Prediction of full load electrical power output of a base load operated combined cycle power plant using machine learning methods

Predicting full load electrical power output of a base load power plant is important in order to maximize the profit from the available megawatt hours. This paper examines and compares some machine learning regression methods to develop a predictive model, which can predict hourly full load electrical power output of a combined cycle power plant. The base load operation of a power plant is influenced by four main parameters, which are used as input variables in the dataset, such as ambient temperature, atmospheric pressure, relative humidity, and exhaust steam pressure. These parameters affect electrical power output, which is considered as the target variable. The dataset, which consists of these input and target variables, was collected over a six-year period. First, based on these variables the best subset of the dataset is explored among all feature subsets in the experiments. Then, the most successful machine learning regression method is sought for predicting full load electrical power output. Thus, the best performance of the best subset, which contains a complete set of input variables, has been observed using the most successful method, which is Bagging algorithm with REPTree, with a mean absolute error of 2.818 and a Root Mean-Squared Error of 3.787.     

An ultra-energy-efficient crosstalk-immune interconnect architecture based on multilayer graphene nanoribbons for deep-nanometer technologies

An ultra-energy-efficient interconnect structure based on multilayer graphene nanoribbon (MLGNR) interconnects for deep-nanometer technologies is proposed herein. First, a low-swing interconnect based on MLGNRs and high-performance interface circuits using carbon nanotube field-effect transistors (CNTFETs) is proposed. Then, an ultra-energy-efficient interconnect structure is obtained by actively shielding such low-swing lines. The structures under study are simulated comprehensively at the 7-nm technology node. The results indicate that the MLGNR interconnect is significantly more energy efficient than its multiwall carbon nanotube (MWCNT) counterpart in the low-voltage regime. Moreover, the proposed approach is superior to its MLGNR counterparts. The proposed structure leads to 86%, 75%, and 31% lower energy consumption over a length of 500 µm as compared with the typical, actively shielded, and low-swing MLGNR interconnects, respectively. Moreover, the impact of the ratio of the widths of the signal line to the shield line on the performance of the interconnects is evaluated. The energy consumption reduction achieved by the proposed approach is mostly preserved even when using minimum-width shield lines on wider signal lines to reduce the area overhead. Moreover, the impact of process variations on the performance of the interconnects is assessed using Monte Carlo simulations, demonstrating the robustness of the proposed approach.

     

Sensors based on saw resonator with “nontraditional formats” of output signal

It is demonstrated that the sensitivity of sensors based on two-port surface acoustic waves (SAW) resonators increases by a factor of Q (Q-loaded quality factor of the resonator) if the output signal of these sensors is represented in the phase format instead of the conventional frequency format. Practical possibility of use of the SAW gas sensor with phase format of transmitted signal as output signal is demonstrated using procedure of the determination of the water contents in gaseous nitrogen with 386 MHz ST-quartz SAW resonator as an example. It is found that the SAW power affects the sensor response owing to the influence of the SAW intensity on the equilibrium amount of water molecules adsorbed on the resonator surface. Extremely high sensor sensitivity was obtained when using phase format of reflected signal as a sensor output. Using such a format of sensor output we have reliably measured the change of temperature of ST-quartz SAW sensor substrate as small as 0.01^∘C.     

双通道USB接口振动测试系统研制

针对机械臂,数控机床等加工设备的振动测试问题,研制了一款便携式双通道振动测试系统.采用ST-3电磁式传感器拾取设备的振动速度信号,利用AD620仪用放大器对传感器的mV级电压输出进行了放大.采用美国NI公司的USB-6008数据采集卡将放大后的信号传至上位机,同时基于LabVIEW平台编写了数字低通滤波,频谱分析和相关分析软件.实际测试表明,该系统可实现较高采样频率的采样,对机械振动信号的分析处理更为高效,同时系统具有体积小,操作简单,人机交互友好等优点.     

