5万m~2地下特大型支撑爆破拆除

介绍了在复杂环境下爆破拆除一地下特大钢筋混凝土支撑的技术难点。由于合理选取爆破参数,采取孔内高段、孔外低段毫秒微差起爆网路,安全防护采取覆盖、近体、保护性三种措施,有效地阻止了飞石对周围建筑物的损害,并对爆破可能产生的危害进行了科学验算,最后分多次爆破圆满完成拆除任务。     

基于VMD和改进TCN的短期光伏发电功率预测

光伏发电功率存在波动性,且光伏出力易受各种气象特征影响,传统TCN网络容易过度强化空间特性而弱化个体特性。针对上述问题,文中提出一种基于VMD和改进TCN的短期光伏发电功率预测模型。通过VMD将原始光伏发电功率时间序列分解为若干不同频率的模态分量,将各个模态分量以及相对应的气象数据输入至改进TCN网络进行建模学习。利用中心频率法确定VMD的最优分解模态分解个数。在传统TCN预测模型的基础上,使用DropBlock正则化取代Dropout正则化以达到抑制卷积层中信息协同的效果,并引入注意力机制自主挖掘并突出关键气象输入特征的影响,量化各气象因素对光伏发电的影响,从而提高预测精度。以江苏省某光伏电站真实数据为例进行仿真实验,结果表明所提预测方法的RMSE为0.62 MW,MAPE为2.03%。     

Efficient utilization of waste heat from molten carbonate fuel cell in parabolic trough power plant for electricity and hydrogen coproduction

Direct steam generating parabolic trough power plant is an important technology to match future electric energy demand. One of the problems related to its emergence is energy storage. Solar-to-hydrogen is a promising technology for solar energy storage. Electrolysis is among the most processes of hydrogen production recently investigated. High temperature steam electrolysis is a clean process to efficiently produce hydrogen. In this paper, steam electrolysis process using solar energy is used to produce hydrogen. A heat recovery steam generator generates high temperature steam thanks to the molten carbonate fuel cell's waste heat. The analytical study investigates the energy efficiency of solar power plant, molten carbonate fuel cell and electrolyser. The impact of waste heat utilization on electricity and hydrogen generation is analysed. The results of calculations done with MATLAB software show that fuel cell produces 7.73 MWth of thermal energy at design conditions. 73.37 tonnes of hydrogen and 14.26 GWh of electricity are yearly produced. The annual energy efficiency of electrolyser is 70% while the annual mean electric efficiency of solar power plant is 18.30%.The proposed configuration based on the yearly electricity production and hydrogen generation has presented a good performance.     

Application of artificial neural network method to predict the breakage properties of PGE bearing chromite ore

In the present investigation, systematic grinding experiments were conducted in a laboratory ball mill to determine the breakage properties of low-grade PGE bearing chromite ore. The population balance modeling technique was used to study the breakage parameters such as primary breakage distribution (B_{i, j}) and the specific rates of breakage (S_i). The breakage and selection function values were determined for six feed sizes. The results stated that the breakage follows the first-order grinding kinetics for all the feed sizes. It was observed that the coarser feed sizes exhibit higher selection function values than the finer feed size. Further, an artificial neural network was used to predict breakage characteristics of low-grade PGE bearing chromite ore. The predicted results obtained from the neural network modeling were close to the experimental results with a correlation of determination R² = 0.99 for both product size and selection function.     

Systematic analysis and multi-objective optimization of an integrated power and freshwater production cycle

