Method suitable for non-intrusive identification of operation load of central air conditioner

Abstract

According to the non-intrusive identification method suitable for the operation load of the central air conditioner, the central air conditioners of different brands are selected, electrical waveforms of the complete operation process of the central air conditioners are recorded, time-frequency domain characteristic analysis is carried out, common characteristics are extracted, and total daily electrical load data and daily air temperature data of a user in the past one year are selected; analyzing the incidence relation between the daily electrical load curve and the daily temperature curve of the user, obtaining a heating load area, a basic load area and a cooling load area, performing linear data fitting with the temperature to obtain a load curve, and subtracting the basic load curve from the cooling load curve to obtain a load curve; further judging whether the central air conditioner is in a starting / stopping state or not, and obtaining the electric quantity of the central air conditioner equipment in the operation period through steady-state power integration by adopting sub-item electric quantity estimation in the starting event and stopping event period of the user air conditioner; and running state tracking and electricity consumption estimation of the central air conditioner are realized.

Description

Technical Field

[0001] The present invention relates to the field of power load monitoring of power systems, and in particular to a non-intrusive identification method for providing accurate characteristics of the operating state and power decomposition of a central air-conditioning operating load. Background Art

[0002] Summer and winter in my country are the peak seasons for power grid electricity consumption. Among them, the electricity consumed by air-conditioning load accounts for a relatively high proportion, and highly overlaps with the peak power consumption of the power grid. Therefore, in the work of power demand response, it is necessary to reasonably optimize the allocation of power resources by adjusting the air-conditioning load.

[0003] The old method of adjusting air conditioning load relies heavily on monitoring the actual situation of users on the spot / investigating them to count and check load resources. This method obviously cannot obtain the air conditioning load situation and start-stop status in real time on a large scale, lacks precise control of air conditioning resources, and causes data distortion.

[0004] Under the existing technology, the method of installing invasive monitoring equipment on the equipment side is usually adopted to realize the monitoring of air conditioning load. However, through actual on-site operations, it is found that the method of installing invasive monitoring equipment has a large investment and requires maintenance, resulting in high costs. Therefore, most electricity users still choose to adopt non-invasive load monitoring solutions.

[0005] However, the existing non-intrusive load monitoring solutions have the problem that there are many types and brands of air conditioners on the market, and the power range varies greatly, so it is impossible to establish a complete load feature library, and the load identification is difficult. Therefore, the non-intrusive load identification technology based on the existing technology can only identify the air-conditioning equipment on the market that has completed field recording and established a feature library, resulting in the established model being incomplete, poor in versatility and inaccurate, and having poor practical effect.

[0006] Based on the above problems, a new method for extracting air-conditioning load characteristics is needed to realize the wide application of identification models. Summary of the invention

[0007] In order to solve the problems that the interventional monitoring equipment under the prior art is relatively high in cost, while the non-intrusive load monitoring solution has the problem of lack of load feature library, high difficulty in load identification, poor versatility and inaccuracy, the present invention aims at the problem that the non-intrusive load identification technical solution of the prior art cannot realize accurate identification of multiple types of air-conditioning loads and the model generalization is not strong. A method for non-intrusive identification of central air-conditioning operating load is proposed. The method improves the integrity of the load identification model by adopting a composite mode based on the air-conditioning load operating characteristics of each household + electrical characteristics extraction, and can complete the parameter training of "one household, one policy" without household surveys, effectively improving the versatility of the identification model, and the identification is accurate and efficient.

[0008] The present invention is a method for non-intrusive identification of central air-conditioning operating load, and its specific steps are as follows:

[0009] 1) First, we selected central air conditioners of different brands, recorded the electrical waveforms of their complete operation process, and performed time-frequency domain feature analysis to extract the common features existing in the operation process of air conditioners of different brands, namely, the startup peak value P f , steady-state power value P s , shutdown power conversion P e , downtime T e and current distortion rate THD, and form a typical air-conditioning operation power curve; in this step, central air-conditioners of different brands are selected, the number of air-conditioners is 5 or more, and the number of electrical waveforms recorded during the complete operation process is 1 or more;

[0010] It should be noted that the steady-state power of different air-conditioning loads is closely related to the air-conditioning capacity and the current temperature difference.