Photonic crystals and microdisk cavities based on GaInAsP-InPsystem

This paper presents a preliminary guide to realize microcavity semiconductor lasers exhibiting spontaneous emission control effects. It includes: 1) theoretical consideration on the effects; 2) processing techniques for semiconductor microcavities; and 3) some demonstrations of photonic crystal and microdisk cavity. It was shown that, even with a spectral broadening of electron transition, thresholdless lasing operation and alternation of spontaneous emission rate are expected in a cavity satisfying the single mode condition that only one mode is allowed in the transition spectrum. An ideal three-dimensional (3-D) photonic crystal has the potentiality for realizing this condition. In two-dimensional (2-D) crystals and microdisk cavities, thresholdless operation is also expected, but the alternation of spontaneous emission rate may be negligible due to the insufficient optical confinement. In the experiment, some processing techniques for GaInAsP-InP system were investigated and methane-based reactive ion beam etching was selected because of the smooth sidewalls and adaptability to arbitrary structures. A GaInAsP-InP 2-D photonic crystal constructed by submicron columns was fabricated using this method. Owing to the slow surface recombination of this material, a polarized photoluminescence and peculiar transmission spectra were observed at room temperature (RT), which can be explained by a photonic band calculation. However, some technical improvement is necessary for clear demonstration of photonic bandgap, which is minimally required for device applications. In contrast to this, a GaInAsP-InP microdisk cavity of 2 μm in diameter, which corresponds to the cavity volume 2.5 times the single-mode condition, has achieved RT lasing with threshold current as low as 0.2 mA. Further reduction of diameter and realization of continuous-wave (CW) operation will provide a significant regime for the observation of spontaneous emission control effects     

Modeling and simulation of temperature drift for ISFET-based pH sensor and its compensation through machine learning techniques

The paper presents modeling and simulation of ion-sensitive field-effect transistor (ISFET)-based pH sensor with temperature-dependent behavioral macromodel and proposes to compensate the temperature drift in the sensor using intelligent machine learning (ML) models. The macromodel is built using SPICE by introducing electrochemical parameters in a metal-oxide-semiconductor field-effect transistor (MOSFET) model to simulate ISFET characteristics. We account for the temperature dependence of electrochemical and semiconductor parameters in our macromodel to increase its robustness. The macromodel is then exported as a subcircuit element, which is used to design the readout interface circuit. A simple constant-voltage, constant-current (CVCC) topology is utilized to generate the data for temperature drift in ISFET pH sensor, which is used to train and test state-of-the-art ML-based regression models in order to compensate the drift behavior. The experimental results demonstrate that the random forest (RF) technique achieves the best performance with very high correlation and low error rate. Corresponding curves for output signal using the trained models show highly temperature-independent characteristics when tested for pH 2, 4, 7, 10, and 12, and we obtained a root mean squared error (RMS) variation of ΔpH ≤ 0.024 over a temperature range of 15°C to 55°C in comparison with ΔpH ≤ 1.346 for uncompensated output signal. This work establishes the framework for integration of ML techniques for drift compensation of ISFET chemical sensor to improve its performance.     

Stigmergic electronic gates and networks

Stigmergy is a communication method based on changing the surrounding environment according to reference feedbacks. It is typical within animal colonies that are able to process even complex information by releasing signals into the environment, which are subsequently received and processed by other elements of the colony. For example, ants searching for food leave traces of a pheromone, like Hansel and Gretel’s breadcrumbs, along the way. When food is found, they return to the anthill reinforcing this pheromone trace as a signal and reminder to all the others. Similar techniques are used in routing software even if stigmergic hardware might be even more efficient, fast, and energy saving. Recently, a stigmergic photonic gate based on soliton waveguides has been proposed; this particular stigmergic hardware can switch the output ratio of the channels as a result of optical feedback. Based on these results, in this study, we analyze stigmergic electronic gates that can be addressed through external feedback, as the photonic ones do. We show that the nonlinear response of such gates must be based on quadratic saturating conductances driven by feedback signals. For this purpose, networks of stigmergic gates require two parallel and communicating current circuits: one to transmit information, and another for feedback signals to control the gate switching. We also show that by increasing the number of terminals per single gate, from 2 × 2 to 3 × 3 or higher, the overall power consumption can be reduced by a few orders of magnitude.

     

基于集群负荷预测的主动配电网多目标优化调度

常规的配电网调度模式中,往往通过可控分布式电源、储能和柔性负荷来调节预测误差和实时波动,粗略地预测负荷值,这使得负荷预测往往不够精准,而且用可控分布式电源、柔性负荷或储能平衡配电网负荷波动,会造成较大的波动成本和备用成本。对此提出一种基于集群负荷预测的主动配电网多目标优化调度方法。采用模糊聚类的方法,对负荷进行集群划分,利用极限学习机对负荷进行集群预测。基于预测值,先以有功调度成本最低进行日前调度,再在日前调度的基础上进行修正,以可控分布式出力修正量最小、储能出力修正量最小、柔性负荷修正量最小为目标进行实时调度。     

