Air conditioner power consumption decomposition method based on temperature and power dependence, product and equipment

Abstract

The application discloses an air conditioner power decomposition method, product and equipment based on temperature and power dependence, and relates to the power supply and distribution field. The application selects all dates in a typical month in spring and autumn as an initial typical day set; filters typical days based on the dependence index of temperature and daily power corresponding to each date in the initial typical day set, obtains a typical working day set and a typical rest day set, and respectively determines the optimal regression model of each based on model selection criteria voting; uses the optimal regression model to respectively fit the benchmark power curve of the corresponding date; and calculates the difference between the total power corresponding to each date and the benchmark power as the final air conditioner power. According to the dependence index based on Copula and the equal-probability ellipse, the application filters the typical days, determines the benchmark power curve by means of the model selection method voting, and finally calculates the air conditioner power which is closer to the true value, thereby improving the accuracy, effectiveness and practicability of the air conditioner power calculation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power supply and distribution, in particular to an air conditioner power decomposition method based on temperature and power dependence, a product and equipment. BACKGROUND

[0002] Air conditioners are high energy consumption devices in residential buildings, office buildings, shopping malls and other places. With the continuous increase of air conditioner installation capacity, the proportion of air conditioner power in total power consumption is becoming larger and larger. Research on air conditioner power decomposition is of great significance for evaluating power consumption status, formulating power consumption scheduling plan and ensuring stable operation of power grid. At present, most of the typical day selection strategies for air conditioner power decomposition problems ignore the corresponding relationship between typical day power and temperature, resulting in deviation of the final air conditioner power calculation result. SUMMARY

[0003] The purpose of the present application is to provide an air conditioner power decomposition method based on temperature and power dependence, a product and equipment, which can improve the accuracy of air conditioner power calculation.

[0004] To achieve the above purpose, the present application provides the following solutions.

[0005] In one aspect, the present application provides an air conditioner power decomposition method based on temperature and power dependence, comprising:

[0006] Selecting all dates in typical spring and autumn months as an initial typical day set;

[0007] Screening typical days based on the dependence index of temperature and daily power corresponding to each date in the initial typical day set to obtain a typical working day subset and a typical rest day subset;

[0008] For the typical working day subset and the typical rest day subset, respectively based on model selection criteria voting to determine the optimal regression model of each;

[0009] Using the optimal regression model to fit the baseline power curve of the corresponding date; the horizontal coordinate of the baseline power curve is the date, and the vertical coordinate is the baseline power;

[0010] Calculating the difference between the total power corresponding to each date and the baseline power as the final air conditioner power.

[0011] Optionally, the selecting all dates in typical spring and autumn months as an initial typical day set specifically comprises:

[0012] Determining the typical spring month as March and the typical autumn month as November, and regarding all dates in March and November as typical days to form an initial typical day set.

[0013] Optionally, the method further comprises:

[0014] dividing all the dates in the initial typical day set into a typical weekday set and a typical weekend set according to weekdays and weekends;

[0015] calculating the dependence σ of the typical daily electricity consumption and its corresponding temperature for the typical weekday set and the typical weekend set respectively;

[0016] if the dependence σ is less than an empirical threshold η, constructing an equiprobable ellipse based on the temperature and the typical daily electricity consumption, and filtering the typical days based on the equiprobable ellipse to obtain a new typical day set and returning to the step of calculating the dependence σ of the typical daily electricity consumption and its corresponding temperature;

[0017] when the dependence σ is greater than or equal to the empirical threshold η, obtaining the typical weekday set and the typical weekend set from the filtered final typical day set.

[0018] Optionally, the step of calculating the dependence σ of the typical daily electricity consumption and its corresponding temperature comprises:

[0019] using the formula to calculate the dependence σ of the typical daily electricity consumption and its corresponding temperature; wherein X1 and X2 are random variables corresponding to the typical daily electricity consumption and the temperature respectively; vector v ∈ [0, 1] 2 ; C(v) is the Copula of X1 and X2; Π(v) is the product of the distribution function of X1 and the distribution function of X2; σ(X1, X2) is abbreviated as σ, 0 ≤ σ ≤ 1.

