A temperature control load aggregation modeling method considering bad data and privacy protection

Abstract

A temperature control load aggregation modeling method taking into account bad data and privacy protection belongs to the field of privacy computing and demand-side response technology, and solves the problems of how to improve the accuracy of temperature control load aggregation modeling and how to effectively protect user privacy information from leakage. The present invention adopts a temperature control load aggregation model, a robust parameter estimation model of a temperature control load aggregation model with bad data resistance, and a robust parameter estimation model of a temperature control load aggregation model with both bad data resistance and privacy information protection to aggregate temperature control loads for temperature control load aggregators to model the temperature control loads, so that they participate in the operation and control of energy systems, while ensuring the protection of user privacy information, enhancing the robustness of the modeling method to bad data, and having good user privacy protection characteristics and resistance to abnormal data, which can promote the power system to tap user-side flexibility resources and enhance operational economy and flexibility.

Description

Technical Field

[0001] The present invention belongs to the technical field of privacy computing and demand-side response, and relates to a temperature control load aggregation modeling method taking bad data and privacy protection into account. Background Art

[0002] With the growing share of variable renewable energy, the power system is undergoing a profound innovation. In order to ensure its safety and economic benefits, the power grid has an increasingly urgent need for flexible regulation resources. Temperature control loads are a demand response resource with great potential. The concept of using temperature control loads to serve power dispatching and management has attracted great attention in academic research and industrial practice.

[0003] However, there are still challenges in converting the elasticity of temperature control loads into practical solutions. The first difficulty is that in the face of a large number of temperature control loads, the information interaction between the energy system and the massive user temperature control loads will generate a huge computing and communication load; secondly, the current methods for tapping the potential of temperature control loads lack effective privacy protection means and mechanisms, and there is a risk of leaking user privacy; in addition, the decline in data quality caused by sensor aging, damage and communication barriers may affect the accuracy of temperature control load modeling.

[0004] In the prior art, the patent application document "A privacy-protected building thermal dynamic aggregation model construction method" with a publication date of January 12, 2024 and a publication number of CN117390893A has the following disadvantages: 1) The least squares estimation method is used to achieve parameter estimation for the building load aggregator model, and bad data has a great impact on the accuracy of parameter estimation and is not robust; 2) It is necessary to fix the parameters in the parameter estimation model separately. and , divide it into two sub-problems, implement a customized privacy protection scheme for each sub-problem, and then iteratively solve the two encrypted sub-problems respectively; therefore, in each iteration, it is necessary to implement the privacy protection scheme for the two sub-problems separately, and the amount of computation and communication is very large. Summary of the invention

[0005] The technical solution of the present invention is used to solve the problems of how to improve the accuracy of temperature control load aggregation modeling and how to effectively protect user privacy information from leakage.

[0006] The present invention solves the above technical problems through the following technical solutions:

[0007] A temperature control load aggregation modeling method considering bad data and privacy protection is applied to a temperature control load aggregator-temperature control load system, including the following steps:

[0008] Step 1: Establish a temperature control load aggregation model, as follows:

[0009] Step 1.1, establish a single-area temperature control load model;

[0010] Step 1.2, establish the aggregation equation;

[0011] Step 1.3, establish a temperature control load aggregation model;

[0012] Step 2: Establish a robust parameter estimation model for the temperature control load aggregation model that is resistant to bad data, as follows:

[0013] Step 2.1, establish the measurement equation of the temperature control load aggregation model;

[0014] Step 2.2, establish a least square parameter estimation model for the temperature control load aggregation model;

[0015] Step 2.3: The introduced Huber loss function is used to improve the least square parameter estimation model of the temperature control load aggregation model in step 2.2, and then a robust parameter estimation model of the temperature control load aggregation model is established;

[0016] Step 3: Establish a robust parameter estimation model for the temperature control load aggregation model that is both resistant to bad data and privacy information, as follows:

[0017] Step 3.1, perform equivalent deformation of the model;

[0018] Step 3.2: Establish a privacy protection mechanism;

[0019] Step 3.3: Solve the privacy-preserving robust parameter estimation model.

[0020] Furthermore, the single-area temperature control load model is specifically:

[0021] in, represents the indoor temperature at time t, Indicates the indoor temperature at time tm, Indicates the thermal power at time tm, Indicates the outdoor temperature at time tm, represents the solar radiation power at time tm, Describe the role of user activity at time t, These are the parameters of the single-area temperature control load model. Characterizes the model order set.