[0011] 2) In this step, for different users and different central air conditioning models, the total daily electricity load data and daily temperature data of the users in the past year are selected, and the correlation between the daily electricity load curve and the daily temperature curve of the users is analyzed by region, specifically:

[0012] Based on the 96-point daily electricity load curve data collected by users and daily temperature data, the load aggregation density D(P t,p ),as follows:

[0013]

[0014] Where: N(P t,p ) is the number of times the temperature is t and the power is p in the daily load curve data of one year;

[0015] t i is the range of temperature variation within a year;

[0016] p k is the range of power variation within a year;

[0017] According to the actual situation, the heating load area, basic load area and cooling load area are divided, the data of different load areas are linearly fitted with the temperature, and the load regression curves of these three areas are obtained by regression clustering:

[0018] p=β k t+b k ,k=1,2,3

[0019] Where: p is power, t is temperature, β k ,b k are the slope and intercept of the kth (k=1, 2, 3) fitting curve respectively;

[0020] The load regression curve is divided into a cooling load curve LC = F C (T), basic load curve LB = F B (T) and heating load LH = F H (T), there are three curves in total. By subtracting the cooling load curve from the basic load curve, the relationship formula between the steady-state power and temperature of the central air conditioner for different users can be obtained;

[0021] 3) Finally, the load identification module updates the identification program according to the obtained steady-state power parameters, and calculates the power characteristic value through the original high-frequency voltage and current data. The obtained power characteristic data is further calculated using the sliding window bilateral CUSUM algorithm to determine whether the central air conditioner is in the start / stop state. During the user's air conditioning start-up event and shutdown event, the sub-item power estimation is used, and the steady-state power integration is used to obtain the power consumption of the central air-conditioning equipment during operation, and finally the operation status tracking and power consumption estimation of the central air conditioner are obtained.

[0022] According to a method for non-intrusive identification of central air-conditioning operating load of the present invention, it is characterized in that the typical air-conditioning operating power curve in step 1) has the following values: startup peak value P f The maximum power value reached during the air conditioner startup process, the steady-state power value P s is the average power value during stable operation of the air conditioner, and the shutdown power is P e is the total power change during the air conditioner shutdown process, and the shutdown time T e It is the overall duration of the air conditioner shutdown process, and the current distortion rate THD is the effective value of all non-fundamental component currents - including harmonics and interharmonics - divided by the effective value of the current fundamental wave 50Hz.

[0023] According to the present invention, a method for non-intrusive identification of central air-conditioning operating load is characterized in that the startup peak value P f Select 1.2 to 3 times the steady-state power value P s , and the shutdown power becomes P e Then select 0.8-~1.2 times the steady-state power value P s .

[0024] According to the present invention, a method suitable for non-intrusive identification of central air-conditioning operating load is characterized in that the "adopting regression clustering method" in the step 2) specifically adopts a Gaussian mixture regression clustering model, sets the number of clusters to 3, and simultaneously calculates a fitted load regression curve.

[0025] According to a method for non-intrusive identification of central air-conditioning operating load of the present invention, it is characterized in that the step 2) includes data preprocessing, which is specifically:

[0026] Set the load aggregation density threshold e1 = 0.6. When the load aggregation density D(P t,p )<=e1, the corresponding sample data point (t,p) is eliminated.

[0027] A method for non-intrusive identification of central air-conditioning operating load according to the present invention is characterized in that the step 2) includes weighting of the sample points after elimination, which is specifically:

[0028] Weight the N remaining sample points (t, p), and the weighted sample points (t i ,p i )=D(P t,p )*(t,p), i=1,2,...,N.

