CN103077297B - A kind of Short-time-interval atmosphere ambient temperature prediction method - Google Patents

Abstract

The invention discloses a kind of Short-time-interval atmosphere ambient temperature prediction method, first, in atmospheric environment, define a sectional area is that the spherical control volume of 1 square metre is as research object, then every setting-up time step-length, the parameter such as solar radiant energy, atmospheric temperature of Survey control body, be greater than 12 constantly little when the sample record duration, start to perform atmospheric temperature predictor, and often perform once through a prediction duration time interval; In commission, for a kth predicted time section, according to parameters such as the atmospheric temperatures that present sample obtains, calculate and determine C _thvalue, if k is more than or equal to 2, upgrades air radiation heat absorption coefficients f value; Finally, utilize solar radiant energy prediction module to predict solar radiant energy distribution in a following prediction duration, obtain Temperature Distribution in prediction duration, program is waited for until the next prediction moment afterwards; The inventive method is applicable to the situation of any place and any weather pattern, and precision is high, and calculated amount is little, can perform in real time online in practical application.

Description

A kind of Short-time-interval atmosphere ambient temperature prediction method

Technical field

The invention belongs to heating ventilation and air conditioner system energy saving technical field, be specifically related to a kind of Short-time-interval atmosphere ambient temperature prediction method.

Background technology

Heating ventilation and air-conditioning system (HVAC) are one of main sources of building energy consumption, are one of major domains of current energy-saving building technology exploitation.The operation of current great majority building HVAC system controls only to consider operating mode at full capacity, and when changing for run duration thermal load, in most cases the controling parameters of system not runs in optimal conditions.HVAC energy efficiency management is by calculating thermal load in advance and predicting, and the controlling value of HVAC system operational factor (as refrigeration (heat) amount, flow, temperature etc.) is optimized, meeting under workload demand prerequisite, ensure that HVAC system operates in least energy consumption state, thus under making the controling parameters of HVAC system remain on optimum condition of work in operational process, reach energy-conservation object always.Atmospheric temperature is one of key factor affecting HVAC thermal load, therefore needs in HVAC energy efficiency management process to predict outside atmosphere environment temperature.The optimization frequency of HVAC energy efficiency management is generally 10-15 minute, and high precision short time interval (sub-hourly) atmospheric temperature forecasting techniques is one of core technology of exploitation HVAC energy efficiency management.

At present about atmospheric temperature Forecasting Methodology mainly contains early stage parameterized analytical models [1-4] and neural network model [5-10] conventional in recent years, but these distinct methods are all aimed at application-specific, in precision and applicability, significant limitation is also existed for the prediction of short time interval environment temperature, the parameterized analytical models proposed as document [1-3] is aimed at agricultural greenhouse application, for predicting near surface air themperature, analyze the heat exchange relationship between air and surface soil and impact.Affect atmospheric temperature change because have [4] such as solar radiant energy, local landform, atmospheric humidity, cloud cover situation and wind speed, the change of these factors has random nature, and therefore parameterized analytical models is difficult to the temperature prediction being applied to short time interval.And artificial neural network (ANN) is suitable for the feature identification to stochastic variable, classification and prediction, and in atmospheric temperature prediction, receive extensive research [5-10], but ANN method has certain limitation for short time interval Forecasting of Atmospheric Environment, it is based on the training to historical data that ANN model is set up, need in forecasting process as humidity, the weather conditions such as wind speed and solar radiant energy are as input parameter [7], the ANN forecast model set up is relevant to locality, do not possess versatility feature, and the training time of setting up forecast model needs is long, need a large amount of historical data, climate change is differed and obtains effective ANN forecast model surely in place frequently.

List of references:

[1]J.W.Deardorff，Efficientpredictionofgroundsurfacetemperatureandmoisture，withinclusionofalayerofvegetation，JournalofGeophysicalResearch，83(1978)：1889-1903.

[2]C.M.Bhumralkar，Numericalexperimentsonthecomputationofgroundsurfacetemperatureinanatmosphericgeneralcirculationmodel，J.Appl.Meteorol.，14(1975)：1246-1258.

