An Online Modeling Method for the Load Voltage-Power Coupling Characteristic Model

Abstract

The present invention discloses an online modeling method for a load voltage-power coupling characteristic model, comprising the following steps: first, designing two regulation characteristic model structures for steady-state and transient control applications; then, when the model parameters need to be updated, using a voltage regulating device to apply a voltage disturbance signal with a small amplitude in the time domain and a wide spectrum in the frequency domain to the feeder, thereby online stimulating the voltage-power coupling characteristics of the feeder load and obtaining valid data of the load voltage-power coupling characteristics in real time; finally, based on the experimental data, using an augmented least squares algorithm to identify the model parameters. The present invention expands the feeder load voltage-power coupling characteristics from traditional static loads to comprehensive loads, constructing a dual-model system of first-order and third-order feeder load voltage-power coupling characteristics, and simultaneously adopting a feeder load voltage-power regulation characteristic model parameter identification method based on a generalized least squares algorithm to achieve comprehensive identification of the feeder load voltage-power regulation characteristics.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric load modeling, and in particular relates to an online modeling method of a load voltage-power coupling characteristic model. Background Art

[0002] The power consumption of an electric load (including both active and reactive power) varies with the load's operating voltage. The load voltage-power coupling characteristic is often used to quantitatively describe the load's power response to the operating voltage. This characteristic is a key input parameter for operational control, such as power system flow calculations, voltage reduction and energy-saving technologies, and power system stability analysis. Understanding this characteristic is crucial.

[0003] Power loads are typically connected to the power grid via low-voltage 220V / 380V levels. These loads are numerous and dispersed, making it difficult and impractical to individually characterize the voltage-power coupling characteristics of each load. Feeders are crucial components of distribution networks and fundamental units of grid operation. Studying the voltage-power coupling characteristics of feeder loads, using feeders as the demarcation boundary, better reflects actual operational needs. However, feeders contain numerous loads, and the voltage-power coupling characteristics of different loads vary. The voltage-power coupling characteristics of the same load also vary with operating conditions, resulting in significant time-varying characteristics for feeder loads. Furthermore, the presence of colored noise caused by random power variations in the loads poses significant challenges for online and accurate identification of feeder load voltage-power coupling characteristics. Furthermore, different control applications have varying requirements for load voltage-power models, making the design of diverse voltage-power coupling model structures for these loads a pressing technical challenge.

[0004] Chinese patent publication number CN 115133538 A discloses a method for predictive control of the voltage model of important loads in a distribution network. The method includes: rapidly estimating the grid state based on the weighted least squares method, obtaining the grid state, and extracting the voltage control sensitivity online; establishing a linearized grid voltage prediction model based on the voltage control sensitivity, combining it with renewable energy forecast information to predict grid voltage at multiple moments in the future, and constructing a rolling optimization model with the goal of minimizing the control deviation between the voltage of important loads and the reference voltage, as well as the equipment adjustment cost. The patent primarily addresses the issue of over-limit and drastic voltage fluctuations in important loads caused by a high proportion of renewable energy connected to the grid; however, it does not consider the issues of large numbers and dispersed distribution of other types of electricity users connected to the grid. Summary of the Invention

[0005] To address the shortcomings of the prior art, the present invention provides an online modeling method for a load voltage-power coupling characteristic model. This modeling method enables online, accurate, and comprehensive modeling of the feeder load voltage-power coupling characteristic, and enables online identification of the load voltage-power coupling characteristic model.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] An online modeling method for a load voltage-power coupling characteristic model includes the following steps:

[0008] (1) Identification of the starting load voltage-power characteristic model;

[0009] (2) Analyze application requirements and determine the load voltage-power coupling characteristic model structure. If the transient regulation process of the load power needs to be considered, execute step (3); otherwise, execute step (4);

[0010] (3) Select a third-order discrete transfer function model as the feeder load voltage-power coupling characteristic model. If the time constant of the model is the same as the time constant of the application, execute step (5);

[0011] (4) Selecting a first-order discrete transfer function model as the feeder load voltage-power coupling characteristic model. If the time constant of the model is the same as the time constant of the application, proceed to step (5);