[0020] Optionally, the step of constructing an equiprobable ellipse based on the temperature and the typical daily electricity consumption comprises:

[0021] under the assumption that the joint distribution of the temperature and the typical daily electricity consumption obeys a Gaussian distribution, constructing an equiprobable ellipse wherein and respectively represent the mathematical expectation of X1 and X2; M -1 is the inverse matrix of the cross-correlation matrix M; r represents the random variable interval distribution threshold.

[0022] Optionally, the step of determining the optimal regression model for the typical weekday set and the typical weekend set respectively based on the model selection criterion voting comprises:

[0023] The daily electricity consumption curves of the typical working day subset and the typical rest day subset are fitted by using a plurality of reference regression models to obtain a plurality of regression models; the plurality of reference regression models include a linear model, a parabolic model, an exponential model and a logarithmic model;

[0024] The score of each regression model under all model selection criteria and the total score of the corresponding regression model are calculated;

[0025] The regression model with the minimum total score is determined as the optimal regression model.

[0026] Optionally, the reference electricity consumption curve corresponding to each date is fitted by using the optimal regression model, and the fitting specifically includes:

[0027] The reference electricity consumption curves of each working day are fitted by using the optimal regression model corresponding to the typical working day subset, and the reference electricity consumption curves of each rest day are fitted by using the optimal regression model corresponding to the typical rest day subset, so that the reference electricity consumption curves of the whole year are obtained.

[0028] In another aspect, the present application further provides a computer program product, comprising a computer program which, when executed by a processor, implements the steps of the air conditioner electricity consumption decomposition method based on temperature and electricity dependence.

[0029] In another aspect, the present application further provides a computer readable storage medium, which stores a computer program, and the computer program, when executed by a processor, implements the steps of the air conditioner electricity consumption decomposition method based on temperature and electricity dependence.

[0030] In still another aspect, the present application further provides a computer device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the steps of the air conditioner electricity consumption decomposition method based on temperature and electricity dependence.

[0031] According to the specific embodiments of the present application, the following technical effects are achieved:

[0032] This invention discloses a method, product, and device for decomposing air conditioning power consumption based on the dependence of temperature and electricity consumption. The method involves selecting all dates in typical spring and autumn months as an initial set of typical days; filtering typical days based on the dependence index of temperature and electricity consumption corresponding to each date in the initial set of typical days to obtain a set of typical working days and a set of typical rest days; determining the optimal regression model for each set of typical working days and typical rest days based on model selection criteria; fitting the baseline power consumption curve for each date using the optimal regression model; the horizontal axis of the baseline power consumption curve represents the date, and the vertical axis represents the baseline power consumption; and calculating the difference between the total power consumption corresponding to each date and the baseline power consumption as the final air conditioning power consumption. This invention filters typical days based on Copula-based dependence indices and equal probability ellipses, and uses a model selection method to determine the baseline power consumption curve through voting. The final calculated air conditioning power consumption is closer to the actual value, improving the accuracy, effectiveness, and practicality of air conditioning power consumption calculation. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 A flowchart of the air conditioner power decomposition method based on temperature and power dependence provided by the present invention;

[0035] Figure 2 This is a schematic diagram of the daily electricity consumption and temperature curves for a certain region in 2019.

[0036] Figure 3 This is a schematic diagram illustrating the screening process for a typical day set in this invention;

[0037] Figure 4 This is a schematic diagram of the equiprobability ellipse constructed in this invention;

[0038] Figure 5 This is a schematic diagram illustrating the process of selecting the optimal regression model in this invention;

[0039] Figure 6 This is a schematic diagram of the air conditioner power consumption decomposition process in the calculation example of the present invention;

[0040] Figure 7 This is a graph showing the daily electricity consumption versus temperature in City A in the calculation example of this invention;

[0041] Figure 8 This is a bubble chart showing the typical days selected from 2019 to 2022 in the calculation examples of this invention;

[0042] Figure 9 A schematic diagram of the benchmark power curve from 2019 to 2022 obtained by calculation in the calculation example of the present application;

[0043] Figure 10 A schematic diagram of the air conditioner power curve from 2019 to 2022 calculated in the calculation example of the present application;

[0044] Figure 11 A schematic diagram of the air conditioner power curve calculated by different methods in the calculation example of the present application. DETAILED DESCRIPTION

[0045] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0046] The air conditioner power decomposition method, product and equipment based on temperature and power dependence of the present application can improve the accuracy of air conditioner power calculation.

