TIME SERIES AND GREY MODEL DIAGNOSIS METHOD USED TO CRACK PROBLEMS

In this paper,the vibration signals in the fatigue crack growth process in a chinese steel used in a mining machinery were analyzed by the frequency spectrum, the time series and grey system model,and the critical criterion for crack initiation was proposed.

Fault Prediction Based on Dynamic Model and Grey Time Series Model in Chemical Processes

This paper combines grey model with time series model and then dynamic model for rapid and in-depth fault prediction in chemical processes. Two combination methods are proposed. In one method, historical data is introduced into the grey time series model to predict future trend of measurement values in chemical process. These predicted measurements are then used in the dynamic model to retrieve the change of fault parameters by model based diagnosis algorithm. In another method, historical data is introduced directly into the dynamic model to retrieve historical fault parameters by model based diagnosis algorithm. These parameters are then predicted by the grey time series model. The two methods are applied to a gravity tank example. The case study demonstrates that the first method is more accurate for fault prediction.

Grey series time-delay predicting model in state estimation for power distribution networks

A new combined model is proposed to obtain predictive data value applied in state estimation for radial power distribution networks. The time delay part of the model is calculated by a recursive least squares algorithm of system identification, which can gradually forget past information. The grey series part of the model uses an equal dimension new information model (EDNIM) and it applies 3 points smoothing method to preprocess the original data and modify remnant difference by GM(1,1). Through the optimization of the coefficient of the model, we are able to minimize the error variance of predictive data. A case study shows that the proposed method achieved high calculation precision and speed and it can be used to obtain the predictive value in real time state estimation of power distribution networks.

Combined hybrid energy storage system and transmission grid model for peak shaving based on time series operation simulation

This study proposes a combined hybrid energy storage system(HESS) and transmission grid(TG) model, and a corresponding time series operation simulation(TSOS) model is established to relieve the peak-shaving pressure of power systems under the integration of renewable energy. First, a linear model for the optimal operation of the HESS is established, which considers the different power-efficiency characteristics of the pumped storage system, electrochemical storage system, and a new type of liquid compressed air energy storage. Second, a TSOS simulation model for peak shaving is built to maximize the power entering the grid from the wind farms and HESS. Based on the proposed model, this study considers the transmission capacity of a TG. By adding the power-flow constraints of the TG, a TSOS-based HESS and TG combination model for peak shaving is established. Finally, the improved IEEE-39 and IEEE-118 bus systems were considered as examples to verify the effectiveness and feasibility of the proposed model.

Hybrid grey model to forecast monitoring series with seasonality

The grey forecasting model has been successfully applied to many fields. However, the precision of GM(1,1) model is not high. In order to remove the seasonal fluctuations in monitoring series before building GM(1,1) model, the forecasting series of GM(1,1) was built, and an inverse process was used to resume the seasonal fluctuations. Two deseasonalization methods were presented , i.e., seasonal index-based deseasonalization and standard normal distribution-based deseasonalization. They were combined with the GM(1,1) model to form hybrid grey models. A simple but practical method to further improve the forecasting results was also suggested. For comparison, a conventional periodic function model was investigated. The concept and algorithms were tested with four years monthly monitoring data. The results show that on the whole the seasonal index-GM(1,1) model outperform the conventional periodic function model and the conventional periodic function model outperform the SND-GM(1,1) model. The mean Absolute error and mean square error of seasonal index-GM(1,1) are 30.69% and 54.53% smaller than that of conventional periodic function model, respectively. The high accuracy, straightforward and easy implementation natures of the proposed hybrid seasonal index-grey model make it a powerful analysis technique for seasonal monitoring series.

Research on Hydrological Time Series Prediction Based on Combined Model

Water level prediction of river runoff is an important part of hydrological forecasting.The change of water level not only has the trend and seasonal characteristics,but also contains the noise factors.And the water level prediction ability of a single model is limited.Since the traditional ARIMA(Autoregressive Integrated Moving Average)model is not accurate enough to predict nonlinear time series,and the WNN(Wavelet Neural Network)model requires a large training set,we proposed a new combined neural network prediction model which combines the WNN model with the ARIMA model on the basis of wavelet decomposition.The combined model fit the wavelet transform sequences whose frequency are high with the WNN,and the scale transform sequence which has low frequency is fitted by the ARIMA model,and then the prediction results of the above are reconstructed by wavelet transform.The daily average water level data of the Liuhe hydrological station in the Chu River Basin of Nanjing are used to forecast the average water level of one day ahead.The combined model is compared with other single models with MATLAB,and the experimental results show that the accuracy of the combined model is improved by 7%compared with the traditional wavelet network under the appropriate wavelet decomposition function and the combined model parameters.