基于协同训练与LS-SVM的集成传感器在线温度补偿

为利用机器学习对集成传感器实现在线补偿,使算法具有标定未知样本和更新样本集的能力,利用协同训练的方式,对最小二乘支持向量回归机进行改进,提出基于协同训练的支持向量回归算法,使用临近法对未知样本进行标定和选择,同时对新的样本空间进行剪枝,在保证反映新样本特性的前提下尽量减少对学习模型影响小的样本数量。实验证明,该算法在泛化能力不下降的情况下提高了回归精度,运用在集成传感器的在线补偿上,能降低获的成本,并取得良好的补偿效果。     

基于机器学习的短期电力负荷预测方法研究

针对短期电力负荷数据具有明显周期性的特点,将基于机器学习引入到短期电力负荷预测领域,提出一种基于岭回归估计的RBF神经网络短期电力负荷预测方法,该方法利用机器学习算法RBF在非线性拟合方面的优势,结合岭回归对RBF神经网络输出层权值进行参数估计,有效消除输入多重共线性问题,采用广义交叉验证法对构建的模型进行评估,寻找最优岭参数,提高了电力负荷预测精度。通过实际负荷预测案例,与传统BP神经网络负荷预测方法进行比对,验证了提出的电力负荷预测方法较传统方法具有较好的稳定性和较高的预测精度,为电力负荷预测提供了新思路。     

基于支持向量机的多回路关口电能表在线监测方法研究

根据多回路关口电能表监测的多通道采样连续性,采样信号的多特征及其耦合性,提出一种基于支持向量机的多回路关口电能表在线监测方法,该方法分别设计了基于监测系统的信号采集单元、数据通信单元、数据处理单元及上位机显示单元,搭建了电能表的多个故障缺陷模型,通过信号采集单元将缺陷模型的三相电流及电压实时数据传送给数据处理单元,数据处理单元对收集的数据进行基础运算后形成对应的有功功率样本数据,并对其进行小波包能量谱离线分析后形成故障特征池,构建了基于支持向量机算法的多分类支持向量机模型并将得到的模型内嵌于上位机显示单元,监测系统运行时上位机显示单元直接基于模型的关联规则,与转换后的采样信息进行特征匹配,实现对电能表运行工况的实时监测。经实例验证结果表明该方法能够实时、准确的对多回路关口电能表进行在线监测,具有较强的应用价值及前景。     

高比例新能源电力系统静态电压稳定裕度在线概率评估

新能源的随机性、波动性及弱调节特性给电力系统静态电压的安全及稳定性带来了挑战。针对此问题，提出一种考虑源荷双侧不确定性的高比例新能源电力系统静态电压稳定裕度在线概率评估方法。首先，基于新能源无功调节特性与传统机组的差异，分析了大量新能源替代传统机组对稳定裕度的影响。然后，分析了新能源出力不确定性对稳定裕度分布范围的影响，并建立源荷不确定性模型以生成典型场景。最后，为了应对新能源快速波动性给稳定裕度带来的影响，提出基于优化ELM-KDE的稳定裕度在线概率评估方法。利用优化极限学习机(extreme learning machine, ELM)预测典型场景稳定裕度并通过核密度估计(kernel density estimation, KDE)准确获得其概率分布函数。构建了静态电压稳定期望裕度和静态电压稳定风险度两个指标对结果进行表征。分别在New England 39和IEEE300节点系统进行了仿真测试，并将结果与传统蒙特卡洛方法计算结果对比，验证了所提方法的有效性。     

采用梁型载荷传感器改造旧式万能试验机

采用弹性梁对旧式液压万能材料试验机下横梁进行改造,为确保弹性梁传感器的灵敏度和精度,根据应用"应力集中"的设计原则,在合理选择传感器结构基础上,采用有限元法对其应变和强度进行计算,合理确定弹性梁的贴片部位,从理论上确定传感器的最大应变方位及传感器的量程。在旧式液压万能试验机上实际应用表明:被测力与输出信号保持严格的对应关系,传感器完全满足测试强度与灵敏度的要求,可实现载荷的电量输出。     

基于支持向量回归的配电网线损计算模型

提出了一种基于支持向量回归的计算配电网线损的可行方法,建立了配电网线损计算的支持向量回归模型。针对有代表性的配电线路的线损与特征参数的样本数据,利用支持向量回归的拟合特性映射线损与特征参数之间复杂的非线性关系,找出配电线路的线损随特征参数变化的规律。为了提高支持向量回归机的学习效率,采用样本分类处理的方法分别对其进行训练,使的计算结果更加符合实际。以配电线路数据为实例,仿真结果验证了所提的方法和模型的有效性和实用性。     