[0047] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0048] As shown in the drawings, Figure 1 The air conditioner power decomposition method based on temperature and power dependence provided by the present application comprises:

[0049] Step 1: Select all dates in a typical month in spring and autumn as an initial typical day set.

[0050] Air conditioner power is mainly the power consumed by refrigeration in summer and heating in winter. In spring and autumn, the temperature is relatively suitable, and there is no air conditioner power by default. Therefore, the dates in spring and autumn are usually selected as typical days, such as in some common air conditioner power calculation methods such as benchmark load comparison method and maximum load comparison method, which select typical days according to this strategy. This typical day selection strategy has obvious shortcomings, such as the phenomenon of back-to-spring in southern China and the phenomenon of early spring cold, which leads to the fact that air conditioners are still needed to adjust indoor temperature in spring and autumn, resulting in air conditioner power consumption, so it is necessary to further improve the selection strategy of typical days. In the method of the present application, the typical month in spring is March, and the typical month in autumn is November. All dates in March and November are selected as typical days to form an initial typical day set.

[0051] Step 2: screening typical days based on the dependence index of temperature and daily electricity consumption corresponding to each date in the initial typical day set, to obtain a typical weekday set and a typical weekend day set.

[0052] The conventional method uses the absolute value of the Pearson correlation coefficient based on temperature and typical daily electricity consumption to evaluate the quality of the typical day set, so as to determine whether the typical day needs to be further filtered. However, the Pearson correlation coefficient can only describe the linear relationship between random variables, while the relationship between temperature and electricity consumption is often nonlinear, as shown in Figure 2 . Figure 2 It is shown that the distribution of daily electricity consumption data in a certain region in 2019, and it can be seen that the daily electricity consumption and the temperature do not satisfy the linear relationship, that is, the quantitative relationship between the electricity consumption and the temperature should not be simplified as a linear relationship. Therefore, it is necessary to evaluate the quality of the typical day set based on an index that can describe the nonlinear relationship between random variables, and then design a filtering method for the typical day set.

[0053] Figure 3 The screening process of the typical day set of the application is shown, referring to Figure 3 , step 2 screens typical days based on the dependence index of temperature and daily electricity consumption corresponding to each date in the initial typical day set, to obtain a typical weekday set and a typical weekend day set, specifically including:

[0054] Step 2.1: dividing all dates in the initial typical day set into a typical weekday set and a typical weekend day set according to weekdays and weekend days.

[0055] Step 2.2: calculating the dependence σ of typical daily electricity consumption and its corresponding temperature for the typical weekday set and the typical weekend day set, respectively.

[0056] In the application, the dependence index SWσ is used to measure the dependence degree of typical daily electricity consumption and its corresponding temperature, and the specific calculation formula is:

[0057]

[0058] wherein X1 and X2 are random variables corresponding to typical daily electricity consumption and temperature, respectively; vector v [0,1] 2 , C(v) is the Copula of X1 and X2, and Π(v) is the product of the distribution function of X1 and the distribution function of X2. Therefore, it is described that the distance of the Copula of X1 and X2 and their distribution product. The dependence index σ(X1,X2) between X1 and X2 can be simply written as σ, 0≤σ≤1, and the greater the value of σ, the higher the dependence degree of X1 and X2. The dependence index SWσ separates the edge distribution and correlation modeling, and can describe the nonlinear relationship between random variables, so as to overcome the deficiency that the Pearson correlation coefficient can only measure the linear correlation coefficient.

