutional adaptive denoising autoencoders for hierarchical feature extraction

Convolutional neural networks (CNNs) are typical structures for deep learning and are widely used in image recognition and classification. However, the random initialization strategy tends to become stuck at local plateaus or even diverge, which results in rather unstable and ineffective solutions in real applications. To address this limitation, we propose a hybrid deep learning CNN-AdapDAE model, which applies the features learned by the AdapDAE algorithm to initialize CNN filters and then train the improved CNN for classification tasks. In this model, AdapDAE is proposed as a CNN pre-training procedure, which adaptively obtains the noise level based on the principle of annealing, by starting with a high level of noise and lowering it as the training progresses. Thus, the features learned by AdapDAE include a combination of features at different levels of granularity. Extensive experimental results on STL-10, CIFAR-10, andMNIST datasets demonstrate that the proposed algorithm performs favorably compared to CNN (random filters), CNNAE (pre-training filters by autoencoder), and a few other unsupervised feature learning methods.     

基于CNN彩色图像边缘检测的车牌定位方法

针对现有车牌定位算法准确率不高、步骤多和速度慢等问题, 提出一种彩色图像车牌定位方法(License plate locating based on CNN color edge detection, LPLCCED). 首先利用细胞神经网络(Cell neural network, CNN)模型导出一种与车牌颜色特征相结合的车牌定位专用边缘检测算法, 将车牌的颜色对约束条件融合到边缘检测算法中, 本文专用边缘检测算法可以大大缩小车牌初步定位的范围. 接下来提出一种针对车牌特征的边缘滤波算法, 最后根据车牌结构和纹理特征对候选区域进行判别验证. 该流程的各个环节都可以通过硬件实现, 为面向智能交通领域的实时车牌识别系统的前期车牌定位处理提供了依据.     

Data mining for unemployment rate prediction using search engine query data

Unemployment rate prediction has become critically significant, because it can help government to make decision and design policies. In previous studies, traditional univariate time series models and econometric methods for unemployment rate prediction have attracted much attention from governments, organizations, research institutes, and scholars. Recently, novel methods using search engine query data were proposed to forecast unemployment rate. In this paper, a data mining framework using search engine query data for unemployment rate prediction is presented. Under the framework, a set of data mining tools including neural networks (NNs) and support vector regressions (SVRs) is developed to forecast unemployment trend. In the proposed method, search engine query data related to employment activities is firstly extracted. Secondly, feature selection model is suggested to reduce the dimension of the query data. Thirdly, various NNs and SVRs are employed to model the relationship between unemployment rate data and query data, and genetic algorithm is used to optimize the parameters and refine the features simultaneously. Fourthly, an appropriate data mining method is selected as the selective predictor by using the cross-validation method. Finally, the selective predictor with the best feature subset and proper parameters is used to forecast unemployment trend. The empirical results show that the proposed framework clearly outperforms the traditional forecasting approaches, and support vector regression with radical basis function (RBF) kernel is dominant for the unemployment rate prediction. These findings imply that the data mining framework is efficient for unemployment rate prediction, and it can strengthen government’s quick responses and service capability.     

基于卷积神经网络特征融合的人脸识别算法

为提高卷积神经网络的识别性能,提出了一种基于多种卷积神经网络模型的特征融合方法。论文通过构建一个深度学习网络,将多种卷积神经网络模型如ResNet、InceptionV3和VGG19提取的特征进行融合,并将融合后的特征应用到人脸识别中,据此训练出特征融合网络模型的网络参数;最后利用计算求出的阈值来区分类别。实验结果表明,在人脸库LFW数据集上,论文算法的人脸识别率可达98%;与现有的单一卷积神经网络相比,论文算法识别率更高。     

基于KNN-ANN算法的边际电价预测

在电力市场中,价格一直受到买卖双方的广泛关注。但是,电价影响因素的不确定性给电价的预测带来难度。针对该问题,提出一种通过结合人工神经网络和KNN算法来进行时间序列预测的模型,用KNN算法找出历史数据中相似的数据子序列集合(最近邻),并用人工神经网络来寻找这些最近邻的最优权重,得出预测的时间序列。以美国纽约州电力市场的电价数据进行实验分析,同时比较了利用ARIMA算法以及Naive I预测的结果,证明该方法简单、有效。     