[0022] Furthermore, the polymerization equation is specifically:

[0023]

[0024] in, is the multi-zone aggregate indoor temperature at time t, For Region The aggregation coefficient, represents a set of regions, Indicates area The room temperature at time t.

[0025] Furthermore, the method for establishing the temperature control load aggregation model is as follows:

[0026] Combining the single-area temperature control load model established in step 1.1 with the aggregation equation established in step 1.2, the temperature control load aggregation model is obtained, as follows:

[0027]

[0028] in, is the multi-zone aggregate indoor temperature at time tm, Indicates area Thermal power at time tm, Describe the role of multi-region user activities at time t, These are all parameters of the multi-zone temperature control load aggregation model.

[0029] Furthermore, the measurement equation for establishing the temperature control load aggregation model is as follows:

[0030] in, represents independent and identically distributed Gaussian noise.

[0031] Furthermore, the method for establishing the least squares parameter estimation model of the temperature control load aggregation model is specifically as follows:

[0032] First define , , , , , , , , , ;

[0033] Based on the measurement equation of the temperature control load aggregation model in step 2.1, the residual vector is obtained:

[0034] in, , the constant matrices c1-c3 are , , , is a K-dimensional 1 vector;

[0035] Then the least squares parameter estimation method is applied, and the regular term is introduced to solve the sparsity problem of the aggregation coefficient, and the least squares parameter estimation model of the temperature control load aggregation model is established as follows:

[0036] in, is the penalty factor.

[0037] Furthermore, the Huber loss function is as follows:

[0038]

[0039] in, is the threshold of the Huber loss function, is the scale estimate of the residual; The value is set to ,in To calculate the median of a data set, Residual vector The median of Adjustment coefficients for estimating scales of residual parameters;

[0040] The robust parameter estimation model for establishing the temperature control load aggregation model is as follows:

[0041]

[0042] in, is the tth standard column vector, and its expression is .

[0043] Furthermore, the method for performing equivalent deformation of the model is specifically as follows:

[0044] Introducing reversible random matrices The robust parameter estimation model of the temperature control load aggregation model in step 2.3 is equivalently transformed:

[0045]

[0046] The reversible random matrix The definition is as follows:

[0047] Therefore, the robust parameter estimation model of the temperature control load aggregation model in step 2.3 is equivalently transformed into the following form:

[0048] Among them, the residual vector It is expressed as follows:

[0049] .

[0050] Furthermore, the specific methods for establishing a privacy protection mechanism are as follows:

[0051] The information required by the temperature control load aggregator to solve the model in step 3.1 is ,remember and They are and No. Column, where By region Generated and privately owned, defined , as well as ,therefore Respectively expressed in the form of and:

[0052]

[0053] Based on the secure aggregation protocol, the process of designing a privacy-preserving computing method is as follows:

[0054] (1) Region Generate random matrix It also exchanges information with other regions and performs the following calculations based on local information and interaction information with other regions:

[0055]

[0056]

[0057]

[0058] (2) Region Upload to temperature-controlled load aggregators;

[0059] (3) Temperature control load aggregator calculation matrix :

[0060]

[0061]

[0062]

[0063] Furthermore, the method for solving the privacy-preserving robust parameter estimation model is specifically as follows:

[0064] The parameter set is , the temperature control load aggregator solves the privacy-preserving robust parameter estimation model through iterative reweighted least squares method. The process is as follows:

[0065] (1) Initialization iteration: , set the iteration convergence threshold ;

[0066] (2) Calculate the least squares parameter estimation model of the temperature control load aggregation model and obtain the initial parameters ;

[0067] (3) Based on the estimated scale of the residual in step 2.3 The formula is the same as the residual vector calculation formula in step 3.1 ;

[0068] (4) For ,like ,Other Otherwise, another ;

[0069] (5) Solution , get the parameters ;

[0070] (6) Based on the estimated scale of the residual in step 2.3 The formula is the same as the residual vector calculation formula in step 3.1 ;

[0071] (7) If , the algorithm terminates; otherwise, , and re-execute steps (4) to (7).

