Development of a Residential Space-Heating Demand Prediction Method under Uncertainly

Dianzheng Fu, Zeyu Zheng, Yiming Tong, Yang Fu2015 · DOI

Abstract

In this paper, a prediction method was advanced for residential space heating demand under the fuzzy environment. This developed method improved the traditional heat load duration curve method via the quantification expression technique for the fuzzy heating durations. The results indicate that the higher the system possibility is, the narrower the interval of the heat demand prediction result becomes, in which the decision maker can adjust the practical heat provisions of the heat sources. Besides, the opposite heat provision variation trend of the main heat source and peak-shaving heat source can be obtained with the rising thermalization coefficient. Compared with the traditional heat load duration curve method, it is helpful for the relative multi-heat-source centralized heating system managers by providing a potential tradeoff analysis between the system possibility and the selectable heat provision ranges of different heat sources.

Introduction

Predicting residential space-heating demand is a critical yet challenging task for urban utility management. Accurate predictions are essential to balance fuel efficiency and greenhouse gas emissions against the necessity of ensuring resident comfort. Traditional predictive models, including modern machine learning approaches, often face significant hurdles: they require extensive historical meteorological and building data that may not be available in many regions, and they frequently treat heating duration as a fixed parameter. However, climate change and seasonal variability introduce significant uncertainty, making the heating season's length "fuzzy." This research proposes a novel method that thrives on this uncertainty by integrating fuzzy set theory with the traditional Heat Load Duration Curve (HLDC) approach.

Methodology

The core of the proposed method is the enhancement of the HLDC method through Fuzzy Set Theory to account for the unpredictable nature of total heating duration ($N_{zh}$). The process involves three primary stages:

  1. Fuzzy Assessment: Instead of a single deterministic value, the heating season length is defined as a "trapezoidal fuzzy number" based on expert judgment (e.g., a range of 140 to 165 days).
  2. $\alpha$-cut Integration: The model employs "$\alpha$-cuts" to slice fuzzy data at various confidence levels (ranging from 0 to 1). This technique converts fuzzy uncertainty into quantifiable intervals—upper and lower bounds—of heat demand.
  3. Multi-Source Allocation: The demand is allocated between a Main Heat Source (MHS), such as a combined heat and power plant, and Peak-Shaving Heat Sources (PHS), like backup boilers. This allocation is governed by a "thermalization coefficient" ($\beta$).
Conceptual framework of the fuzzy heat load duration curve method.

By utilizing linearity approximations, the method avoids overly complex fuzzy algorithms, making it a practical tool for engineering applications where computational resources or specialized expertise may be limited.

Case Study: Shenbei New District

The method was validated using a case study of the Shenbei New District in Shenyang, China, covering a heating area of 5.97 million square meters. The region has an average outdoor temperature of -5.7°C and a design temperature of -19°C. Expert input defined the fuzzy heating duration as $N_{zh} = (140, 165, 150, 155)$ days. The system includes a mix of ordinary and energy-conserving buildings with heat load indices of 64 and 45 $W/m^2$, respectively.

Key Results

The application of the fuzzy HLDC method yielded interval-based predictions that offer more flexibility than traditional deterministic models. A primary finding is the "possibility tradeoff": as the system possibility ($\alpha$-cut) increases, the prediction interval narrows. This provides managers with a mathematical visualization of risk; higher confidence levels result in tighter demand ranges, while lower confidence levels highlight a broader range of potential "edge case" demand.

Heat demand predictions for main and peak-shaving sources across different confidence levels and thermalization coefficients.

For instance, at a moderate confidence level ($\alpha = 0.5$) and a standard thermalization coefficient ($\beta = 0.55$), the predicted demand for the MHS was an interval of [2.20, 2.43] PJ. Increasing the confidence level from 0.2 to 0.8 reduced the spread of the MHS demand from 0.32 PJ to 0.13 PJ. Furthermore, the results demonstrated an inverse relationship: as $\beta$ increases, the demand on the MHS rises while the demand on PHS decreases.

Conclusion

The developed method provides a robust framework for predicting heat demand in data-scarce and uncertain environments. By offering "selectable heat provision ranges," it empowers system managers to make informed tradeoffs between system possibility and operational flexibility. While the model currently relies on expert-defined fuzzy numbers and assumes static building data, it serves as a valuable foundation for future real-time energy management systems that could integrate IoT sensor data to further refine these fuzzy boundaries.

Summary updated

Abstract

in this paper, a prediction method was advanced for residential space heating demand under the fuzz environment. This developed method improved the traditional heat load duration curve method via the quantification expression technique for the fuzzy heating durations. The results indicate that the higher the system possibility is, the narrower the interval of the heat demand prediction result becomes, in which the decision maker can adjust the practical heat provisions of the heat sources. Besides, the opposite heat provision variation trend of the main heat source and peak-shaving heat source can be obtained with the rising thermalization coefficient. Compared with the traditional heat load duration curve method, it is helpful for the relative multi-heat-source centralized heating system managers by providing a potential tradeoff analysis between the system possibility and the selectable heat provision ranges of different heat sources.
BibTeX
@misc{fu2015development,
  author = {Dianzheng Fu and Zeyu Zheng and Yiming Tong and Yang Fu},
  title = {{Development of a Residential Space-Heating Demand Prediction Method under Uncertainly}},
  year = {2015},
  doi = {10.2991/gmee-15.2015.45},
}
Figure