[3]A.K.Blackada，Modelingthenocturnalboundarylayer，ProceedingsoftheThirdSymposiumonAtmosphericTurbulence，DiffusionandAirQuality,pp.46-49，AmericanMeteorologicalSociety，Boston，Mass.，1976.

[4]H.Swaid，M.E.Hoffman，PredictionofurbanairtemperaturevariationsusingtheanalyticalCTTCmodel，EnergyandBuilding，14(1990)：313-224.

[5]L.Bodri，V.Cermak，Predictionofsurfaceairtemperaturesbyneuralnetwork，examplebasedonthree-yeartemperaturemonitoringatSporilovstation，Stud.Geophys.Geod.，47(2003)：173-184.

[6]A.Jain，RW.McClendon，G.Hoogenboom，Freezepredictionforspecificlocationsusingartificialneuralnetworks，TransactionsoftheASABE，49(6)：1955-1962.

[7]B.A.Smith，R.W.McClendon，G.Hoogenboom，Improvingairtemperaturepredictionwithartificialneuralnetworks，Int.J.ComputationalIntelligence，3(2006)：179-186.

[8]R.F.Chevalier，G.Hoogenboom，R.W.McClendon，J.A.Paz，Supportvectorregressionwithreducedtrainingsetsforairtemperatureprediction：acomparisonwithartificialneuralnetworks，NeuralComput.&Applic.，20(2011)：151-159.

[9]A.L.Labajo，J.L.Labajo，Analysisoftemporalbehaviorofclimatevariablesusingartificialneuralnetworks：anapplicationtomeanmonthlymaximumtemperaturesontheSpanishCentralPlateau，Atmosfera，24(2011)：267-285.

[10]M.Afzali，A.Afzali，G.Zahedi，Thepotentialofartificialneuralnetworktechniqueindailyandmonthlyambientairtemperatureprediction，Int.J.EnvironmentalScienceandDevelopment，3(2012)：33-38.

[11]G.A.F.Seber，C.J.Wild，NonlinearRegression，JohnWiley＆Sons，Inc.，1989，p.254

Summary of the invention

For solving above-mentioned problems of the prior art, the object of the present invention is to provide a kind of Short-time-interval atmosphere ambient temperature prediction method, the inventive method is applicable to the situation of any place and any weather pattern, and precision is high, calculated amount is little, can perform in real time online in practical application.

The design philosophy of the inventive method is:

First, define in atmospheric environment sectional area be the spherical control volume of 1 square metre as research object, namely the change of this control volume temperature represents atmospheric temperature change.The heat that control volume temperature variation depends on heat capacity of air and receives, wherein the change of heat capacity of air is relevant with temperature with air humidity, exchange heat outside control volume and control volume realizes mainly through radiation and convection current, consider that predicted time interval is shorter, can suppose that control volume and surrounding air temperature in prediction period is identical, therefore can ignore control volume and extraneous convection heat transfer.Radiation heat has multiple ingredient, comprises the direct radiation heat coming from (1) sun; (2) cloud layer and ground return; (3) cloud layer and terrestrial radiation heat.Wherein (1) and (2) directly depends on local solar radiant energy, and (3) and cloud layer, surface temperature are relevant, and have indirect dependence with solar radiant energy, its change has a lag-effect in time with solar radiant energy.Therefore the change of atmospheric temperature can be summed up as only relevant with local solar radiant energy, and key content of the present invention proposes a kind of new method to determine the dependence between atmospheric temperature and solar radiant energy.

For achieving the above object, the technical solution adopted in the present invention is:

A kind of Short-time-interval atmosphere ambient temperature prediction method, comprises the steps:

Step 1: first, define in atmospheric environment sectional area be the spherical control volume of 1 square metre as research object, the sampling time step-length of setup control body environmental variance, setting prediction duration;

Step 2: every the time step of a setting, measure solar radiant energy, atmospheric temperature, pressure and relative humidity, the time span continued when sample record is greater than 12 constantly little, then start to perform atmospheric temperature predictor, and often perform an atmospheric temperature prediction through the time interval of a prediction duration;

Step 3: in atmospheric temperature prediction implementation, for a kth predicted time section, first according to atmospheric temperature, pressure and relative humidity that present sample obtains, calculated by equation (1) ~ (5) and determine C _thvalue, if k equals 1, then sets a value (as 0.0025) for air radiation heat absorption coefficients f in interval [0.001,0.01] scope, if predicted time segment index k is more than or equal to 2, adopts equation (6) to upgrade the value of air radiation heat absorption coefficients f:

$\ln \left(p_s\right)=$

$54.842763-\frac{6763.22}{T}-4.21\ln \left(T\right)+0.000367T+\tanh \left[0.0415(T-218.8)\right][53.878-$

$\frac{1331.22}{T}-9.44523\ln \left(T\right)+0.014025T]---\left(2\right)$

$\mathrm{\ρ}=\frac{p(1-x)}{287.058T}+\frac{\mathrm{px}}{461.495T}---\left(3\right)$

$C_{\mathrm{th}}=\frac{3}{4\mathrm{\π\ρ}{c}_p{r}^3}---\left(5\right)$

$f=\frac{A(T_0^k-T_0^{k-1})}{C_{\mathrm{th}}^k[{\∫}_{t_0^k-12}^{t_0^k}{f}_{\mathrm{bl}}\left(t\right){\stackrel{\·}{q}}_{\mathrm{rad}}\left(t\right)\mathrm{dt}-{\∫}_{t_0^{k-1}-12}^{t_0^{k-1}}{f}_{\mathrm{bl}}\left(t\right){\stackrel{\·}{q}}_{\mathrm{rad}}\left(t\right)\mathrm{dt}]}---\left(6\right)$

In formula: ---air absolute humidity,

X---relative air humidity,

P---atmospheric pressure,

P _s---water vapor in air saturation pressure,

T---control volume temperature,

ρ---control volume atmospheric density,

C _p---control volume air specific heat,

R---control volume radius

F---the photothermal absorption coefficient of control volume,

T ₀---Current Temperatures,

T ₀---current time, unit: hour

F _bi(t)---BoxLucas function, expression formula is: parameter a in formula ₁, a ₂default value can be taken as 0.5 and 3.8, can do certain adjustment for these two parameter values of different regions.

A---BoxLucas function is at the integration of interval [0,12], and expression formula is: $A={\∫}_0^{12}{f}_{\mathrm{bl}}\left(t\right)\mathrm{dt}$

Interval [t ₀-12, t ₀] q _rad---the solar radiant energy in past 12 hours;

Step 4: utilize solar radiant energy prediction module to predict solar radiant energy distribution in a following prediction duration, thus obtain Temperature Distribution in a following prediction duration according to equation (10), program is waited for until the next prediction moment afterwards, specific as follows:

The thermal balance equation of control volume is (7):

$\frac{4}{3}{\mathrm{\π\ρc}}_p{r}^3\frac{\mathrm{dT}}{\mathrm{dt}}=f\stackrel{\·}{q}---\left(7\right)$

Integration is carried out to these equation both sides and obtains equation (8):

（8）

In formula: T is control volume temperature, ρ is atmospheric density, c _pfor air specific heat, r is control volume radius, for built-up radiation hot strength, f is the photothermal absorption coefficient of control volume;

Compare with reflection with direct solar radiation, the heat radiation that cloud layer and earth surface itself produce has certain hysteresis quality in time, consider that cloud layer and the heat radiation of earth surface own are solar radiant energy lag-effects in time, the radiation heat in equation (7) calculating can be regarded as the past period solar radiant energy the accumulation of thermal effect, if thermal effect duration time lag of solar radiation is 12 hours, and feature time lag of solar radiant energy thermal effect follows BoxLucas model [11], and is defined as equation (9):

$f_{\mathrm{decay}}\left(t\right)=\frac{f_{\mathrm{bl}}\left(t\right)}{A}---\left(9\right)$

In formula: $f_{\mathrm{bl}}\left(t\right)=\frac{a_1(e^{-a_{2}t}-e^{-a_{1}t})}{a_1-a_2},$ $A={\∫}_0^{12}{f}_{\mathrm{bl}}\left(t\right)\mathrm{dt}$

T is obtained by equation (5), (8) and (9) _pthe prediction expression (10) of moment atmospheric temperature.