[0059] In formula (1), calculating the dependence index SWσ requires estimating the Copula of temperature and typical daily electricity. In the present application, the Copula function is estimated by using an empirical formula, i.e.:

[0060]

[0061] wherein v in formula (1) is a two-dimensional vector v = [[v1, v2], each dimension taking [0, 1], i.e. v1, v2 in formula (2) belongs to [0, 1], so the vector v ∈ [0, 1] 2 In formula (1), C(v) is the Copula function of X1 and X2, and the left side of the equal sign in formula (2) is the estimated value of the Copula function, and the right side of the equal sign in formula (2) is a 1 function, which takes 1 when the two conditions in the brackets are met, and takes 0 otherwise. Xj(j = 1 or 2) represents the observed value of Xj; is the estimated distribution function of the random variable Xj, and j takes 1 or 2, corresponding to and

[0062] Specifically,

[0063] wherein x is a random variable, is the observed value of the random variable, and N is the number of observed values. Formula (3) and formula (2) have the same principle, and the difference lies in that the left side of the equal sign in formula (3) is an estimated value of a distribution function, and the right side is a 1 function, which takes 1 when the conditions in the brackets are met, and takes 0 otherwise.

[0064] Step 2.3: If the dependence σ is less than the empirical threshold η, an equiprobable ellipse is constructed based on the temperature and the typical daily electricity, and the typical days are filtered based on the equiprobable ellipse, to obtain a new set of typical days and return to the step of calculating the dependence σ of the typical daily electricity and the corresponding temperature.

[0065] In the present application, the empirical threshold η of the dependence index σ is set to 0.8, i.e. when the dependence index σ ≥ 0.8, it is considered that the dependence between the set of typical days and the temperature is obvious, and there is no need to filter the outliers in the set of typical days; otherwise, it is considered that the overall consistency of the set of typical days is poor, and there are outliers in the set, and an equiprobable ellipse needs to be constructed to filter the outliers in the typical days.

[0066] Under the assumption that the joint distribution of temperature and typical daily electricity obeys a Gaussian distribution, the following equiprobable ellipse is constructed:

[0067]

[0068] wherein and E(X1) and E(X2) are the mathematical expectations of X1 and X2, respectively; M is the cross-correlation matrix, M -1 is the inverse matrix of the cross-correlation matrix; r is a random variable interval distribution threshold.

[0069]

[0070] Given a confidence level γ of the typical day outliers, the parameter r is obtained by solving the following equation:

[0071]

[0072] Equation (6) can be solved by a standard normal distribution lookup table. At this time, the probability of the typical day falling outside the equal-probability ellipse is:

[0073]

[0074] In the present application, the confidence level γ is set to 5%, that is, assuming that the joint distribution of temperature and typical daily electricity consumption obeys a Gaussian distribution, an ellipse is fitted in a two-dimensional plane, so that the probability of points falling outside the equal-probability ellipse is 5%, so that there is a 95% probability of falling within the ellipse. The equal-probability ellipse constructed in the present application is as shown in Figure 4 , Figure 4 where λ1 and λ2 are eigenvalues of the cross-correlation matrix M, and the dashed rectangle is the circumscribed rectangle of the innermost ellipse , and the circumscribed ellipse (the outermost ellipse) of the circumscribed rectangle is the equal-probability ellipse Ω represents a rectangular space, j takes [1, 2], at this time, there are two value ranges corresponding to the length and width of the circumscribed rectangle, respectively.

[0075] When the dependence index σ is less than 0.8, it is considered that the overall consistency of the typical day set is poor, and there are outliers in the set. At this time, the equal-probability ellipse constructed by the present application is used to filter the outliers in the typical day set, that is, to filter the points falling outside the equal-probability ellipse, to obtain a new typical day set to continue the judgment of the size of the dependence σ, until all the typical day electricity consumption and temperature dependence σ in the new typical day set are greater than or equal to 0.8, to obtain the final typical day set.

[0076] Step 2.4: When the dependence σ is greater than or equal to the empirical threshold η, the final typical day set after filtering is used to obtain a typical workday subset and a typical rest day subset.

[0077] All the dates remaining in the final typical day set are divided into a typical workday subset and a typical rest day subset according to workdays and rest days.

[0078] Step 3: For the typical workday subset and the typical rest day subset, the optimal regression model of each is determined based on the model selection criteria voting.

[0079] On the basis of the typical working day set and the typical rest day set determined in step 2, a regression analysis method is used to fit the benchmark power curve, i.e. the benchmark power curve is determined by fitting a regression curve of the typical day set and the time label.

