توسعه یک شبکه عصبی RDP برای عیب یابی و تشخیص مصرف انرژی ساختمان
|کد مقاله||سال انتشار||تعداد صفحات مقاله انگلیسی||ترجمه فارسی|
|6363||2013||6 صفحه PDF||سفارش دهید|
نسخه انگلیسی مقاله همین الان قابل دانلود است.
هزینه ترجمه مقاله بر اساس تعداد کلمات مقاله انگلیسی محاسبه می شود.
این مقاله تقریباً شامل 4390 کلمه می باشد.
هزینه ترجمه مقاله توسط مترجمان با تجربه، طبق جدول زیر محاسبه می شود:
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Energy and Buildings, Volume 62, July 2013, Pages 133–138
Fault detection and diagnosis (FDD) is an important issue in building energy conservation. This paper proposes a new option for solving this problem at the building level by using a recursive deterministic perceptron (RDP) neural network. Results show a higher than 97% level of generalization in all the designed experiments. Based on this high detection ability of RDP model, a new diagnostic architecture is proposed. Our experiments demonstrate that it is able to not only report correct source of faults but also sort sources in the order of degradation likelihood.
Buildings play an important role in the total energy use and environmental implications in the world. In Europe, it accounts for 40% of the total energy used and 36% of the total CO2 emissions . The energy usage of buildings is not always done under normal conditions. For example, in order to keep a room temperature at a comfortable level, more than the normally required heating energy is used. Another example might be one on which normal energy is consumed but room temperature has not reached its expected value. Together with heating energy, electricity and cooling energy consumption may also experience abnormality. We call these abnormal energy consumptions faults. They might be caused by performance degradation, poor maintenance or improper operation of the installed electrical systems. The early detection of these faults in energy usage is crucial for building operation and energy conservation. Based on this information, building users and engineers are able to adjust their corresponding equipment for enhancing energy usage as well as saving energy. This type of non-abrupt faults are difficult to monitor and even more difficult to diagnose. This paper therefore introduces an effective artificial neural network (ANN) model, based on the recursive deterministic perceptron (RDP) neural network, to implement fault detection and diagnosis (FDD) of building energy consumption. Based on the knowledge from previous faulty consumption, this model is able to report faults automatically and with a high degree of accuracy. It also shows high performance in a newly designed fault diagnosis procedure. This paper is organized as follows. Section 2 introduces the recent work related to FDD of building energy consumption. Section 3 describes RDP neural network model. Section 4 presents how to do fault detection with RDP and reports the experimental results. Section 5 proposes a new diagnosis approach involving RDP neural networks. The last section concludes this paper.
نتیجه گیری انگلیسی
An RDP neural network model is used to detect and diagnose faults in the entire building level systems. Four equipments under normal and abnormal conditions are simulated to evaluate this model. Experiments demonstrate that this model is highly accurate in detecting all possible faults, including one equipment degradation and all four equipments degradation. More specifically, on training set, the accuracy remains 100% and on testing set it achieves higher than 97%. Based on the high detection ability of RDP, we proposed a new fault diagnostic architecture. Other than correctly reporting the source of faults, it can list the equipments in order, according to their possibilities of experiencing malfunction. This attribute allows us to diagnose faults caused by more than one source. We are even able to deal with the problem in which several equipments cause similar faults. In future work, we will evaluate this model on real consumption data, test other classification models and compare their performance. In addition, we consider parallelizing this model to achieve higher performance.