[0029] A method for non-intrusive identification of central air-conditioning operating load according to the present invention is characterized in that the linear data fitting in step 2) is specifically a fitting optimization using a Gaussian mixture regression clustering model, as follows:

[0030] 1) Optimization objective function:

[0031]

[0032] Where: β k ,b k are the slope and intercept terms of the kth (k=1,2,3) fitting curve;

[0033] π k is the proportion of the kth regression fitting class in the total categories;

[0034] σ kis the standard deviation of the error distribution of the kth fitted curve;

[0035] The number of samples is N, and (t i ,p i );

[0036] is the probability density function of the normal distribution, which is as follows:

[0037]

[0038] Among them, e is the base of natural logarithm, and its value is 2.718;

[0039] 2) Initialization parameters:

[0040] β k =rand(1);b k =rand(1);

[0041] (k=1,2,3), where rand(1) is a random number (0,1) with a stop error limit of η=10 -5

[0042] 3) E-step expectation maximization:

[0043] Calculate the probability p of the i-th (i=1,2,…,N) sample to the k-th (k=1,2,3) fitting curve ik , the ratio π k ,as follows:

[0044]

[0045] 4) Calculate the regression coefficient β of the new fitted regression curve k ,b k , and calculate the standard deviation of the regression fit Among them, the new β k ,b k pass Fitting regression solution. The details are as follows:

[0046]

[0047] 5) Determine the new β k ,b k With the previous step β k ,b k When the absolute value of the difference between the two is greater than the shutdown error limit η=10^(-5), continue to repeat steps 7.3) and 7.4). If not, stop the loop and output the final fitting regression curve p=β k t+bk ,k=1,2,3.

[0048] A method for non-intrusive identification of central air-conditioning operating load according to the present invention is characterized in that the sliding window bilateral CUSUM algorithm in step 3) is specifically:

[0049] Assume that the power characteristic time series is The mean calculation window and transient detection window are W m and W d , the window length is defined as m, n, where currently m = n = 128, and the power averages in the two windows are M m and M d , the calculation methods are as follows:

[0050]

[0051] Where k is the first sampling point of the mean calculation window.

[0052] According to a method for non-intrusive identification of central air-conditioning operating load of the present invention, it is characterized in that the identification method of "determining whether the central air-conditioning is in the start / stop state" in step 3) is specifically:

[0053] 1) Identify the air conditioner startup status and define the startup event accumulation and The calculation formula is:

[0054]

[0055] Where β is the fluctuation level when the power is stable. d >M m +3*β, it is determined to be a suspected device startup event. Starts to accumulate from 0. If after 5 minutes The value is 1.2 to 3 times the steady-state power of the user's central air-conditioning, and it is determined that the user's central air-conditioning equipment is started;

[0056] 2) Identify the air conditioner shutdown state and define the cumulative and The calculation formula is:

[0057]

[0058] Where β is the fluctuation level when the power is stable. d <M m +3*β, it is determined to be a suspected equipment downtime event. At this time, Start from 0, if 1 minute later The value is 0.8 to 1.2 times of the user's central air-conditioning steady-state power, and it is determined that the user's central air-conditioning equipment is shut down.

[0059] A method suitable for non-intrusive identification of central air-conditioning operating load according to the present invention is characterized in that the step 3) of "using steady-state power integration to obtain the power consumption of the central air-conditioning equipment during operation" is specifically, during the user's air-conditioning startup event and shutdown event, using sub-item power estimation, and using the steady-state power integral ∫p·dt to obtain the power consumption of the central air-conditioning equipment during operation.

[0060] The following beneficial effects are obtained by using the non-intrusive identification method for central air-conditioning operating load of the present invention:

[0061] 1. The present invention is a method for non-intrusive identification of central air-conditioning operating load, which provides a new feature extraction dimension and feature extraction and identification method, can realize the integrity of the load identification model, and can complete the parameter training of "one household one policy" without home survey and wave recording;

[0062] 2. The present invention provides a method for non-intrusive identification of central air-conditioning operating load, which solves the problems of poor generalization of load identification models and large and difficult-to-obtain training data requirements for equipment to be identified. It can effectively improve the versatility of the identification model, and has accurate and efficient identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A typical central air conditioning equipment power curve diagram for a method of non-intrusive identification of central air conditioning operating load according to the present invention;

[0064] Figure 2 A typical central air conditioning equipment power and air temperature relationship diagram for a method for non-intrusive identification of central air conditioning operating load according to the present invention. DETAILED DESCRIPTION

[0065] The following is a further description of the technical means, creative features, objectives and effects achieved by a method for non-intrusive identification of central air-conditioning operating load of the present invention in conjunction with the accompanying drawings and embodiments.