[0080] The traditional method selects a model with the largest determination coefficient as the optimal regression model to fit the benchmark power curve by calculating the determination coefficient R2 of each model for the selection of the multiple regression model. Although this method is simple and easy to implement, it has the disadvantage that the model selection index is too single and the actual optimal regression model cannot be selected. Therefore, the present application selects the optimal regression model by voting based on multiple model selection criteria. The model selection criteria used in the present application include Akaike information criterion (AIC), Bayesian information criterion (BIC) and Deviance information criterion (DIC).

[0081] The reference regression models used include linear model, parabolic model, exponential model and logarithmic model:

[0082] P1=j1t+k1 (8)

[0083] P2=j2t 2 +k2t+l2 (9)

[0084]

[0085] P4=j4t+k4lnt (11)

[0086] Where j1, j2, j3, j4, k1, k2, k3, k4, l2 are estimated parameters in the model, which can be estimated by the least square method.

[0087] Figure 5 The process of voting to select the optimal regression model of the present application is shown, referring to Figure 5 , step 3 is based on the model selection criteria to vote to determine the optimal regression model of each set of typical working days and typical rest days, specifically including:

[0088] Step 3.1: For the typical working day set and the typical rest day set, respectively, a plurality of reference regression models are used to fit the daily power curve, obtaining a plurality of regression models.

[0089] The optional reference regression models of the present application include linear model (8), parabolic model (9), exponential model (10) and logarithmic model (11) and the like. Considering that the power consumption on weekdays is obviously different from that on holidays, the present application subdivides the typical day set into a typical weekday subset and a typical holiday subset, and selects optimal regression models for the two subsets respectively to fit the corresponding benchmark power curves.

[0090] Step 3.2: Calculate the score of each regression model under all model selection criteria and the total score as the corresponding regression model.

[0091] First, four reference regression models are used to fit the daily power curves of the typical weekday / holiday subsets respectively, and four regression models are obtained. Then, the scores of the four regression models under AIC, BIC and DIC criteria are calculated, respectively denoted as and i = 1, 2, 3, 4. Next, the scores of the four regression models under the same model selection criterion are normalized; finally, the score sum of each regression model under all model selection criteria is calculated as the regression model P j (j = 1, 2, 3, 4) final total score:

[0092]

[0093] Step 3.3: Determine the regression model with the minimum total score as the optimal regression model.

[0094] The total score C j ∈ [0, 1] calculated in formula (12) is also called a weighted model selection index. The smaller the total score under the above model selection criteria, the closer the weighted model selection index to 0, indicating that the performance of the corresponding regression model is better. Therefore, the regression model corresponding to the minimum total score in (12) is the optimal regression model.

[0095] Step 4: Use the optimal regression model to fit the benchmark power curve of the corresponding date respectively; the horizontal coordinate of the benchmark power curve is the date, and the vertical coordinate is the benchmark power.

[0096] After obtaining the filtered typical day set, the present application uses four different reference regression models to fit, find the parameters in the equation, determine the equation; then uses the model selection criteria to determine the optimal equation; the determined optimal equation is the optimal regression model obtained by fitting, that is, the equation of the benchmark power curve. In other words, the optimal regression model is an equation curve, whose horizontal coordinate is the date and the vertical coordinate is the daily power. Such a curve is the benchmark power curve. At the same time, it should be noted that the fitting is performed separately for weekdays and holidays, and different benchmark power curve equations are fitted.

[0097] Step 5: Calculate the difference between the total power and the benchmark power on each date as the final air conditioner power.

[0098] Based on the benchmark power curve calculated in step 4, the daily air conditioner power can be calculated. Specifically, the points above the benchmark power equation curve calculated in step 4 are the benchmark power values arranged by date, and the difference between the total power and the benchmark power is calculated as the final air conditioner power, i.e.:

[0099] P AIR = P L -P B (13)

[0100] where P AIR , P L , and P B are the air conditioner power, the total power, and the benchmark power, respectively.