[0012] (5) Selecting a pseudo-random binary signal that has statistical characteristics similar to white noise and is easy to generate in engineering as the voltage disturbance signal, and optimizing and determining the parameters of the pseudo-random binary signal;

[0013] (6) Based on the application's requirements for model update frequency, use the voltage regulator to apply two rounds of voltage disturbance signals periodically or as needed, and use the measurement equipment to record the voltage and power data during the test;

[0014] (7) preprocessing the test data recorded in step (6);

[0015] (8) Using the preprocessed data and the selected model structure, the generalized least squares algorithm is used to identify the model parameters.

[0016] Furthermore, the third-order discrete transfer function model in step (3) is as follows:

[0017]

[0018] Where G(z) is the feeder load voltage-power coupling characteristic model; ΔP and ΔV are the changes in power and voltage, respectively; m1, m2, m3, n1, n2, and n3 are all model parameters.

[0019] Furthermore, the first-order discrete transfer function model in step (4) is as follows:

[0020]

[0021] Where K is the steady-state gain of the voltage-power characteristic.

[0022] Furthermore, the parameters of the pseudo-random binary signal in step (5) include the interval period Ts, the sequence length Np and the time domain amplitude a.

[0023] Furthermore, the calculation formula of the interval period Ts is:

[0024] Ts=KsTa

[0025] Where Ta is the time constant of the application, Ks is the interval period coefficient, and according to operating experience, Ks ≤ 0.5.

[0026] Furthermore, the calculation formula of the sequence length Np is:

[0027] Np=KnTP

[0028] Wherein, TP is the response time of the load voltage-power characteristic. According to the field test results, TP is 0.5-0.7s. Based on operating experience, the sequence length coefficient Kn is 1.1-1.3.

[0029] Furthermore, the time domain amplitude a is within the allowable range of the national standard, the power grid company and the voltage regulating equipment, that is, a≤AST, a≤ADSO, a≤AVT, where AST is the maximum feeder voltage offset allowed by the national standard, ADSO is the maximum feeder voltage offset allowed by the power grid company, and AVT is the maximum voltage regulation allowed by the voltage regulating equipment, where AST and ADSO are both 0.07pu, and AVT is 0.1-0.15pu.

[0030] Furthermore, the voltage regulating device in step (6) includes a transformer, a dynamic voltage restorer, and an energy storage voltage source.

[0031] Furthermore, the preprocessing method for the test data in step (7) includes bad data removal, high-frequency filtering, detrending and zero-meaning.

[0032] Power load modeling is an important foundational research area. The accuracy of power load models largely determines the accuracy of power system simulation and analysis. Furthermore, the rapid development of modern load devices and power grids presents new challenges to power load modeling. The continuous introduction of various new energy-saving electrical equipment has led to new changes in the composition of power loads. The diverse management and control methods of modern power grids have further complicated the operating conditions of these loads. During a disturbance, the overall load characteristics may undergo strong nonlinear discrete changes due to factors such as the grid's self-healing control and supply-demand interactions. In such situations, traditional power load models are no longer applicable, and new power load modeling methods must be explored.

[0033] Existing methods for power load modeling are generally divided into statistical synthesis and population identification. In the statistical synthesis method, the average characteristics of electrical equipment can be obtained experimentally or provided by the manufacturer, which is consistent with the physical characteristics of electrical equipment. Furthermore, this method uses statistical methods to form a comprehensive load model, which is consistent with the fact that the comprehensive load is composed of various electrical equipment. Therefore, the statistical synthesis method can control the error of the load model within a certain range, making it a relatively practical power load modeling method. However, due to the large number of electrical equipment at the load node, the statistical work is too arduous. Furthermore, the statistical synthesis method simplifies the determination of the average characteristics of electrical equipment and the statistical calculation of load component ratios, which inevitably affects modeling accuracy. The population identification method has two advantages in application: first, it treats the load group as a whole for identification, eliminating the need for statistical analysis and understanding of the complex internal composition of the load, making it more convenient to operate. Second, because the objective function of the population identification method is to minimize the deviation between the fitted load curve and the measured load curve, this method can accurately simulate local load characteristics. However, the overall identification method also has two shortcomings: first, due to the limitations of load disturbance data, this method can usually only simulate a certain local characteristic of the load, rather than the overall characteristic; second, like other nonlinear optimization problems, the overall identification method also inevitably has the problem of local optimal solution rather than overall optimal solution in the parameter optimization process.