[0066] A method for non-intrusive identification of central air-conditioning operating load, the specific steps are as follows:

[0067] 1) First, we selected central air conditioners of different brands, recorded the electrical waveforms of their complete operation process, and performed time-frequency domain feature analysis to extract the common features existing in the operation process of air conditioners of different brands, namely, the startup peak value P f , steady-state power value P s , shutdown power conversion P e , downtime Te and current distortion rate THD, and form a typical air-conditioning operation power curve; in this step, central air-conditioners of different brands are selected, the number of air-conditioners is 5 or more, and the number of electrical waveforms recorded during the complete operation process is 1 or more;

[0068] 2) In this step, for different users and different central air conditioning models, the total daily electricity load data and daily temperature data of the users in the past year are selected, and the correlation between the daily electricity load curve and the daily temperature curve of the users is analyzed by region, specifically:

[0069] Based on the 96-point daily electricity load curve data collected by users and daily temperature data, the load aggregation density D(P t,p ),as follows:

[0070]

[0071] Where: N(P t,p ) is the number of times the temperature is t and the power is p in the daily load curve data of one year;

[0072] t i is the range of temperature variation within a year;

[0073] p k is the range of power variation within a year;

[0074] According to the actual situation, the heating load area, basic load area and cooling load area are divided, the data of different load areas are linearly fitted with the temperature, and the load regression curves of these three areas are obtained by regression clustering:

[0075] p=β k t+b k ,k=1,2,3

[0076] Where: p is power, t is temperature, β k ,b k are the slope and intercept of the kth (k=1, 2, 3) fitting curve respectively;

[0077] The load regression curve is divided into a cooling load curve LC = F C (T), basic load curve LB = F B (T) and heating load LH = F H (T), there are three curves in total. By subtracting the cooling load curve from the basic load curve, the relationship formula between the steady-state power and temperature of the central air conditioner for different users can be obtained;

[0078] 3) Finally, the load identification module updates the identification program according to the obtained steady-state power parameters, and calculates the power characteristic value through the original high-frequency voltage and current data. The obtained power characteristic data is further calculated using the sliding window bilateral CUSUM algorithm to determine whether the central air conditioner is in the start / stop state. During the user's air conditioning start-up event and shutdown event, the sub-item power estimation is used, and the steady-state power integration is used to obtain the power consumption of the central air-conditioning equipment during operation, and finally the operation status tracking and power consumption estimation of the central air conditioner are obtained.

[0079] The typical air conditioning operating power curve in step 1) has the following values: startup peak value P f The maximum power value reached during the air conditioner startup process, the steady-state power value P s is the average power value during stable operation of the air conditioner, and the shutdown power is P e is the total power change during the air conditioner shutdown process, and the shutdown time T e It is the overall duration of the air conditioner shutdown process, and the current distortion rate THD is the effective value of all non-fundamental component currents - including harmonics and interharmonics - divided by the effective value of the current fundamental wave 50Hz.

[0080] Start peak P f Select 1.2 to 3 times the steady-state power value P s , and the shutdown power becomes P e Then select 0.8-~1.2 times the steady-state power value P s .

[0081] In step 2), "adopting a regression clustering method" specifically adopts a Gaussian mixture regression clustering model, sets the number of clusters to 3, and simultaneously calculates a fitted load regression curve.

[0082] Step 2) includes data preprocessing, which is specifically:

[0083] Set the load aggregation density threshold e1 = 0.6. When the load aggregation density D(P t,p )<=e1, the corresponding sample data point (t,p) is eliminated.

[0084] Step 2) includes weighting the sample points after elimination, which is specifically:

[0085] Weight the N remaining sample points (t, p), and the weighted sample points (t i ,p i )=D(P t,p )*(t,p), i=1,2,...,N.