Predicting the Number of Beijing Science and Technology Personnel Based on GM(1,N) Model

In this paper, based on the Science and Technology Statistics in Beijing Statistical Yearbook, grey theory is used to study the relationship among S&T (Science and Technology) activities personnel, R&D (research and development) personnel FTE (Full Time Equivalent), intramural expenditure for R&D and Patent Application Amount. According to the grey correlation coefficient, screening of grey GM(1,N) prediction variables, the grey prediction model is established. Meanwhile, time series model and GM(1,1) model are established for patent applications and R&D personnel equivalent FTE. By comparing the simulating results with the real data, the absolute relative error of prediction models is less than 10%. The results of the prediction model are tested. In order to improve the prediction accuracy, the mean values of the predicted values of the two models are brought into the GM(1,N) model to predict the number of scientific and technical personnel in Beijing during 2015-2025. Forecast results show that the number of science and technology personnel in Beijing will grow with exponential growth trend in the next ten years, which has a certain reference value for predicting the science and technology activities and formulating the policy in Beijing.

考虑时序特征的深圳港集装箱吞吐量组合方法预测

集装箱吞吐量预测对港口企业运营及决策具有重要的作用。传统集装箱吞吐量预测方法存在预测精度不高的缺点。为解决这一问题,提出了一种考虑季节性和不确定性的SARIMA-XGBoost组合预测方法。针对集装箱吞吐量的季节性特征,选取季节性自回归移动平均模型(seasonal autoregressive integrated moving average model,SARIMA)捕捉周期性特征和线性特征;针对集装箱吞吐量中的不确定性因素,选取极致梯度提升树算法(extreme gradient boosting,XGBoost)自适应学习时间序列数据中的复杂模式和非线性特征。通过选取优化指标并计算分配权重的方式实现了预测模型中线性和非线性特征的有效融合,从而提升预测精度。通过对深圳港2013—2022年集装箱吞吐量月度数据进行实证研究和对比分析,结果表明SARIMA-XGBoost组合方法预测精度最高、稳定性好,验证了该组合方法在集装箱吞吐量预测中的有效性。

配电网负荷预测中信号分解和预测模型组合的双层优化策略

负荷时间序列的波动性和非线性特征的加剧对负荷预测方法提出了更高的要求,而常规组合预测方法针对海量负荷数据存在应用局限性问题。为此,提出了配电网负荷预测中时序分解方法和预测模型组合的双层优化策略。首先针对某一负荷预测数据,在时序信号分解层配置权重,以负荷均方根误差最小寻优各分解方法的权重系数,进而获得各时序信号分解方法的最优组合;在此基础上,在预测模型层进行组合方案寻优,通过配置权重系数以获得各预测模型的最优组合,进一步提升负荷预测的精度。仿真结果表明,所提策略可根据预测对象的特征优化组合各信号分解方法和预测模型,降低了配电网负荷序列的非平稳性对预测精度的影响。

基于多模型比选耦合的降水预测

变化环境下水文时间序列的模拟预测难度不断加大,以往研究大多聚焦在模型的不同组合尝试与应用探索,但缺乏针对不同组合模型适用性与稳定性的系统研究。选择经验模态分解(empirical mode decomposition,EMD)、总体平均经验模态分解(ensemble empirical mode decomposition,EEMD)、改进的总体平均经验模态分解(modified ensemble empirical mode decomposition,MEEMD)和变分模态分解(variational mode decomposition,VMD)4种常用的分解算法,与多元线性回归(multivariable linear regression,MLR)、随机森林(random forest,RF)、BP神经网络(back propagation,BP)、卷积神经网络(convolutional neural networks,CNN)和长短期记忆神经网络(long short-term memory,LSTM)5种具有代表性的模型结合,构建20种基于“分解-预测-重构”模式的组合模型,并以华北地区密云、官厅两流域年和汛期降水为例,进行模型适用性与稳定性综合对比分析。结果表明:单一模型对密云流域年降水和汛期降水的预测结果优于官厅流域,但整体预测结果均不理想;结合分解算法后的组合模型预测结果明显优于单一模型,且该预测结果存在正负误差抵消现象,因此有助于进一步提高组合模型的整体预测精度;与基于EMD系列的分解算法相比,VMD算法对模型预测精度提升效果最显著,组合模型适用性和稳定性整体上表现为VMD-MLR>VMD-LSTM>VMD-BP>VMD-CNN。