[0101] The present application studies the unsupervised decomposition of air conditioner power and proposes an air conditioner power decomposition method based on typical day screening and regression curve voting. The present application method screens typical days based on equal probability ellipses and designs a voting method to select the optimal regression model to fit the benchmark power curve. The accuracy, effectiveness, and practicality of the air conditioner power decomposition method proposed in the present application are verified by the experimental results of real power consumption data.

[0102] A specific calculation example of the present application method is provided below, and the air conditioner power decomposition process is shown in Figure 6 The calculation example involves data collected from City A in southern China. The climate of City A belongs to the typical subtropical monsoon climate, which is warm and humid. The summer climate is hot, and it is the period with the highest air conditioner power consumption in a year; although the winter is not cold, it will still consume some heating air conditioner power if it encounters a long period of cooling weather. The daily power consumption curve of City A from 2019 to 2022 is shown in Figure 7 It can be seen that: in each summer, the daily power consumption and the daily average temperature are at the highest level in a year; during the period from October to December each year, although the temperature is decreasing, the daily power consumption remains roughly at the same level; in winter, from January, the daily power consumption will experience a stage of first decline and then rise.

[0103] In this calculation example, the dates in March and November, which have almost no air conditioner power consumption, are defined as the initial typical day set. According to the typical day selection strategy based on the dependence of temperature and power in the present application method, the initial typical day set is subdivided into weekdays and weekends, and typical day screening is performed respectively. Table 1 shows the typical day screening results in 2019.

[0104] Table 1 Typical day screening results in 2019

[0105]

[0106] To better illustrate the effectiveness of the typical screening strategy proposed in the present application, the specific typical days to be removed and the corresponding power and temperature of these typical days are specifically listed in Table 1, and the dependence index of the power and temperature of the typical day set before and after screening is given. It can be seen that the typical days to be removed are either too high in power value (possibly caused by too large data acquisition error) or the temperature is lower than the comfortable temperature interval (in the present application, the temperature interval without air conditioning cooling or heating is [20℃, 25℃]); after screening, the dependence index of the typical day set and the corresponding temperature increases significantly.

[0107] The screening statistical results of the typical day set composed of all typical days from 2019 to 2022 are shown in Table 2.

[0108] Table 2 Screening statistical results of typical days from 2019 to 2022

[0109]

[0110] It can be found from Table 2 that the daily power and daily average temperature dependence is high in 2020, and no removal is needed to meet the threshold, indicating that extreme weather phenomena in March and November of 2020 are less than in other three years; a total of 3 days are removed in 2021, and the abnormal dates are mainly concentrated in the working days of November. Because the temperature suddenly dropped in several days, while the daily power data did not change significantly, so it had a negative impact on the dependence index; 8 days and 7 days were removed in 2019 and 2022 respectively, and the removed dates were relatively concentrated. The main reason is that meteorological factors cause the temperature to suddenly drop or rise on a certain day, which differs from the temperature of adjacent days, while the power changes less in the short term than the temperature, thereby affecting the value of the dependence index. For example, most of the removed dates in 2022 are in March, because the local area experienced a large range of low temperature and snow weather in the spring, which caused the temperature to intermittently decrease at the beginning of March, and the local people rarely turned on the air conditioner in the spring, ultimately leading to the deterioration of the consistency of daily power and temperature.

[0111] For ease of observation, the typical day set and the removed abnormal typical days are both visualized in Figure 8 . Figure 8 In the figure, the horizontal coordinate is the temperature of the typical day, and the vertical coordinate is the power of the typical day. It can be seen that the removed typical days have obvious physical meaning, i.e. outliers.

[0112] Further, the baseline power fitting curves based on the screened typical working day set and typical holiday set from 2019 to 2022 are shown in Figure 9 . For ease of comparison, Figure 9The fitting curves of all four regression models are plotted. Table 3 shows the selection results of the regression models in the process of fitting the benchmark power for 2019-2022: in 2019, the power function model was selected as the final optimal regression model for the fitting of the weekday benchmark curve; the logarithmic function model was selected as the final optimal regression model for the fitting of the weekend benchmark curve; in 2020 and 2021, the power function model was selected as the optimal regression model for the fitting of the weekday and weekend benchmark curves; in 2022, the logarithmic function model was selected as the optimal regression model for the fitting of the weekday and weekend benchmark curves.

