Using the newly available TOPEX Microwave Radiometer (TMR) data spanning 1993 through 2002. a 10 yr climatology of oceanic water vapor [OWVI is constructed, of which the distribution and variation at various spatial-temporal scales we investigated The new dataaet confirms most of the well-known OWV features, and yields a number of interesting findings, due to its high quality, long duration, and unique orbit. I) The TMR-derived climatology compares well in both overall pattern and general statistics, with similar results based on radiosondes and other satellites. Climatological comparisons with sea surface temperature and oceanic precip itation suggest thai the western Pacific warm pool is "mirrored" in the atmosphere as a "'wet pool," whereas the meteorological equator is refleclcd in OWV as a Iransocean equatorial wet belt. 21 It is lound that El Nifio (La Nifia) events are accompanied by a significant increase (decease) in the amount of OWV between 10s and 10N with a somewhat unexpected Southern Hemisphere dominance. This is parieularly evidem during the 1997198 El Nino when the interannual variability of OWV reaches a record high. Composite maps of annual OWV anomalies disclose a dipnlelike pattern in the western equatorial Pacific with a phase opposition between El Nino and La Nina years. 3) The annual amplitude of OWV is characterized by six cross-continent wet belts located largely in the subtropics of both hemispheres. The phase patterns of the annual and semiannual variations are hemispherically divided, and climatologically co.elated, respectively. North (south) of the intertropical convergence zone (ITCZ), a majorily of the oceanic areas have their water vapor maximum in August (February). Early peaks in July are found over a few continental shelf regions of the Northern t-lemisphere (NH), while late peaks in March are fouiid in the tropical oceans of the Southern Hemisphere (SHh Moreover. two delayed maximums in September are visible in the interior North Pacific and North Adandtic, respectively. 4) The daily cycle of OWV is strongly eoupled with its seasonal cycle, and is therefore unstable in nature But a double peak streture with a general hemispheric phase reversal can still be identified. 5) The ratio of the NH versus SH OWV is roughly 1.17:1, and tile relative importance of the interannuah annual, semiannuah diurnal, and semidlurnal variations in terms of mean amplitude is approximately I.fi:5:1.2:1:1. [n view of these encouraging results, further exploration of present and future "attlmele-borne" radiomeler data will no, doubt lead to an improved and complementary understanding of the OWV system in many aspecls.

本文叙述了面向对象技术构造电力变压器绝缘诊断知识库的原理。采用这种面向对象技术使得整个专家知识呈分布式、层次式特征，从而克服了传统产生式专家系统的固有局限。本文还用“黑板结构”模式改变了基本结构专家系统的构造并用计算机程序实现，取得了较为满意的效果。

This paper describes the theory of on-line detection of six kinds of gases-H2, CO, CH4, C2H4, C2H2, C2H6-which are dissolved in transformer oil, and analyzes the characterististics of the microcomputer detection system. The back propagation neural network (BPNN) is applied to diagnosing the transformer faults, and the results of verification show that the BPNN diagnosis method is effective

变压器油中溶解气体在线监测对变压器绝缘状态分析具有积极的指导作用，在线监测用气体传感器是实施该技术 的关键。本文针对半导体气体传感器的交叉敏感特性，提出了将气体传感器阵列与人工神经网络技术相结合，形成一种智能传感器，用于单一气体的定性识别和定量检测；利用六个半导体气体传感器组成传感器阵列，采用BP神经网络进行模式识别。通过大量的试验证明了本文提出的智能传感器可有效提高H2、CO、CH4、C2H4、C2H2、C2H6六种气体的分辨率和检测灵敏度。