基于邻域粗糙集和帝王蝶优化的特征选择算法

针对经典的帝王蝶优化（MBO）算法不能很好地处理连续型数据,以及粗糙集模型对于大规模、高维复杂的数据处理能力不足等问题,提出了基于邻域粗糙集（NRS）和MBO的特征选择算法。首先,将局部扰动和群体划分策略与MBO算法结合,并构建传输机制以形成一种二进制MBO（BMBO）算法;其次,引入突变算子增强算法的探索能力,设计了基于突变算子的BMBO（BMBOM）算法;然后,基于NRS的邻域度构造适应度函数,并对初始化的特征子集的适应度值进行评估并排序;最后,使用BMBOM算法通过不断迭代搜索出最优特征子集,并设计了一种元启发式特征选择算法。在基准函数上评估BMBOM算法的优化性能,并在UCI数据集上评价所提出的特征选择算法的分类能力。实验结果表明,在5个基准函数上,BMBOM算法的最优值、最差值、平均值以及标准差明显优于MBO和粒子群优化（PSO）算法;在UCI数据集上,与基于粗糙集的优化特征选择算法、结合粗糙集与优化算法的特征选择算法、结合NRS与优化算法的特征选择算法、基于二进制灰狼优化的特征选择算法相比,所提特征选择算法在分类精度、所选特征数和适应度值这3个指标上表现良好,能够选择特征数少且分类精度高的最优特征子集。     

二层分解技术在电价预测中的应用研究

针对电价波动幅度大且预测精度低的问题,提出了二层分解技术与神经网络组成的电价多步预测模型。该模型首先采用集合经验模式分解将原始电价序列分解为一系列分量,变分模态分解将第一层分解产生的最高频率分量进一步分解为一系列模态分量,所有分量采用神经网络模型进行预测,并使用纵横交叉算法对神经网络的参数进行优化,最后叠加所有子序列,得出预测电价值。仿真结果表明,所提出的模型相比其他混合模型具有更好的预测性能,且实用价值高。     

An expert system of price forecasting for used cars using adaptive neuro-fuzzy inference

An expert system for used cars price forecasting using adaptive neuro-fuzzy inference system (ANFIS) is presented in this paper. The proposed system consists of three parts: data acquisition system, price forecasting algorithm and performance analysis. The effective factors in the present system for price forecasting are simply assumed as the mark of the car, manufacturing year and engine style. Further, the equipment of the car is considered to raise the performance of price forecasting. In price forecasting, to verify the effect of the proposed ANFIS, a conventional artificial neural network (ANN) with back-propagation (BP) network is compared with proposed ANFIS for price forecast because of its adaptive learning capability. The ANFIS includes both fuzzy logic qualitative approximation and the adaptive neural network capability. The experimental result pointed out that the proposed expert system using ANFIS has more possibilities in used car price forecasting.     

基于卷积神经网络的发动机故障预测方法

近年来,卷积神经网络(CNN)等深度学习方法的发展为发动机故障诊断和预测带来了新的思路。CNN具有局部连接、权值共享、池化操作以及多层结构等特点,能够有效提取局部特征,降低网络的训练难度,使CNN具有很强的学习能力和特征表达能力。开展了深度卷积神经网络故障预测方法研究,实现了面向发动机气路故障预测算法架构。利用基于发动机试验仿真数据对该方法进行了验证,并与其他几种常见的基于数据驱动的预测方法进行了比较,验证结果表明本文提出的基于卷积神经网络的预测方法具有较好的可行性和效果,可作为开展发动机PHM技术研究的参考。     

基于新的森林优化算法的特征选择算法

针对传统的基于森林优化算法的特征选择算法在初始化阶段、候选森林生成阶段和更新阶段存在的问题,提出了一种新的基于森林优化算法的特征选择算法。该算法在初始化阶段采用皮尔森相关系数和L1正则化方法代替随机初始化策略;在候选森林生成阶段,采用优劣树分开和差额补足的方法解决优劣树不完备问题;在更新阶段,将与最优树精度相同但维度不同的树木添加到森林中。在实验中,所提算法采用与传统的基于森林优化算法的特征选择算法相同的实验数据和实验参数,分别测试了小维度、中维度和大维度数据。实验结果表明,在2个大维度数据和2个中维度数据上,所提算法的分类精度和维度缩减能力均高于传统的基于森林优化算法的特征选择算法。实验结果验证了所提算法在处理特征选择问题的有效性。