$T_p=T_0+\frac{C_{\mathrm{th}}f}{A}[{\∫}_{t_p-12}^{t_p}{f}_{\mathrm{bl}}\left(t\right){\stackrel{\·}{q}}_{\mathrm{rad}}\left(t\right)\mathrm{dt}-{\∫}_{t_0-12}^{t_0}{f}_{\mathrm{bl}}\left(t\right){\stackrel{\·}{q}}_{\mathrm{rad}}\left(t\right)\mathrm{dt}]---\left(10\right)$

Described prediction duration is less than 1 hour.

Described sampling time step-length is 1 ~ 5 minute.

The inventive method is applicable to the situation of any place and any weather pattern, and precision is high, and calculated amount is little, can perform in real time online in practical application.When predicting that duration is less than 1 hour, predicted value and actual observed value meet well, and wherein when predicting that duration is 30 minutes, absolute error is less than 1K, when prediction duration is that 1 little absolute error is constantly less than 3K.

Accompanying drawing explanation

Fig. 1 is solar radiant energy thermal effect characteristic curve time lag based on BoxLucas model.

Fig. 2 compares with observed reading the predicted value of the atmospheric temperature during-8 days on the 1st April in 2012 of Singapore.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is described in further detail.

The present embodiment, by predicting the atmospheric temperature during-8 days on the 1st April in 2012 of Singapore, illustrates a kind of Short-time-interval atmosphere ambient temperature prediction method of the present invention, comprises the steps:

Step 1: first, define in atmospheric environment sectional area be the spherical control volume of 1 square metre as research object, the sampling time step-length of setup control body environmental variance is 3 minutes;

Step 2: every the time step of a setting, the solar radiant energy of Survey control body, atmospheric temperature, pressure and relative humidity, the time span continued when sample record is greater than 12 constantly little, then start to perform atmospheric temperature predictor, and often perform an atmospheric temperature prediction through the time interval of a prediction duration;

Step 3: in atmospheric temperature prediction implementation, for a kth predicted time section, first according to atmospheric temperature, pressure, relative humidity, atmospheric density and air specific heat that present sample obtains, calculated by equation (1) ~ (5) and determine C _thvalue, when k equals 1, if f value is 0.0025, if predicted time segment index k is more than or equal to 2, adopts equation (6) to upgrade the value of air radiation heat absorption coefficients f;

$\ln \left(p_s\right)=$

$54.842763-\frac{6763.22}{T}-4.21\ln \left(T\right)+0.000367T+\tanh \left[0.0415(T-218.8)\right][53.878-$

$\frac{1331.22}{T}-9.44523\ln \left(T\right)+0.014025T]---\left(2\right)$

$\mathrm{\ρ}=\frac{p(1-x)}{287.058T}+\frac{\mathrm{px}}{461.495T}---\left(3\right)$

$C_{\mathrm{th}}=\frac{3}{4\mathrm{\π\ρ}{c}_p{r}^3}---\left(5\right)$

$f=\frac{A(T_0^k-T_0^{k-1})}{C_{\mathrm{th}}^k[{\∫}_{t_0^k-12}^{t_0^k}{f}_{\mathrm{bl}}\left(t\right){\stackrel{\·}{q}}_{\mathrm{rad}}\left(t\right)\mathrm{dt}-{\∫}_{t_0^{k-1}-12}^{t_0^{k-1}}{f}_{\mathrm{bl}}\left(t\right){\stackrel{\·}{q}}_{\mathrm{rad}}\left(t\right)\mathrm{dt}]}---\left(6\right)$

In formula: ---air absolute humidity,

X---relative air humidity,

P---atmospheric pressure,

P _s---water vapor in air saturation pressure,

T---control volume temperature,

ρ---control volume atmospheric density,

C _p---control volume air specific heat,

R---control volume radius

F---the photothermal absorption coefficient of control volume,

T ₀---Current Temperatures,

T ₀---current time, unit: hour

F _bi(t)---BoxLucas function, expression formula is: parameter a in formula ₁, a ₂default value can be taken as 0.5 and 3.8, can do certain adjustment for these two parameter values of different regions.

