基于注意力机制的GRA-EMD-BILSTM锂电池性能衰退趋势预测

随着电动汽车充放电次数的增加，其电池组的电池性能会逐渐衰退，长此以往会影响电动汽车的日常使用，甚至增加电动汽车发生故障的概率。目前针对长里程下的锂电池性能衰退趋势预测工作很少。针对深度学习在预测工作暴露出的不足，提出了一种以注意力机制为基础，联合了灰色关联度分析(GRA)和经验模态分解(EMD)的双向长短期记忆网络(BILSTM)，提出的网络能够有效地解决深度学习中出现的冗余特征、数据噪声对预测模型的影响。实验结果表明，在开源和非开源数据集中，该模型相比其他网络具有更好的预测性能。     

基于组合深度学习的光伏功率日前概率预测模型

为准确量化复杂场景下光伏预测功率的不确定性,提出了一种基于时序卷积网络-注意力机制-长短期记忆网络组合的光伏功率短期概率预测方法。首先,基于多种相关性分析方法选出与光伏功率强相关的气象因素;然后,基于时序卷积网络的特征提取能力和长短期记忆网络的时序特征建模能力,并结合注意力机制和分位数回归,建立组合深度学习预测模型;最后,采用核密度估计方法生成连续概率密度函数。以实际集中式和分布式光伏电站为案例进行分析,结果表明：与长短期记忆网络、时序卷积网络、时序卷积网络-注意力机制和时序卷积网络-长短期记忆网络相比,所提方法在确保最优预测区间的同时,可以提升概率密度预测的性能。     

多通道长短期记忆卷积网络的风速预测

为提高风速的预测性能,提出了多通道长短期记忆网络和卷积网络相结合的风速预测方法。预测模型由多个长短期记忆子网络及卷积网络组成。各子网络选择不同长度的历史数据作为输入,分别实现未来风速值的计算,避免了单一网络输入数据长度参数难以确定的问题。卷积网络将各子网络的计算结果进行卷积、最大池化操作,并通过全连接层计算风速序列的预测值。为避免预测误差累积及漂移,利用误差动态补偿方法对预测值进行校正,获得最终的预测结果。多通道长短期记忆卷积网络可用于风速的超短期预测中,仿真实验结果表明,与现有基于深度学习的预测网络相比,该网络能够更好地拟合实际风速序列的变化趋势,表现出更优的预测性能。     

基于卷积神经网络特征提取的风电功率爬坡预测

为提高风电功率爬坡预测的准确性,提出了一种基于卷积神经网络、长短期记忆网络和注意力机制的风电功率爬坡预测方法。首先,针对风电功率爬坡发生次数少、特征复杂、预测模型难以对小样本爬坡事件有效学习的问题,使用卷积神经网络对风电功率序列进行特征提取。然后,使用长短期记忆网络建立预测模型,解决风电功率的长时依赖问题,并在模型中加入注意力机制对长短期记忆网络单元的输出进行加权,从而加强风电特征的学习,提高爬坡预测准确度。仿真验证表明,模型对风电功率爬坡预测有较高的准确性。     

基于改进Q学习算法和组合模型的超短期电力负荷预测

单一模型在进行超短期负荷预测时会因负荷波动而导致预测精度变差,针对此问题,提出一种基于深度学习算法的组合预测模型。首先,采用变分模态分解对原始负荷序列进行分解,得到一系列的子序列。其次,分别采用双向长短期记忆网络和优化后的深度极限学习机对每个子序列进行预测。然后,利用改进Q学习算法对双向长短期记忆网络的预测结果和深度极限学习机的预测结果进行加权组合,得到每个子序列的预测结果。最后,将各个子序列的预测结果进行求和,得到最终的负荷预测结果。以某地真实负荷数据进行预测实验,结果表明所提预测模型较其他模型在超短期负荷预测中表现更佳,预测精度达到98%以上。     

基于注意力机制和 CNN-BiLSTM 模型的 航空发动机剩余寿命预测

航空发动机作为飞机的主要动力源,其可靠性是保证飞机安全的关键。 剩余使用寿命预测对于提高航空发动机的可用 性和降低其寿命周期成本具有重要意义。 针对现有的预测算法存在对航空发动机多维数据特征提取不足的问题,提出了一种 基于注意力机制的卷积神经网络和双向长短期网络融合模型。 首先,采用卷积神经网络提取特征和双向长短期记忆网络获取 特征中的长短期依赖关系;其次,使用注意力机制来突出特征中的重要部分,提高模型预测的准确率。 为验证所提出方法的有 效性,在 C-MAPSS 数据集上进行了实验。 实验表明,模型可以准确地预测出航空发动机的剩余使用寿命,并比传统方法有着更 高的预测精度。     