[0113] Table 3: Regression model selection results for benchmark curves in 2019-2022

[0114]

[0115]

[0116] In Table 3, the linear function, quadratic function, power function, and logarithmic function correspond to the linear model, parabolic model, exponential model, and logarithmic model in step 3, respectively.

[0117] Further, the optimal regression model for work / weekend is used to fit the benchmark power curve for all dates throughout the year, and the final air conditioner power is obtained by subtracting the benchmark power from the total power. Figure 10 The air conditioner power curve for 2019-2022 decomposed using the method of the present application is shown. By observing the air conditioner power curve and temperature curve in Figure 10 , it is found that the change trends of the two are roughly the same, mainly because the air conditioner power is most affected by temperature, especially in summer, and the most intuitive performance is that the higher the temperature, the greater the air conditioner power consumption.

[0118] To verify the effectiveness of the method of the present application (the method herein), the benchmark load comparison method and the linear decomposition method are used to calculate the air conditioner power of A city from 2019 to 2022, and compared with the method of the present application. Figure 11 The air conditioner power results calculated by the three methods in 2020 are shown. By observing, the three curves have the same trend. The difference is that, compared with the method of the present application, the air conditioner power obtained by the two comparison methods fluctuates relatively large. This is because the benchmark load comparison method and the linear decomposition method both select all the typical days to fit the benchmark curve, which leads to more negative days in the calculated air conditioner power, such as in the interval from January to April. The negative day ratio of the air conditioner power calculated by the three methods is shown in Table 4.

[0119] Table 4: Negative day ratio of air conditioner power in 2020

[0120] Air conditioner power consumption calculation method Air conditioner power consumption negative value proportion Benchmark load comparison method 34.15％ Linear decomposition method 31.15％ Invention method 26.78％

[0121] It can be seen that the method of the present application benefits from the voting selection of the typical day filtering and the optimal regression model, and obtains the lowest proportion of negative air conditioning power. Meanwhile, the dependence indexes of the three methods are close, as shown in Table 5.

[0122] Table 5 Dependence indexes of three air conditioning decomposition methods

[0123] Air conditioner power consumption calculation method Dependency index Benchmark load comparison method 0.96 Linear decomposition method 0.99 Invention method 0.96

[0124] Although the linear decomposition method has a higher dependence, it contains too many negative values of air conditioning power in the calculation result, which is not consistent with the actual life situation. In contrast, the method of the present application not only has fewer negative values of air conditioning power, but also has a superior dependence index, verifying the accuracy, effectiveness and practicability of the method of the present application.

[0125] The method of the present application proposes a typical day selection strategy, uses the Copula-based dependence index to evaluate the quality of the initial typical day set, and under the assumption of Gaussian joint distribution, uses the temperature data and the initial typical day set to construct the equiprobable ellipse, and based on the equiprobable ellipse, realizes the filtering of the outliers in the typical day. Based on the typical day set after the outlier filtering, a variety of regression models are used to fit the benchmark curve of the power, and based on the three model selection criteria, the optimal benchmark curve is determined by voting, and the calculation of the air conditioning power is realized. Further, according to the actual calculation of the air conditioning power based on the real power data of a certain region, the effectiveness of the method of the present application is verified through the comparison test.

[0126] In some embodiments, the present application also provides a computer program product comprising a computer program which, when executed by a processor, implements the steps of the air conditioning power decomposition method based on the temperature and power dependence.

[0127] In some embodiments, the present application also provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the air conditioning power decomposition method based on the temperature and power dependence.

[0128] In some embodiments, the present application also provides a computer device comprising a processor, a memory, an input / output interface (I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store to-be-processed transactions. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with external terminals through network connection. The computer program, when executed by the processor, is capable of implementing the steps of the air conditioner power decomposition method based on temperature and power dependence.

[0129] The principles and implementation manners of the present application are described herein by applying specific examples, and the above examples are only used to help understand the method of the present application and its core idea; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation manners and application scopes will have changes. In conclusion, the content of the present specification should not be understood as a limitation of the present application.