基于SVM-STL-LSTM的区域短期电力负荷预测研究

针对区域电力负荷的时间序列数据随机性强、预测精度低及单一模型的数据特征提取能力差等问题,提出了一种支持向量机(SVM)、STL时序分解法、长短期记忆神经网络(LSTM)组合的电力负荷预测模型。该模型利用SVM对时间序列的电力负荷数据进行初始预测,并通过STL时序分解法对残差序列进行时序分解,从而提高残差序列的稳定性,减小其随机性,最后用LSTM对SVM的预测误差进行修正。试验结果证明,该方法利用误差修正可有效处理随机性强的数据,有利于预测结果的稳定性,提高预测精度。

基于自回归模型的RBCC隔离段激波串位置识别与压力值预估

为了清楚客观地判断火箭基组合循环发动机(Rocket-based combined-cycle,RBCC)隔离段激波串位置,将Ma=6,4,3.5工况下直连试验中实测得到的RBCC隔离段测压点压力数据按照时间的先后顺序排列形成一时间序列,建立自回归(Auto-Regressive,AR)模型并计算赤池信息准则(Akaike information criterion,AIC)值,完成了不同工况下激波串前缘位置的识别。研究表明:当隔离段测压点没有受到激波串影响时,实时压力值仅存在微弱波动,模型AIC值变化较为平稳;当激波串运动至测压点处时,该点压力升高,振荡幅度明显增加,AIC值随之瞬时增大。取同一时间段内发动机沿程测压点中首个AIC值增加500以上,并在不改变工况的情况下始终保持较大值的测点位置为激波串前缘位置。与压比法相比,时间序列分析法能敏感监测到实时压力值的升高和振荡,激波串前缘位置识别更为准确。通过建立自回归模型还可以实现激波串内部压力值预估,记录连续160 ms内Ma=6,4,3.5工况下测压点压力数据,采样频率1 kHz,使用前80 ms数据建立自回归模型,完成后80 ms压力值预估及准确性检验,得到三个工况下预估平均误差分别为3.21%,7.68%,6.49%。

基于邻域互信息的组合预测最优子集选择算法

为在候选模型集中高效地选择时间序列组合预测的最优子集,提出一种CSPSO-NMI-MRMR最优子集选择算法。利用邻域互信息(neighborhood mutual information, NMI)度量相关性和冗余度,避免数值型数据的离散化,按最大相关最小冗余原则(minimal redundancy and maximal relevance, MRMR)筛选最优子集;邻域互信息中的邻域参数与子集选择效果密切相关,采用CSPSO算法寻找最优邻域参数,充分利用布谷鸟算法(cuckoo search, CS)和粒子群优化算法(particle swarm optimization, PSO)的优势,兼顾搜索效率和全局搜索能力;在寻参过程中设计一种淘汰策略,优化邻域参数的寻优区间并淘汰部分单模型,减少计算量。仿真结果表明,所提方法在预测精度、运行时间和稳健性上效果更优。

基于ConvLSTM-LSTM的短期负荷预测方法

长短时记忆(LSTM)网络和结合卷积神经网络(CNN)的CNN-LSTM预测模型由于其网络模型本身的缺陷,限制了预测精度的提高。针对以上问题,提出了一种结合卷积长短时记忆(ConvL⁃STM)网络的ConvLSTM-LSTM负荷预测模型。利用ConvLSTM网络充分提取时序特征,将提取到的信息输入到LSTM网络中进行进一步的选择性记忆,并输出预测结果。将该模型与CNN-LSTM网络模型、LSTM网络模型、以及门控循环单元(GRU)网络模型进行了对比,结果显示所构建的Con⁃vLSTM-LSTM模型的预测效果均优于对比模型,在精度评价指标平均绝对百分比误差(MAPE)上,分别减小了1.10%、1.54%、1.91%。

