Research On Ground Heat Balance In Ground Source Heat Pump Systems Based On Wavelet Theory

Ground source heat pump (GSHP), which is high efficiency, hasenormous development potentials. It extracts heat from soil underground forheating buildings in winter, and rejects heat to soil underground for coolingbuildings in summer. Therefore, the temperature of soil undergroundinfluences the performance of the GSHP system directly. As the diversity ofbuilding function and the complexity of soil heat transfer, it is hard to predictthe temperature variations of soil underground accurately. In this paper, amathematical method based on the theory of wavelet and auto-regression wasused to analyze and forecast the building load and underground soiltemperature in the GSHP system. The main works are summarized asfollows:(1) The methods used in the fields of financial economics analysis,shock load forecasting in electric network and rainfall prediction werereferenced. First of all, complex time sequences were decomposed andreconstructed at single scale by wavelet transformation respectively. Then, different auto-regressive moving average (ARIMA) models were built toforecast the reconstruction data sequences of different scales. Next, theforecasting results of each scale were accumulated to get the final forecastingresults of the raw time sequences. Finally, the errors were analyzed to testifythe accuracy of mathematical model.(2) Building load sequences and soil temperature sequences in theGSHP system were analyzed and forecasted in different time scales using themathematical model mentioned above. The research showed, building loadsequences had best annual periodicity, better daily periodicity, good weeklyperiodicity and bad monthly periodicity; while the soil temperature sequenceshad satisfactory annual periodicity and which was the shortest cycle.Consequently, it is better to analyze and forecast the building load sequencesand soil temperature sequences in time scales measured in year or day,neither week nor month.(3) Building load sequences at different levels had different influence onsoil temperature sequences. The smoothed low frequency sequences ofbuilding load and soil temperature, which were obtained by waveletdecomposition, were regression analyzed to obtain the regression equations ateach level. Comparison of different regression equations showed that the firstderivative of soil temperature low frequency sequences and building load low frequency sequences had linear relationship under the conditions of fixedGSHP units; and the regression equation coefficients were related to GSHPunits parameters, the running mode of GSHP, the building load and so on.(4) The mathematical forecasting model based on the theory of waveletand auto-regression can be used to guide the practical GSHP system designand optimization qualitatively.(5) Compared with traditional soil heat transfer numerical calculationand simulation, this mathematical model is easy computation and highlyaccurate in certain cases. However, it depends on a large number of test data,and the universality of forecasting results hinges on the number of sampledata used in the statistical analysis.