Claims

1. An air conditioner power decomposition method based on temperature and power dependence, characterized by, The method comprises the following steps: selecting all dates in typical months in spring and autumn as an initial typical day set; screening typical days based on a dependence index of temperature and daily electricity consumption corresponding to each date in the initial typical day set to obtain a typical working day set and a typical rest day set; The step of screening typical days based on a dependence index of temperature and daily electricity consumption corresponding to each date in the initial typical day set to obtain a typical working day set and a typical rest day set comprises the following steps: dividing all dates in the initial typical day set into a typical working day set and a typical rest day set according to working days and rest days; calculating the dependence σ of typical daily electricity consumption and corresponding temperature for the typical working day set and the typical rest day set respectively; if the dependence σ is less than an empirical threshold η, constructing an equiprobability ellipse based on temperature and typical daily electricity consumption, and filtering typical days based on the equiprobability ellipse to obtain a new typical day set and return to the step of calculating the dependence σ of typical daily electricity consumption and corresponding temperature; when the dependence σ is greater than or equal to the empirical threshold η, obtaining the typical working day set and the typical rest day set from the filtered final typical day set; The step of calculating the dependence σ of typical daily electricity consumption and corresponding temperature comprises the following steps: The dependence σ of the typical daily electricity consumption on the corresponding temperature is calculated using the formula where X1 and X2 are random variables corresponding to the typical daily electricity consumption and the temperature, respectively; the vector v e [0, 1] 2 ; C(v) is the Copula of X1 and X2; Π(v) is the product of the distribution function of X1 and the distribution function of X2; σ(X1, X2) is abbreviated as σ, and 0≤σ≤1; The step of constructing an equiprobability ellipse based on temperature and typical daily electricity consumption comprises the following steps: Under the assumption that the joint distribution of temperature and typical daily electricity consumption obeys a Gaussian distribution, an equiprobable ellipse is constructed wherein, and E[X1] and E[X2] denote the mathematical expectations of X1 and X2, respectively; M -1 M-1 denotes the inverse matrix of the cross-correlation matrix M; r denotes a random variable interval distribution threshold value; voting to determine the optimal regression model for the typical working day set and the typical rest day set respectively based on a model selection criterion; fitting the benchmark electricity consumption curve of the corresponding date using the optimal regression model; the abscissa of the benchmark electricity consumption curve is date, and the ordinate is benchmark electricity consumption; calculating the difference between the total electricity consumption and the benchmark electricity consumption corresponding to each date as the final air conditioner electricity consumption.

2. The temperature and electricity dependency based air conditioner electricity decomposition method of claim 1, wherein, The step of selecting all dates in typical months in spring and autumn as an initial typical day set comprises the following steps: determining the typical months in spring as March and in autumn as November, and regarding all dates in March and November as typical days to form the initial typical day set.

3. The temperature and electricity dependency based air conditioner electricity decomposition method of claim 2, wherein, The step of voting to determine the optimal regression model for the typical working day set and the typical rest day set respectively based on a model selection criterion comprises the following steps: fitting the daily electricity consumption curve of the typical working day set and the typical rest day set using a plurality of reference regression models to obtain a plurality of regression models; the plurality of reference regression models include a linear model, a parabolic model, an exponential model and a logarithmic model; calculating the score of each regression model under all model selection criteria and the total score of the corresponding regression model; determining the regression model with the smallest total score as the optimal regression model.

4. The air conditioner power decomposition method based on temperature and power dependence according to claim 3, characterized by, The step of fitting the benchmark electricity consumption curve of the corresponding date using the optimal regression model comprises the following steps: fitting the benchmark electricity consumption curve of each working day using the optimal regression model corresponding to the typical working day set, and fitting the benchmark electricity consumption curve of each rest day using the optimal regression model corresponding to the typical rest day set to obtain the benchmark electricity consumption curve of the whole year.

5. A computer program product comprising a computer program, characterized in that, The computer program is executed by a processor to implement the steps of the air conditioner electricity consumption decomposition method based on temperature and electricity consumption dependence according to any one of claims 1-4.

6. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by a processor, implements the steps of the air conditioner power decomposition method based on temperature and power dependency of any one of claims 1-4.

7. A computer device comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the air conditioner power decomposition method based on temperature and power dependency of any one of claims 1-4.

Images