[0086] The linear data fitting of step 2) is specifically the fitting optimization using the Gaussian mixture regression clustering model, as follows:

[0087] 1) Optimization objective function:

[0088]

[0089] Where: β k ,b k are the slope and intercept terms of the kth (k=1,2,3) fitting curve;

[0090] π k is the proportion of the kth regression fitting class in the total categories;

[0091] σ k is the standard deviation of the error distribution of the kth fitted curve;

[0092] The number of samples is N, and (t i ,p i );

[0093] is the probability density function of the normal distribution, which is as follows:

[0094]

[0095] Among them, e is the base of natural logarithm, and its value is 2.718;

[0096] 2) Initialization parameters:

[0097] β k =rand(1);b k =rand(1);

[0098] (k=1,2,3), where rand(1) is a random number (0,1) with a stop error limit of η=10 -5

[0099] 3) E-step expectation maximization:

[0100] Calculate the probability p of the i-th (i=1,2,…,N) sample to the k-th (k=1,2,3) fitting curve ik , the ratio π k ,as follows:

[0101]

[0102] 4) Calculate the regression coefficient β of the new fitted regression curve k ,b k , and calculate the standard deviation of the regression fit Among them, the new β k ,b k pass Fitting regression solution. The details are as follows:

[0103]

[0104] 5) Determine the new β k ,b k With the previous step β k ,b k When the absolute value of the difference between the two is greater than the shutdown error limit η=10^(-5), continue to repeat steps 7.3) and 7.4). If not, stop the loop and output the final fitting regression curve p=β k t+b k ,k=1,2,3.

[0105] 8. The method for non-intrusive identification of central air-conditioning operating load according to claim 1, characterized in that the sliding window bilateral CUSUM algorithm in step 3) is specifically:

[0106] Assume that the power characteristic time series is The mean calculation window and transient detection window are W m and W d , the window length is defined as m, n, where currently m = n = 128, and the power averages in the two windows are M m and M d , the calculation methods are as follows:

[0107]

[0108] Where k is the first sampling point of the mean calculation window.

[0109] 9. A method for non-intrusive identification of central air conditioner operating load as claimed in claim 1, characterized in that the identification method of "determining whether the central air conditioner is in a start / stop state" in step 3) is specifically:

[0110] 9.1) Identify the air conditioner startup status and define the startup event accumulation and The calculation formula is:

[0111]

[0112] Where β is the fluctuation level when the power is stable. d >M m +3*β, it is determined to be a suspected device startup event. Starts to accumulate from 0. If after 5 minutes The value is 1.2 to 3 times the steady-state power of the user's central air-conditioning, and it is determined that the user's central air-conditioning equipment is started;

[0113] 9.2) Identify the air conditioner shutdown state and define the cumulative sum of the start events The calculation formula is:

[0114]

[0115] Where β is the fluctuation level when the power is stable. d <M m +3*β, it is determined to be a suspected equipment downtime event. At this time, Start from 0, if 1 minute later The value is 0.8 to 1.2 times of the user's central air-conditioning steady-state power, and it is determined that the user's central air-conditioning equipment is shut down.

[0116] 10. A method for non-intrusive identification of central air-conditioning operating load as described in claim 1, characterized in that the "using steady-state power integration to obtain the power consumption of the central air-conditioning equipment during operation" in the step 3) is specifically to use sub-item power estimation during user air-conditioning startup events and shutdown events, and use the steady-state power integral ∫p·dt to obtain the power consumption of the central air-conditioning equipment during operation.

[0117] Example

[0118] According to the above steps, the data of a certain user from June 11 to July 7 was used to verify the load identification results. The result data (as shown in Table 1 below) was compared with the data of the rail meter installed on the air conditioner side. The analysis showed that the actual start and stop times of the user's air conditioning equipment were 23 times, the load identification model identified the start and stop times 23 times, the load type identification accuracy was 100%, and the load state identification accuracy was 100%. During the operation of the air conditioner, the rail meter measured 9036kW·h of electricity, and the algorithm estimated 9282kW·h of electricity. The electricity estimation deviation was 3%. It can be seen that the identification model of the method for non-intrusive identification of central air conditioning operation load of the present invention is highly versatile, and the identification is accurate, efficient and has small errors.