A---BoxLucas function is at the integration of interval [0,12], and expression formula is: $A={\∫}_0^{12}{f}_{\mathrm{bl}}\left(t\right)\mathrm{dt}$

Interval [t ₀-12, t ₀] q _rad---the solar radiant energy in past 12 hours;

Step 4: utilize solar radiant energy prediction module to predict solar radiant energy distribution in a following prediction duration, thus obtain Temperature Distribution in a following prediction duration according to equation (10), program is waited for until the next prediction moment afterwards; Specific as follows:

The thermal balance equation of control volume is (7):

$\frac{4}{3}{\mathrm{\π\ρc}}_p{r}^3\frac{\mathrm{dT}}{\mathrm{dt}}=f\stackrel{\·}{q}---\left(7\right)$

Integration is carried out to these equation both sides and obtains equation (8):

$T={\∫C}_{\mathrm{th}}\left(t\right)f\stackrel{\·}{q}\left(t\right)\mathrm{dt}---\left(8\right)$

In formula: T is control volume temperature, ρ is atmospheric density, c _pfor air specific heat, r is control volume radius, for built-up radiation hot strength, f is the photothermal absorption coefficient of control volume;

Compare with reflection with direct solar radiation, the heat radiation that cloud layer and earth surface itself produce has certain hysteresis quality in time, consider that cloud layer and the heat radiation of earth surface own are solar radiant energy lag-effects in time, the radiation heat in equation (7) calculating can be regarded as the past period solar radiant energy the accumulation of thermal effect, if thermal effect duration time lag of solar radiation is 12 hours, and feature time lag of solar radiant energy thermal effect follows BoxLucas model [11], and is defined as equation (9):

$f_{\mathrm{decay}}\left(t\right)=\frac{f_{\mathrm{bl}}\left(t\right)}{A}---\left(9\right)$

In formula: $f_{\mathrm{bl}}\left(t\right)=\frac{a_1(e^{-a_{2}t}-e^{-a_{1}t})}{a_1-a_2},$ $A={\∫}_0^{12}{f}_{\mathrm{bl}}\left(t\right)\mathrm{dt}$

T is obtained by equation (5), (8) and (9) _pthe prediction expression (10) of moment atmospheric temperature.

$T_p=T_0+\frac{C_{\mathrm{th}}f}{A}[{\∫}_{t_p-12}^{t_p}{f}_{\mathrm{bl}}\left(t\right){\stackrel{\·}{q}}_{\mathrm{rad}}\left(t\right)\mathrm{dt}-{\∫}_{t_0-12}^{t_0}{f}_{\mathrm{bl}}\left(t\right){\stackrel{\·}{q}}_{\mathrm{rad}}\left(t\right)\mathrm{dt}]---\left(10\right)$

As shown in Figure 1, be solar radiant energy thermal effect characteristic curve time lag based on BoxLucas model, as can be seen from the figure, solar radiant energy thermal effect reaches peak value at 1 hours, thermal effect reduces gradually afterwards, and its thermal effect is close to 0 after 12 hours, almost negligible.

As shown in Figure 2, compare with observed reading to the predicted value of the atmospheric temperature during-8 days on the 1st April in 2012 of Singapore for adopting the inventive method, can find out, when predicting that duration is less than 1 hour, predicted value and actual observed value meet well, wherein when predicting that duration is 30 minutes, absolute error is less than 1K, when prediction duration is that 1 little absolute error is constantly less than 3K.

Claims (3)

1. a Short-time-interval atmosphere ambient temperature prediction method, is characterized in that: comprise the steps:

Step 1: first, define in atmospheric environment sectional area be the spherical control volume of 1 square metre as research object, the sampling time step-length of setup control body environmental variance, setting prediction duration;

Step 2: every the time step of a setting, measure solar radiant energy, atmospheric temperature, pressure and relative humidity, the time span continued when sample record is greater than 12 constantly little, then start to perform atmospheric temperature predictor, and often perform an atmospheric temperature prediction through the time interval of a prediction duration;

Step 3: in atmospheric temperature prediction implementation, for a kth predicted time section, first according to atmospheric temperature, pressure and relative humidity that present sample obtains, calculated by equation (1) ~ (5) and determine C _thvalue, if k equals 1, then sets a value for air radiation heat absorption coefficients f in interval [0.001,0.01] scope, if k is more than or equal to 2, adopts equation (6) to upgrade the value of air radiation heat absorption coefficients f;