基于CNN-BiLSTM网络的锂离子电池健康状态检测方法

锂离子电池健康状态(SOH)是锂离子电池可靠运行的重要参考指标，为提高电池健康状态检测的精确性，提出一种基于CNN-BiLSTM网络的锂电池健康状态检测方法。该方法使用CALCE锂离子电池容量衰减数据集，提取电池健康因子(HI)作为模型输入数据，同时利用灰色关联分析法(GRA)验证HI选取的合理性，采用卷积神经网络(CNN)、双向长短期记忆神经网络(BiLSTM)构建网络模型，对电池容量进行预测，实现锂离子电池健康状态检测。实验结果表明，该方法SOH检测的平均绝对误差为1.3%,均方根误差为1.78%,精确度和可靠性较高。     

基于改进LSTM的脉冲大倍率工况下锂电池SOC估计

锂离子电池是电力系统中不可或缺的重要储能元件。针对脉冲大倍率放电下锂离子电池荷电状态(State of Charge,SOC)预测问题,采用改进的长短期记忆循环神经网络(Long Short-term Memory,LSTM)搭建三元锂电池SOC预测模型。所用方法在原有LSTM基础上增加两个门控单元,通过增强LSTM内部输入和输出的交互,提高模型的动态逼近能力。在脉冲大倍率放电工况下,将所用方法与BP神经网络(Back Propagation,BP)、LSTM神经网络相比较,验证了算法在脉冲放电下的预测性能。实验结果表明,改进方法可准确表征三元锂电池工作特性,满足了SOC估计的实际需求。     

考虑特征重要性值波动的MI-BILSTM短期负荷预测

在短期负荷预测中,含有循环单元的深度学习模型应用广泛,但训练时采用的权值共享结构具有时不变性,忽略了输入特征(气象、日期、历史负荷值等)在不同时刻下给负荷变化带来的不同影响,即权值共享结构无法追踪输入特征的重要性值波动.针对此问题,提出一种考虑特征重要性值波动的互信息(MI)-双向长短期记忆(BILSTM)网络预测方法...     

基于双向长短期记忆深度学习模型的短期风功率预测方法研究

风功率的准确预测对电力系统的规划、调度运行等方面均具有重要意义。该文以风功率预测误差最小为目标,提出了一种基于双向长短期记忆深度学习模型的短期风功率预测方法,包括3层（输入层、隐含层和输出层）网络结构的详细设计以及网络训练过程。输入层负责对原始数据进行预处理以满足网络输入要求,隐含层采用双向长短期记忆单元构建以提取输入数据的非线性特征,输出层提供预测结果,网络训练采用Adam优化方法。在此基础上,基于实际风电场采集数据为算例,对该文所提出模型进行训练与测试,验证了该文所提方法的可行性与优越性。     

Medium term power planning with bilateral contracts

This paper addresses the optimal management of hydropower resources on medium term. The objective is to maximize the expected revenue of a producer, and the decision variables are generation and forward contracts in each period for each scenario. Stochastic linear and nonlinear programming has been used as a framework for modeling and solution. Results are exposited for a Norwegian power producer participating in Nord Pool, the Nordic power exchange.     

Multidimensional linear systems: the extra term

An Erratum has been published for this article in International Journal of Circuit Theory and Applications 1999; 27(5):523. A central role in the theory of discrete‐time linear systems is played by the idea that every such system has an input–output map that can be represented by a convolution or the familiar generalization of a convolution. This thinking involves an oversight that was recently corrected by adding an additional term to the representation. Here we give and discuss a corresponding result for the important case of representations of multidimensional continuous‐space system maps. Copyright © 1999 John Wiley & Sons, Ltd.     

节气负荷预测方法研究

针对电力负荷预测的实际困难，提出了一种进行负荷预测的新思路，即采用节气负荷作为建模数据，并根据负荷呈现出的较为明显的时序性、周期性特点，将数据分离成趋势分量、节气周期分量，以及时间噪声及白噪声，采用双因子ARIMA模型对数据进行拟合，并以BP网络方法完成负荷预测。据此，着重论述了电力负荷预测中建模数据的选择、预处理方法及其对预测精度的影响。     

Fuzzy approach for short term load forecasting

The main objective of short term load forecasting (STLF) is to provide load predictions for generation scheduling, economic load dispatch and security assessment at any time. The STLF is needed to supply necessary information for the system management of day-to-day operations and unit commitment. In this paper, the ‘time’ and ‘temperature’ of the day are taken as inputs for the fuzzy logic controller and the ‘forecasted load’ is the output. The input variable ‘time’ has been divided into eight triangular membership functions. The membership functions are Mid Night, Dawn, Morning, Fore Noon, After Noon, Evening, Dusk and Night. Another input variable ‘temperature’ has been divided into four triangular membership functions. They are Below Normal, Normal, Above Normal and High. The ‘forecasted load’ as output has been divided into eight triangular membership functions. They are Very Low, Low, Sub Normal, Moderate Normal, Normal, Above Normal, High and Very High. Case studies have been carried out for the Neyveli Thermal Power Station Unit-II (NTPS-II) in India. The fuzzy forecasted load values are compared with the conventional forecasted values. The forecasted load closely matches the actual one within ±3%.     