[0119]

[0120] Table 1-Comparison of central air conditioning identification and verification results for a certain user

[0121] The present invention provides a method for non-intrusive identification of central air-conditioning operating load, provides a new feature extraction dimension and feature extraction and identification method, can realize the integrity of the load identification model, and can complete the parameter training of "one household, one policy" without home surveys and wave recording; and the present invention solves the problems of poor generalization of the load identification model and large demand for and difficult to obtain training data for the equipment to be identified, can effectively improve the versatility of the identification model, and has accurate identification and high efficiency.

[0122] The present invention is applicable to the field of power load monitoring and power consumption analysis of central air-conditioning operation load.

[0123] In addition, those skilled in the art should recognize that the above embodiments are only used to illustrate the present application, and are not used to limit the present application. As long as they are within the spirit of the present application, any changes or modifications to the above embodiments will fall within the scope of the claims of the present application.

Claims

1. A method for non-intrusive identification of central air-conditioning operating load, the specific steps of which are as follows: 1) First, we selected central air conditioners of different brands, recorded the electrical waveforms of their complete operation process, and performed time-frequency domain feature analysis to extract the common features existing in the operation process of air conditioners of different brands, namely, the startup peak value P f , Steady-state power value P s , shutdown power conversion P e , downtime T e and current distortion rate THD, and form a typical air-conditioning operation power curve; in this step, central air-conditioners of different brands are selected, the number of air-conditioners is 5 or more, and the number of electrical waveforms recorded during the complete operation process is 1 or more; 2) In this step, for different users and different central air conditioning models, the total daily electricity load data and daily temperature data of the users in the past year are selected, and the correlation between the daily electricity load curve and the daily temperature curve of the users is analyzed by region, specifically: Based on the 96-point daily electricity load curve data collected by users and daily temperature data, the load aggregation density D(P t,p ),as follows: Where: N(P t,p ) is the number of times the temperature is t and the power is p in the daily load curve data of one year; t i is the range of temperature variation within a year; p k is the range of power variation within a year; According to the actual situation, the heating load area, basic load area and cooling load area are divided, the data of different load areas are linearly fitted with the temperature, and the load regression curves of these three areas are obtained by regression clustering: p6β k t+b k ,kS1,2,3 Where: p is power, t is temperature, β k ,b k are the slope and intercept of the kth (k=1, 2, 3) fitting curve respectively; The load regression curve is divided into a cooling load curve LC = F C (T), basic load curve LB = F B (T) and heating load LH = F H (T), there are three curves in total. By subtracting the cooling load curve from the basic load curve, the relationship formula between the steady-state power and temperature of the central air conditioner for different users can be obtained; 3) Finally, the load identification module updates the identification program according to the obtained steady-state power parameters, and calculates the power characteristic value through the original high-frequency voltage and current data. The obtained power characteristic data is further calculated using the sliding window bilateral CUSUM algorithm to determine whether the central air conditioner is in the start / stop state. During the user's air conditioning start-up event and shutdown event, the sub-item power estimation is used, and the steady-state power integration is used to obtain the power consumption of the central air-conditioning equipment during operation, and finally the operation status tracking and power consumption estimation of the central air conditioner are obtained.

2. A method for non-intrusive identification of central air-conditioning operating load as claimed in claim 1, characterized in that: The typical air conditioning operation power curve in step 1) has the following values: startup peak value P f The maximum power value reached during the air conditioner startup process, the steady-state power value P s is the average power value during stable operation of the air conditioner, and the shutdown power is P e is the total power change during air conditioning shutdown, and the shutdown time T e It is the overall duration of the air conditioner shutdown process, and the current distortion rate THD is the effective value of all non-fundamental component currents - including harmonics and interharmonics divided by the effective value of the current fundamental wave 50Hz.

3. A method for non-intrusive identification of central air-conditioning operating load as claimed in claim 1, characterized in that: The starting peak value P f Select 1.2 to 3 times the steady-state power value P s , and the shutdown power becomes P e Then select 0.8-~1.2 times the steady-state power value P s .