[0072] The advantages of the present invention are:

[0073] The present invention adopts a temperature control load aggregation model, a robust parameter estimation model of a temperature control load aggregation model with resistance to bad data, and a robust parameter estimation model of a temperature control load aggregation model with both resistance to bad data and protection of privacy information to aggregate temperature control loads for temperature control load aggregators to model the temperature control loads, so that they participate in the operation and control of the energy system while ensuring the protection of user privacy information, thereby enhancing the robustness of the modeling method to bad data, having good user privacy protection characteristics and resistance to abnormal data, and can promote the power system to tap user-side flexibility resources and enhance operational economy and flexibility; the present invention has high modeling accuracy, good prediction effect, good resistance to bad data, and can effectively protect user privacy information from leakage. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 is a flow chart of a temperature control load aggregation modeling method taking bad data and privacy protection into account according to the present invention;

[0075] Figure 2 It is a temperature control load aggregator-temperature control load system structure diagram of the present invention;

[0076] Figure 3 It is a comparison chart of the predicted value and the actual value of the aggregation state of the temperature control load aggregation model when the outlier ratio is set to 5% in the temperature control load aggregation modeling method taking into account bad data and privacy protection of the present invention;

[0077] Figure 4 This is a comparison chart of the encrypted indoor temperature-related information in region 1 of the temperature control load aggregation modeling method of the present invention that takes into account bad data and privacy protection. DETAILED DESCRIPTION

[0078] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in combination with the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0079] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and specific embodiments:

[0080] Embodiment 1

[0081] A temperature control load aggregation modeling method taking into account bad data and privacy protection in this embodiment is applied to a temperature control load aggregator-temperature control load system, including the following steps:

[0082] Step 1: Establish a temperature control load aggregation model.

[0083] Step 1.1: Establish a single-area temperature control load model, as follows:

[0084] in, represents the indoor temperature at time t, Indicates the indoor temperature at time tm, Indicates the thermal power at time tm, Indicates the outdoor temperature at time tm, represents the solar radiation power at time tm, Describe the role of user activity at time t, These are the parameters of the single-area temperature control load model. Characterizes the model order set.

[0085] Step 1.2: Establish the aggregation equation as follows:

[0086]

[0087] in, is the multi-zone aggregate indoor temperature at time t, For Region The aggregation coefficient, represents a set of regions, Indicates area The room temperature at time t.

[0088] Step 1.3: Establish a temperature control load aggregation model.

[0089] Combining the single-area temperature control load model established in step 1.1 with the aggregation equation established in step 1.2, the temperature control load aggregation model is obtained, as follows:

[0090]

[0091] in, is the multi-zone aggregate indoor temperature at time tm, Indicates area Thermal power at time tm, Describe the role of multi-region user activities at time t, These are all parameters of the multi-zone temperature control load aggregation model.

[0092] Step 2: Establish a robust parameter estimation model for the temperature control load aggregation model that is resistant to bad data.

[0093] Step 2.1: Establish the temperature control load aggregation model measurement equation, as follows:

[0094] in, represents independent and identically distributed Gaussian noise.

[0095] Step 2.2: Establish the least squares parameter estimation model of the temperature control load aggregation model, as follows:

[0096] The method for establishing the least squares parameter estimation model of the temperature control load aggregation model is specifically as follows:

[0097] First define , , , , , , , , , ;

[0098] Based on the measurement equation of the temperature control load aggregation model in step 2.1, the residual vector is obtained:

[0099] in, , the constant matrices c1-c3 are , , , is a K-dimensional 1 vector;

[0100] Next, the least squares parameter estimation method is applied, and the regularization term is introduced to solve the sparsity problem of the aggregation coefficient, and the least squares parameter estimation model of the temperature control load aggregation model is established as follows:

[0101] in, is the penalty factor.

[0102] Step 2.3: The introduced Huber loss function improves the least squares parameter estimation model of the temperature control load aggregation model in step 2.2, and then establishes a robust parameter estimation model for the temperature control load aggregation model.

[0103] The Huber loss function is introduced as follows:

[0104]

[0105] in, is the threshold of the Huber loss function, is the scale estimate of the residual; The value is set to ,in To calculate the median of a data set, Residual vector the median of Adjustment coefficients for estimating scales of residual parameters;

[0106] The robust parameter estimation model for establishing the temperature control load aggregation model is as follows:

[0107] in, is the tth standard column vector, and its expression is .

[0108] Step 3: Establish a robust parameter estimation model for the temperature control load aggregation model that is both resistant to bad data and protects privacy information.