灰色Verhulst-时间序列组合模型在沉降监测中的应用

针对传统单一的灰色模型、时间序列模型在进行沉降监测数据模型存在精度较低的问题,本文提出了一种灰色Verhulst模型和时间序列模型的组合优化模型。首先,对灰色Verhulst模型、时间序列模型的基本理论进行阐述;其次,建立了灰色Verhulst-时间序列模型,即以灰色Verhulst模型得到的样本数据拟合值作为时间序列模型进行预测的样本值;最后,通过以某建筑物沉降监测点为例进行计算并进行灰色Verhulst模型、时间序列模型以及灰色Verhulst-时间序列模型的预测结果对比。实验结果表明,灰色Verhulst-时间序列组合优化模型的预测精度最高,均方差为0.3275 mm,验证了该组合模型更贴合建筑物沉降的变形趋势。

时间序列模型在经济分析中的应用——陕西省GDP分析与预测

相对精准的GDP分析与预测可以为国家或地区在进行经济发展战略部署及相关发展措施的制定方面提供重要参考依据。文章借助时间序列的相关定义和方法,分析和模拟了影响陕西省GDP的主要因素及其与第一、二、三产业的关系,并进一步采用数据分析方法,建立不同的模型并加以比较,得出了最佳的预测模型,并对第一、二、三产业与GDP进行了预测。

基于时间序列模型与灰色模型的组合预测模型的研究

为了能有效地提高预测模型的精度,提出了组合预测模型。本文首先利用ARIMA模型对时间序列数据进行模型的识别和拟合,然后由比较可知优化后的GM(1,1)模型拟合和预测效果好于GM(1,1)模型,最后通过赋予合理权重结合ARIMA模型和优化后的GM(1,1)模型两种方法得到ARIMA-GM的组合预测模型。预测结果表明:组合模型的预测准确性高于各个模型单独使用时的准确性,组合模型发挥了各个单一模型的优势。

奇异谱分析在故障时间序列分析中的应用

由于日益增长的飞行安全和飞机维护质量需求,飞机使用可靠性已经成为一个重要的研究领域。从某航空公司波音737飞机使用过程中现场所记录的18年的故障数据出发,应用奇异谱分析(SSA)方法,对故障时间序列进行了建模和预测,进一步以预测结果的均方根误差(RMSE)最小为优化目标对SSA模型参数进行了优选。在此基础上,提出了一种更为广泛的模型组合方法和实现算法,这种方法采用不同的时间序列模型来构造SSA分解出的趋势、周期和残差等成分。通过与三次指数平滑(Holt-Winters)、自回归移动平均(ARIMA)2种时间序列模型的实验结果对比,SSA及其参数优选和模型组合方法在故障时间序列分析中具有更好的拟合和预测精度。

考虑气象因素的相似聚类短期负荷组合预测方法

提出以气象负荷和长期趋势负荷之和为聚类中心对历史负荷数据进行相似搜索的方法,该方法可提高预测样本同被预测日负荷的相似度,从而增加预测结果的可信度和精确度。以权重优化组合的方式采用多种负荷预测方法进行组合负荷预测,应用实例证明该方法可体现不同地区、不同类型、不同气象敏感度的负荷特性,因而具有广泛的自适应性,对于负荷总量较小、变动范围较大且受天气因素影响明显的地区具有较好的预测精度。

组合预测模型在猪肉价格预测中的应用研究

本文在分析了神经网络、灰系统和时间序列预测模型的基础上,设计了将其中两种模型组合的预测方法。该方法的主要思想是利用回归预测思想将预测分为因素预测和结果预测两部分,并分别采用不同预测模型进行预测,从而达到提高预测精度的目的。利用该方法对吉林省近期的生猪价格进行预测,实验结果表明,该方法比单个预测方法有更好的预测效果,并且通过对不同组合的实验结果的分析发现,灰系统与神经网络相结合的方法具有更高的预测精度。