This study represents the results of the analysis and optimization of an integrated system for cogenerating electricity and freshwater. This setup consists of a Solid Oxide Fuel cell (SOFC) for producing electricity. Unburned fuel of the SOFC is burned in the afterburner to increase the temperature of the SOFC's outlet gasses and operate a Gas turbine (GT) to produce additional power and operate the air compressor. At the bottom of this cycle, a combined setup of a Multi-Effect Desalination (MED) and Reverse Osmosis (RO) is considered to produce freshwater from the unused heat capacity of the GT's exhaust gasses. Also, a Stirling engine is used in the fuel supply line to increase the fuel's temperature. Using LNG and the Stirling engine will replace the fuel compressor with a pump which increases the system performance and eliminates the need for the expansion valve. To study the system performance a mathematical model is developed in Engineering Equation Solver (EES) program. Then, the system's simulated data from the EES has been sent to MATLAB to promote the best operating condition based on the optimization criteria. An energetic, exergetic, economic, and environmental analysis has been performed and a Non-dominated Sorting Genetic Algorithm (NSGA-II) is used to achieve the goal. The two-objective optimization is performed to maximize the exergetic efficiency of the proposed system while minimizing the system's total cost of production. This cost is a weighted distribution of the Levelized Cost of Electricity (LCOE) and Levelized Cost of freshwater (LCOW). The results showed that the exergetic and energetic efficiencies of the system can reach 73.5% and 69.06% at the optimum point. The total electricity production of the system is 99 MW. The production cost is 11.71 Cents/kWh, of which 1.04 Cents/kWh is emission-related and environmental taxes. The freshwater production rate is 42.44 kg/s which costs 4.38 USD/m³.     

基于卷积神经网络的5G蜂窝网络无线定位方法

5G蜂窝网络发展迅猛，其覆盖面积将逐渐增大，因此使用5G蜂窝网络进行定位是有研究潜力的研究方向。本文提出一种新的深度学习技术来实现高效、高精度和低占用的定位，以代替传统指纹定位过程中繁重的指纹库生成以及距离计算。该方法建立了一个特殊的卷积神经网络，并根据5G天线信号的接收信号强度指示、相位和到达角等特征量，选择合适的输入数据格式构造样本组建训练集，对该卷积神经网络进行训练。训练得到的卷积神经网络可以替代指纹定位中的庞大指纹库，非常有利于直接在5G移动设备端实现定位。虽然卷积神经网络在训练过程中需要大量时间，但在训练完毕后直接进行分类定位的速度非常快，可以保障定位实现的实时性。本文所实现的卷积神经网络权重与偏置所占内存不到0.5 MB，且能够在实际应用环境中以95%的定位准确率以及0.1 m的平均定位精度实现高精度定位。     

疫情下电磁场理论课程的网络教学

本文主要总结了新冠疫情期间作者的电磁场理论课程在线教学经验。对比分析了录播和直播的优缺点后,选择录播教学方式。基于超星网络教学平台,展示了录播网络教学的具体措施,包括网上答疑和学习效果检查以及在线批改作业等。给出了网络教学可以为线下教学继续使用的方法和手段,为疫情结束后的正常教学提供了新的网络教学补充措施。     

5G核心网业务模型的智能化预测研究

摘 要：核心网业务模型的建立是5G网络容量规划和网络建设的基础，通过现有方法得到的理论业务模型是静态不可变的且与实际网络存在偏离。为了克服现有5G核心网业务模型与现网模型适配性较差以及规划设备无法满足用户实际业务需求的问题，提出了一种长短期记忆（long short-term memory，LSTM）网络与卷积LSTM （convolution LSTM，ConvLSTM）网络双通道融合的 5G 核心网业务模型预测方法。该方法基于人工智能（artificial intelligence，AI）技术以实现高质量的核心网业务模型的智能预测，形成数据反馈闭环，实现网络自优化调整，助力网络智能化建设。     

Learn decision trees with deep visual primitives

In this paper, we strive to propose a self-interpretable framework, termed PrimitiveTree, that incorporates deep visual primitives condensed from deep features with a conventional decision tree, bridging the gap between deep features extracted from deep neural networks (DNNs) and trees’ transparent decision-making processes. Specifically, we utilize a codebook, which embeds the continuous deep features into a finite discrete space (deep visual primitives) to distill the most common semantic information. The decision tree adopts the spatial location information and the mapped primitives to present the decision-making process of the deep features in a tree hierarchy. Moreover, the trained interpretable PrimitiveTree can inversely explain the constituents of the deep features, highlighting the most critical and semantic-rich image patches attributing to the final predictions of the given DNN. Extensive experiments and visualization results validate the effectiveness and interpretability of our method.