$\begin{array}{c}\ln \left(p_s\right)=\\ 54.842763-\frac{6763.22}{T}-4.21\ln \left(T\right)+0.000367T+\tanh \[0.0415\left(T-218.8\right)\]\[53.878-\\ \frac{1331.22}{T}-9.44523\ln \left(T\right)+0.014025T\]\end{array}---\left(2\right)$

$\mathrm{\ρ}=\frac{p(1-x)}{287.058T}+\frac{px}{461.495T}---\left(3\right)$

$C_{th}=\frac{3}{4{\mathrm{\π\ρc}}_p{r}^3}---\left(5\right)$

$f=\frac{A(T_0^k-T_0^{k-1})}{C_{th}^k\[{\∫}_{t_0^k-12}^{t_0^k}{f}_{bl}\left(t\right){\stackrel{\·}{q}}_{rad}\left(t\right)dt-{\∫}_{t_0^{k-1}-12}^{t_0^{k-1}}{f}_{bl}\left(t\right){\stackrel{\·}{q}}_{rad}\left(t\right)dt\]}---\left(6\right)$

In formula: ---air absolute humidity,

X---relative air humidity,

P---atmospheric pressure,

P _s----water vapor in air saturation pressure,

T---control volume temperature,

ρ---control volume atmospheric density,

C _p---control volume air specific heat,

R---control volume radius

F---the photothermal absorption coefficient of control volume,

T ₀---Current Temperatures,

T ₀---current time, unit: hour

F _bl(t)---BoxLucas function, expression formula is: parameter a in formula ₁, a ₂default value is taken as 0.5 and 3.8;

A---BoxLucas function is at the integration of interval [0,12], and expression formula is: $A={\∫}_0^{12}{f}_{bl}\left(t\right)dt;$

Interval [t ₀-12, t ₀] ---the solar radiant energy in past 12 hours;

Step 4: utilize solar radiant energy prediction module to predict solar radiant energy distribution in a following prediction duration, thus obtain Temperature Distribution in a following prediction duration according to equation (10), program is waited for until the next prediction moment afterwards; Specific as follows:

The thermal balance equation of control volume is (7):

$\frac{4}{3}{\mathrm{\π\ρc}}_p{r}^3\frac{dT}{dt}=f\stackrel{\·}{q}---\left(7\right)$

Integration is carried out to these equation both sides and obtains equation (8):

$T=\∫C_{th}\left(t\right)f\stackrel{\·}{q}\left(t\right)dt---\left(8\right)$

In formula: T is control volume temperature, ρ is atmospheric density, c _pfor air specific heat, r is control volume radius, for built-up radiation hot strength, f is the photothermal absorption coefficient of control volume;

Compare with reflection with direct solar radiation, the heat radiation that cloud layer and earth surface itself produce has certain hysteresis quality in time, consider that cloud layer and the heat radiation of earth surface own are solar radiant energy lag-effects in time, the built-up radiation hot strength in equation (7) calculating regard the past period solar radiant energy as the accumulation of thermal effect, if thermal effect duration time lag of solar radiation is 12 hours, and feature time lag of solar radiant energy thermal effect follows BoxLucas model, and is defined as equation (9):

$f_{decay}\left(t\right)=\frac{f_{bl}\left(t\right)}{A}---\left(9\right)$

In formula: $f_{bl}\left(t\right)=\frac{a_1(e^{-a_{2}t}-e^{-a_{1}t})}{a_1-a_2},$ $A={\∫}_0^{12}{f}_{bl}\left(t\right)dt$

T is obtained by equation (5), (8) and (9) _pthe prediction expression (10) of moment atmospheric temperature;

$T_p=T_0+\frac{C_{th}f}{A}\[{\∫}_{t_p-12}^{t_p}{f}_{bl}\left(t\right){\stackrel{\·}{q}}_{rad}\left(t\right)dt-{\∫}_{t_0-12}^{t_0}{f}_{bl}\left(t\right){\stackrel{\·}{q}}_{rad}\left(t\right)dt\]---\left(10\right)$

In formula: T ₀---Current Temperatures.

2. Forecasting Methodology according to claim 1, is characterized in that: described prediction duration is less than 1 hour.

3. Forecasting Methodology according to claim 1, is characterized in that: described sampling time step-length is 1 ~ 5 minute.