基于聚类分析的配电台区拓扑识别方法

拓扑识别是配电台区的技术热点之一,拓扑关系是电网普遍需求.在不额外增加拓扑识别硬件的条件下,利用台区同期电能数据进行拓扑识别,是有别于专用拓扑装置的另一种方法.研究了基于基尔霍夫定律的智能装置父子关系的特征条件和数学组合算法,并研究了基于聚类分析的拓扑识别算法,实现了从台区总出线开关到用户电能表的拓扑识别过程.提出了智...     

A novel fuzzy state feedback controller for power system stabilization

A novel fuzzy logic controller (FLC) is developed for damping electromechanical modes of oscillations and enhancing power system synchronous stability. The proposed controller is based on state feedback control system. The input signals to the controller are the weighted sum of the mechanical states and the electrical states weighted sum, while its output signal is added to the conventional fixed-gain PI controller output signal to give the excitation control signal. The simulation results using the proposed controller are carried out on an synchronous machine infinite bus system. Moreover, a comparison between the conventional fixed-gain PI controller and the proposed FLC is presented. The results validate the effectiveness of the proposed FLC in terms of less overshoot/undershoot and enhancing the power system stability over a wide range variation of operating conditions.     

基于单片机的多点温度监控系统设计

设计了基于单片机的多点温度监控系统,旨在解决传统的单点线式控制的不足。系统包括主机和3个从机模块,从机使用数字温度传感器DS18B20对温度进行检测,将温度信号传输给单片机,单片机对温度信号进行处理,处理后的数据送给无线模块nRF24L01,然后将数据发送给主机的nRF24L01模块,主机的nRF24L01接收到温度信号后,送给主机单片机进行处理,控制LCD1602进行温度显示,当温度超过上下限值时,单片机会启动外围电路的蜂鸣器进行报警。实际测试结果表明,该系统所测量的温度误差小、系统性能良好、稳定可靠,具有较好的推广价值。     

基于极端天气情况的风-储系统平抑控制策略

提出了一种储能系统的功率控制方法,实现了极端天气情况下风电场出力波动的快速平抑。该控制框架融合了机器学习算法与模型预测控制方法,由基于在线序贯极限学习机的神经网络模型预测优化时域范围的风电功率,储能的充放电功率指令通过MPC进行滚动优化,保证储能系统的运行约束得到满足。仿真实验表明该方法能够实现储能系统的快速充、放电管理,利用准确的风电功率预测,降低了极端天气下风电场功率陡降对电网的不利影响,使得风-储联合系统注入电网的功率更接近给定值。     

最小二乘支持向量机预测绝缘子等值附盐密度

考虑到气象因子条件对绝缘子的等值附盐密度影响复杂,难以建立精确数学模型等问题,提出了一种最小二乘支持向量机的绝缘子在一定的气象因子条件下的等值附盐密度预测新模型。以温度、湿度、风速等主要气象因子为输入,绝缘子等值附盐密度为输出,通过最小二乘支持向量机模型,拟合输入与输出之间的复杂非线性函数关系。以现场采集的气候数据为样本对模型进行学习训练,用训练好模型预测绝缘子在一定气候条件下的等值附盐密度。实践表明该方法具有建模速度快、预测精度高、操作简便等优点,不仅克服了常规的BP预测模型的不足,而且性能优于标准支持向量机预测模型。     

基于虚拟直流机的直流微电网电压稳定控制策略

针对大量恒功率负荷接入直流微电网致使直流微电网失稳的问题,提出了一种基于虚拟直流机(VDCM)的直流微电网电压稳定控制策略。控制策略以储能双向DC/DC变流器为研究对象,基于直流电机原理,以电感电流为反馈量在传统下垂控制的基础上引入VDCM环节,增强系统阻尼,降低恒功率负荷对系统稳定性的影响。通过建立所提控制策略下的直流微电网小信号模型,利用阻抗匹配原则分析相关参数变化时系统的稳定性,并将其与传统的VDCM控制策略进行对比。最后,搭建仿真模型和硬件实验平台,验证所提控制策略的有效性。结果表明：所提控制策略使变流器具备了直流电机的惯量和阻尼特性,在提升系统稳定性的同时,也在一定程度上改善了系统动态响应性能,且其控制效果优于传统的VDCM控制策略。