[0109] Step 3.1: Perform equivalent deformation of the model

[0110] Introducing reversible random matrices The robust parameter estimation model of the temperature control load aggregation model in step 2.3 is equivalently transformed:

[0111]

[0112] The reversible random matrix The definition is as follows:

[0113] Therefore, the robust parameter estimation model of the temperature control load aggregation model in step 2.3 is equivalently transformed into the following form:

[0114] Among them, the residual vector It is expressed as follows:

[0115] .

[0116] Step 3.2: Establish a privacy protection mechanism

[0117] The information required by the temperature control load aggregator to solve the model in step 3.1 is ,remember and They are and No. Column, where By region Generated and privately owned, defined , as well as ,therefore Respectively expressed in the form of and:

[0118]

[0119] Based on the secure aggregation protocol, the process of designing a privacy-preserving computing method is as follows:

[0120] (1) Region Generate random matrix It also exchanges information with other regions and performs the following calculations based on local information and interaction information with other regions:

[0121]

[0122]

[0123]

[0124] (2) Region Upload to temperature-controlled load aggregators;

[0125] (3) Temperature control load aggregator calculation matrix :

[0126]

[0127]

[0128] .

[0129] Step 3.3: Solving the privacy-preserving robust parameter estimation model

[0130] The parameter set is , the temperature control load aggregator solves the privacy-preserving robust parameter estimation model through iterative reweighted least squares method. The process is as follows:

[0131] (1) Initialization iteration: , set the iteration convergence threshold ;

[0132] (2) Calculate the least squares parameter estimation model of the temperature control load aggregation model and obtain the initial parameters ;

[0133] (3) Based on the estimated scale of the residual in step 2.3 The formula is the same as the residual vector calculation formula in step 3.1 ;

[0134] (4) For ,like ,Other Otherwise, another ;

[0135] (5) Solution , get the parameters ;

[0136] (6) Based on the estimated scale of the residual in step 2.3 The formula is the same as the residual vector calculation formula in step 3.1 ;

[0137] (7) If , the algorithm terminates; otherwise, , and re-execute steps (4) to (7).

[0138] The technical solution of the present invention establishes a robust parameter estimation model for the temperature control load aggregation model by introducing the Huber loss function as the objective function of the parameter estimation model. The method has good robustness and resistance to bad data. The privacy protection mechanism is improved by following the principle of separation of encryption and computation, that is, the privacy protection scheme only needs to be executed once in the initial stage, and the encrypted parameter estimation model is uniformly transmitted to the temperature control load aggregator for solution. In this way, the amount of calculation and communication is greatly reduced, making the current patent more valuable for practical application.

[0139] Test verification

[0140] like Figure 2 As shown in the figure, the temperature control load aggregator-temperature control load system consists of a temperature control load aggregator and seven buildings (containing 64 areas in total). The temperature control load aggregator aggregates the temperature control loads in the seven buildings. There are 1440 simulation data in total, of which 1080 data are training sets and the remaining 360 data are test sets. The simulation time interval is set to 30 minutes, the user activity cycle is 24 hours, and the random matrix obeys a normal distribution with a mean of 0.1 and a standard deviation of 0.1. The temperature control load aggregation model order M is set to 2, and the penalty coefficient Set to 100. Introduce salt and pepper noise to the original, and its noise ratio is , The probability distribution is as follows:

[0141]

[0142] set up , , the hyperparameters in the Huber loss function , adjustment coefficient ,set up ,in is a variable parameter representing the probability.

[0143] Figure 3 The method of the present invention is Comparison chart of predicted value and actual value of the aggregation state of the temperature control load aggregation model when ; Figure 4 This is a comparison of the information about the indoor temperature of area 1 before and after encryption in the method of the present invention. It can be seen that the method of the present invention has high accuracy, good prediction effect, good resistance to bad data, and can effectively protect user privacy information from leakage.

[0144] An electronic device includes a memory and a processor, wherein the memory is used to store a program that supports the processor to execute the temperature control load aggregation modeling method that takes into account bad data and privacy protection in Example 1, and the processor is configured to execute the program stored in the memory.

[0145] A storage medium stores a computer program, which, when executed by a processor, executes the steps of the temperature control load aggregation modeling method taking into account bad data and privacy protection in embodiment 1.