4. A method for non-intrusive identification of central air-conditioning operating load as claimed in claim 1, characterized in that: The "adopting a regression clustering method" in step 2) specifically adopts a Gaussian mixture regression clustering model, sets the number of clusters to 3, and simultaneously calculates a fitted load regression curve.

5. A method for non-intrusive identification of central air-conditioning operating load as claimed in claim 1, characterized in that: The step 2) includes data preprocessing, which is specifically: Set the load aggregation density threshold e1 = 0.

6. When the load aggregation density D(P t,p )<=e1, the corresponding sample data point (t,p) is eliminated.

6. A method for non-intrusive identification of central air-conditioning operating load as claimed in claim 1, characterized in that: The step 2) includes weighting the sample points after elimination, which is specifically: Weight the N remaining sample points (t, p), and the weighted sample points (t i ,p i )=D(P t,p )*(t,p), i=1,2,...,N.

7. A method for non-intrusive identification of central air-conditioning operating load as claimed in claim 1, characterized in that: The linear data fitting of step 2) is specifically a fitting optimization using a Gaussian mixture regression clustering model, as follows: 7.1) Optimization objective function: Where: β k ,b k are the slope and intercept terms of the kth (k=1,2,3) fitting curve; π k is the proportion of the kth regression fitting class in the total categories; σ k is the standard deviation of the error distribution of the kth fitted curve; The number of samples is N, which is (t i ,p i ); is the probability density function of the normal distribution, which is as follows: Among them, e is the base of natural logarithm, and its value is 2.718; 7.2) Initialization parameters: β k =rand(1);b k =rand(1); (k=1,2,3), where rand(1) is a random number generated from (0,1). Stop error limit η=10 -5 7.3) E-step expectation maximization: Calculate the probability p of the i-th (i=1,2,…,N) sample to the k-th (k=1,2,3) fitting curve ik , the ratio π k ,as follows: 7.4) Calculate the regression coefficient β of the new fitted regression curve k ,b k , and calculate the standard deviation of the regression fit Among them, the new β k ,b k pass Fitting regression solution. The details are as follows: 7.5) Determine the new β k ,b k With the previous step β k ,b k When the absolute value of the difference between the two is greater than the shutdown error limit η=10^(-5), continue to repeat steps 7.3) and 7.4). If not, stop the loop and output the final fitting regression curve p=β k t+b k ,k=1,2,3.

8. A method for non-intrusive identification of central air-conditioning operating load as claimed in claim 1, characterized in that: The sliding window bilateral CUSUM algorithm of step 3) is specifically as follows: Assume that the power characteristic time series is The mean calculation window and transient detection window are W m and W d , the window length is defined as m, n, where currently m = n = 128, and the power averages in the two windows are M m and M d , the calculation methods are as follows: Where k is the first sampling point of the mean calculation window.

9. A method for non-intrusive identification of central air-conditioning operating load as claimed in claim 1, characterized in that: The identification method of "determining whether the central air conditioner is in the start / stop state" in step 3) is specifically: 9.1) Identify the air conditioner startup status and define the startup event accumulation and The calculation formula is: Where β is the fluctuation level when the power is stable. d >M m +3*β, it is determined to be a suspected device startup event. Start from 0, if after 5 minutes The value is 1.2 to 3 times the steady-state power of the user's central air-conditioning, and it is determined that the user's central air-conditioning equipment is started; 9.2) Identify the air conditioner shutdown state and define the cumulative sum of the start events The calculation formula is: Where β is the fluctuation level when the power is stable. d <M m +3*β, it is determined to be a suspected equipment downtime event. At this time, Start from 0, if 1 minute later The value is 0.8 to 1.2 times of the user's central air-conditioning steady-state power, and it is determined that the user's central air-conditioning equipment is shut down.

10. A method for non-intrusive identification of central air-conditioning operating load as claimed in claim 1, characterized in that: The step 3) of "using steady-state power integration to obtain the power consumption of the central air-conditioning equipment during operation" specifically includes, during the user's air-conditioning startup event and shutdown event, using sub-item power estimation, and using the steady-state power integration ∫p·dt to obtain the power consumption of the central air-conditioning equipment during operation.

Images