一种新型蓄电池充电控制策略研究

在风光互补发电系统中,蓄电池作为风光互补发电系统的储能设备,在整个发电系统中起着非常重要的作用,而蓄电池充电控制方法的优劣,是影响其使用寿命的一个重要因素。文章对风光互补发电系统中蓄电池的常用充电控制方法进行了分析,提出了一种新的控制策略,即将蓄电池充电电压、充电电流采样值分别与蓄电池事先设定好的恒压过充电压、恒流充电限流值和浮充电流值进行实时比较,来判断采取哪种合理的充电方式,对蓄电池起到尽可能的保护,从而延长了风光互补发电系统中蓄电池使用寿命。     

基于H.264的改进阈值的可变自适应十字搜索算法

H.264视频编解码标准的算法复杂度高,不能满足视频监控和无线传输中要求的高压缩比、低码率和实时性的要求,通过分析可知运算量最大的是运动估计模块,本文对该模块的算法进行优化。在ARPS算法的基础上结合预测的初始点和自适应调节臂长的搜索策略的优点,采用可变阈值的早期终止策略和简单的搜索模板的算法。实验表明该算法在有效减少搜索点数的情况下并不会对图像质量产生较大的影响,提高了运动估计算法的搜索速率和鲁棒性。     

防止过拟合的属性约简

近年来,机器学习的过拟合问题备受关注,尤其在属性约简中。为解决这一难题,提出一种融合集成策略和去除操作的算法。首先将训练模型数据平分为◢M份;然后将其中M◣-1份采用集成策略进行潜在约简计算;最后将剩余的一份进行提前测试,一旦发生过拟合则将刚加入的属性从潜在约简集中去除。利用提前测试潜在属性约简的方法来防止过拟合现象的发生,几组UCI数据的实验结果说明了新算法的有效性,同时为丰富和发展属性约简提供了一种新的方向。     

SIMOP: Efficient reactive distillation optimization using stochastic optimizers

A methodology to improve the efficiency of stochastic methods applied to the optimization of chemical processes with a large number of equality constraints is presented. The methodology is based on two steps: (a) the optimization of the simulation step, which involves the optimum choice of design variables and subsystems to be simultaneously solved; (b) the optimization of the nonlinear programming (NLP) problem using stochastic methods. For the first step a flexible tool (SIMOP) is used, whereby different numerical procedures can be easily obtained, taking into account the problem formulation and specific characteristics, the need for specific initialization schemes and the efficient solution of systems of nonlinear equations. This methodology was applied to the optimization of a reactive distillation process for the production of ethylene glycol. Due to the complexity of the mathematical model, several different numerical procedures were generated, and their influence on the computational burden and on the reliability and accuracy of the optimization to reach the global optimum were studied. The results obtained suggest that in addition to the choice of design variables, the structure of subsystems associated to numerical procedures has a considerable impact on the performance of the optimizers.     

Fast and robust face recognition via coding residual map learning based adaptive masking

Robust face recognition (FR) is an active topic in computer vision and biometrics, while face occlusion is one of the most challenging problems for robust FR. Recently, the representation (or coding) based FR schemes with sparse coding coefficients and coding residual have demonstrated good robustness to face occlusion; however, the high complexity of l₁-minimization makes them less useful in practical applications. In this paper we propose a novel coding residual map learning scheme for fast and robust FR based on the fact that occluded pixels usually have higher coding residuals when representing an occluded face image over the non-occluded training samples. A dictionary is learned to code the training samples, and the distribution of coding residuals is computed. Consequently, a residual map is learned to detect the occlusions by adaptive thresholding. Finally the face image is identified by masking the detected occlusion pixels from face representation. Experiments on benchmark databases show that the proposed scheme has much lower time complexity but comparable FR accuracy with other popular approaches.     

Towards adaptive learning with improved convergence of deep belief networks on graphics processing units

In this paper we focus on two complementary approaches to significantly decrease pre-training time of a deep belief network (DBN). First, we propose an adaptive step size technique to enhance the convergence of the contrastive divergence (CD) algorithm, thereby reducing the number of epochs to train the restricted Boltzmann machine (RBM) that supports the DBN infrastructure. Second, we present a highly scalable graphics processing unit (GPU) parallel implementation of the CD-k algorithm, which boosts notably the training speed. Additionally, extensive experiments are conducted on the MNIST and the HHreco databases. The results suggest that the maximum useful depth of a DBN is related to the number and quality of the training samples. Moreover, it was found that the lower-level layer plays a fundamental role for building successful DBN models. Furthermore, the results contradict the pre-conceived idea that all the layers should be pre-trained. Finally, it is shown that by incorporating multiple back-propagation (MBP) layers, the DBNs generalization capability is remarkably improved.     

Adaptive block size for dense QR factorization in hybrid CPU–GPU systems via statistical modeling

QR factorization is a computational kernel of scientific computing. How can the latest computer be used to accelerate this task? We investigate this topic by proposing a dense QR factorization algorithm with adaptive block sizes on a hybrid system that contains a central processing unit (CPU) and a graphic processing unit (GPU). To maximize the use of CPU and GPU, we develop an adaptive scheme that chooses block size at each iteration. The decision is based on statistical surrogate models of performance and an online monitor, which avoids unexpected occasional performance drops. We modify the highly optimized CPU–GPU based QR factorization in MAGMA to implement the proposed schemes. Numerical results suggest that our approaches are efficient and can lead to near-optimal block sizes. The proposed algorithm can be extended to other one-sided factorizations, such as LU and Cholesky factorizations.     

建筑卫生陶瓷行业发展战略选择

通过对佛山、东莞等地的陶瓷产区相关企业的调查研究,在SWOT分析的基础上,研究了相关的竞争战略,归纳总结出现阶段适合建筑卫生陶瓷企业的战略:即以低成本为竞争生存方式,差异性为竞争发展动力,国际化为发展长期最终目标。     

ZW多产品间歇过程调度及在线调整

针对多产品间歇过程调度提出了分层递阶的Petri网建模方法，利用赋时Petri和Petri网的简化技术描述不同层次的生产问题，具有很强的模型描述能力。另外，在调度决策层可以方便地集成优化策略和在线调整算法，使问题求解更加灵活。