[0146] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A temperature control load aggregation modeling method taking into account bad data and privacy protection, applied in a temperature control load aggregator-temperature control load system, characterized in that: The following steps are involved: Step 1: Establish a temperature control load aggregation model, as follows: Step 1.1: Establish a single-area temperature control load model, specifically: in, represents the indoor temperature at time t, Indicates the indoor temperature at time tm, Indicates the thermal power at time tm, Indicates the outdoor temperature at time tm, represents the solar radiation power at time tm, Describe the role of user activity at time t, These are the parameters of the single-area temperature control load model. Characterize the set of model orders; Step 1.2: Establish the aggregation equation, specifically: in, is the multi-zone aggregate indoor temperature at time t, For Region The aggregation coefficient, represents a set of regions, Indicates area The indoor temperature at time t; Step 1.3: Establish a temperature control load aggregation model as follows: Combining the single-area temperature control load model established in step 1.1 with the aggregation equation established in step 1.2, the temperature control load aggregation model is obtained, as follows: in, is the multi-zone aggregate indoor temperature at time tm, Indicates area Thermal power at time tm, Describe the role of multi-region user activities at time t, All are parameters of the multi-region temperature control load aggregation model; Step 2: Establish a robust parameter estimation model for the temperature control load aggregation model that is resistant to bad data, as follows: Step 2.1: Establish the temperature control load aggregation model measurement equation as follows: in, represents independent and identically distributed Gaussian noise; Step 2.2: Establish the least squares parameter estimation model of the temperature control load aggregation model. The specific method is as follows: First define , , , , , , , , , ; Based on the measurement equation of the temperature control load aggregation model in step 2.1, the residual vector is obtained: in, , the constant matrices c1-c3 are , , , is a K-dimensional 1 vector; Then the least squares parameter estimation method is applied, and the regular term is introduced to solve the sparsity problem of the aggregation coefficient, and the least squares parameter estimation model of the temperature control load aggregation model is established as follows: in, is the penalty factor; Step 2.3: The introduced Huber loss function is used to improve the least square parameter estimation model of the temperature control load aggregation model in step 2.2, and then a robust parameter estimation model of the temperature control load aggregation model is established; The Huber loss function is as follows: in, is the threshold of the Huber loss function, is the scale estimate of the residual; The value is set to ,in To calculate the median of a data set, Residual vector The median of Adjustment coefficients for estimating scales of residual parameters; The robust parameter estimation model for establishing the temperature control load aggregation model is as follows: in, is the tth standard column vector, and its expression is ; Step 3: Establish a robust parameter estimation model for the temperature control load aggregation model that is both resistant to bad data and privacy information, as follows: Step 3.1: Perform equivalent deformation of the model. The specific method is as follows: Introducing reversible random matrices The robust parameter estimation model of the temperature control load aggregation model in step 2.3 is equivalently transformed: The reversible random matrix The definition is as follows: Therefore, the robust parameter estimation model of the temperature control load aggregation model in step 2.3 is equivalently transformed into the following form: Among them, the residual vector It is expressed as follows: ； Step 3.2: Establish a privacy protection mechanism as follows: The information required by the temperature control load aggregator to solve the model in step 3.1 is ,remember and They are and No. Column, where By region Generated and privately owned, defined , as well as ,therefore Respectively expressed in the form of and: Based on the secure aggregation protocol, the process of designing a privacy-preserving computing method is as follows: (1) Region Generate random matrix It also exchanges information with other regions and performs the following calculations based on local information and interaction information with other regions: (2) Region Upload to temperature-controlled load aggregators; (3) Temperature control load aggregator calculation matrix : ； Step 3.3: Solve the privacy-preserving robust parameter estimation model. The specific method is as follows: The parameter set is , the temperature control load aggregator solves the privacy-preserving robust parameter estimation model through iterative reweighted least squares method. The process is as follows: (1) Initialization iteration: , set the iteration convergence threshold ; (2) Calculate the least squares parameter estimation model of the temperature control load aggregation model and obtain the initial parameters ; (3) Based on the estimated scale of the residual in step 2.3 The formula is the same as the residual vector calculation formula in step 3.1 ; (4) For ,like ,Other Otherwise, another ; (5) Solution , get the parameters ; (6) Based on the estimated scale of the residual in step 2.3 The formula is the same as the residual vector calculation formula in step 3.1 ; (7) If , the algorithm terminates; otherwise, , and re-execute steps (4) to (7).

Images