Neural network based short term load forecasting

The artificial neural network (ANN) technique for short-term load forecasting (STLF) has been proposed previously. In order to evaluate ANNs as a viable technique for STLF, one has to evaluate the performance of ANN methodology for practical considerations of STLF problems. The authors make an attempt to address these issues. The results of a study to investigate whether the ANN model is system dependent, and/or case dependent, are presented. Data from two utilities are used in modeling and forecasting. In addition, the effectiveness of a next 24 h ANN model in predicting 24 h load profile at one time was compared with the traditional next 1 h ANN model     

基于短期预测信息和长期值函数近似的大规模电动汽车实时随机优化调度算法

针对大规模电动汽车(Electric Vehicle, EV)和可再生能源接入背景下主动配电网的实时随机调度问题,提出了一种结合短期预测信息和长期值函数近似的双层实时调度模型。为应对大量EV接入后的维数灾问题,首先提出双层调度框架,上层建立EV集群模型,下层根据EV特性提出功率分配算法对每辆EV制定充电计划,实现上层集群指令的完全消纳并满足各EV充电的需求。同时,为应对EV行为、实时电价及可再生能源出力不确定性的问题,实时优化时采用预测算法预测短期内未来接入的EV行为、可再生能源最大出力与实时电价,并通过值函数近似评估短期决策后系统的值函数,从而实现EV集群充电计划、可再生能源调度计划与购电计划的实时分阶段决策。仿真算例表明,所提模型可以实现大规模EV接入下主动配电网的实时随机调度,同时具备良好的鲁棒性。     

基于MPC的超短期优化调度策略研究

为实现多时间尺度协调调度模式,基于模型预测控制(MPC)理论对超短期优化调度策略进行了研究。以系统发电机组运行成本和排放量费用最小为目标函数,通过求解前瞻时间内的多时段优化问题,为实时调度提供初始策略,以实际调度结果和新的预测信息作为反馈信息进行滚动优化调度。该方法将日前调度时间尺度和实时调度时间尺度的信息联系起来,可有效应对预测信息波动对系统调度的影响。算例仿真结果表明:考虑多时段整体优化,可以更好地在机组之间合理的优化分配负荷,提高整个火电厂运行的经济性。同时验证了MPC方法的鲁棒性和收敛性。     

Sensitivity analysis of an optimal short term hydro-thermalschedule

The authors deal with the sensitivity analysis of an optimal short term hydro-thermal schedule. Methodological and practical aspects of the system economy loss (SEL) calculation caused by incorrect input data are investigated. It was shown and checked by numerical simulations that errors in the hydro plant characteristics can be studied through errors in the short term load forecasting. In a special case when differentiability conditions are satisfied (economic dispatching problem) analytical formulas for the SEL are developed. Some of the important results of sensitivity calculations concerning relevant errors in thermal cost coefficients, hydro plant characteristics, forecasted water inflows and forecasted system demands are presented. The critical input data have been selected and their tolerance margins established to preserve the beneficial effects of an optimal short term schedule     

Long term distribution network planning considering urbanity uncertainties

This paper addresses the problem of long term distribution network planning under urbanity uncertainties. Unpredictable urbanity plans are expected facts in developing/under developed countries. This type of uncertainties make it difficult to implement designed network in the future and leads to increasing operational costs including loss and outage costs. In this paper we presented a novel approach for distribution network planning which in addition of eliminating harmful effects of urbanity uncertainties, leads to easy management and operation of resulted network. In this approach several points of study region with high accessibility are selected as candidate embranchment points and optimal connection configuration of load points to the embranchment points is determined by genetic algorithm considering investment, loss and also customer interruption costs. Afterward, final structure of network is designed by branch exchange method considering the embranchment points as representative to load points in their service area. The performance of the proposed approach is assessed on a test distribution network.     

短期风速-风电功率预测方法

传统灰色风速预测模型累加处理时不能预测突变风速,使风电功率预测误差过大.采用数值逼近算法对传统灰色GM(1,1)预测模型进行优化改进,以优化的灰色GM(1,1)预测模型对未来时段风速进行预测,突变风速预测误差降低了34.3％.再将优化风速预测模型和时间序列动态神经网络相结合,构建出风电功率预测模型.应用该模型对酒泉地区某风电场现场数据进行仿真测试,预测效果可信度大于93％.