[0034] Compared with the prior art, the present invention has the following positive and beneficial effects:

[0035] (1) The present invention actively applies an optimized voltage disturbance signal to fully stimulate the load voltage-power coupling characteristics without affecting the normal operation of the system. Compared with the existing non-disturbance mode of directly obtaining data, the present invention realizes the online acquisition of effective data required for the identification of the feeder load voltage-power regulation characteristics; at the same time, the present invention expands the feeder load voltage-power coupling characteristics from traditional static loads to comprehensive loads, and constructs a first-order and third-order feeder load voltage-power coupling characteristics dual model system, which can accurately describe the steady-state and transient regulation process of the feeder load power, and can realize the comprehensive identification of the feeder load voltage-power regulation characteristics.

[0036] (2) The present invention further considers the influence of colored noise caused by random fluctuations in feeder loads, and adopts a feeder load voltage-power regulation characteristic model parameter identification strategy based on a generalized least squares algorithm, which can distinguish normal power signals from noise power signals, thereby achieving accurate identification of the feeder load voltage-power regulation characteristics. Moreover, the identification method in the present invention conforms to the development trend of high observability and high controllability of the distribution network, and can be implemented using a variety of voltage regulation and measurement devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart of the online modeling method of the load voltage-power coupling characteristic model of the present invention. DETAILED DESCRIPTION

[0038] In order to make the purpose and technical solution of the present invention clearer, the present invention is further described in detail below with reference to the accompanying drawings.

[0039] The voltage-power characteristics of different loads vary significantly. Some loads operate with constant impedance, such as electric heating loads, where the load power has a quadratic relationship with the operating voltage. Some loads operate with constant current, such as refrigerators, where the load power has a linear relationship with the operating voltage. Some loads operate with constant power, such as computers, where the load power does not respond to changes in the operating voltage. Feeder load composition changes constantly, and the voltage-power characteristics of feeder loads exhibit significant time-varying characteristics. Therefore, it is necessary to identify these characteristics online to help operators efficiently control the grid.

[0040] This embodiment identifies the overall voltage-power coupling characteristics of the feeder load. A voltage disturbance signal is applied to the feeder headend using the voltage regulator in a 10kV substation. The voltage and power at the feeder headend during the disturbance are recorded in real time by a measurement device within the substation, allowing for online identification of the voltage-power coupling characteristics of the feeder load.

[0041] See attached Figure 1 As shown, an online modeling method for a load voltage-power coupling characteristic model includes the following steps:

[0042] (1) The load power-voltage coupling characteristic model developed in this embodiment is set in the substation automation operation program, and the program operator can implement a one-key start of the load power-voltage coupling characteristic model identification program;

[0043] Feeder load power-voltage coupling characteristic models are essential input data for various advanced applications. Different applications have different model requirements, and operators should flexibly determine them based on prior information. For example, advanced applications with smaller timescales, such as stability analysis, need to consider the transient regulation of feeder loads and therefore require a more precise model structure. However, advanced applications with larger timescales, such as power flow calculations, only consider the static state of feeder load power and therefore require a simpler steady-state model.

[0044] (2) When the transient regulation process of the feeder load needs to be considered, the motor is the main dynamic load type of the feeder load. Considering that the motor load generally adopts a third-order model and the discrete model is often used in actual engineering, the third-order discrete transfer function is selected as the feeder load power-voltage characteristic model; the third-order model can fully represent the nonlinear process of the feeder load. At the same time, the model parameters are limited, which not only reduces the parameter identification complexity but also enhances the model generalization ability;

[0045] (3) When the transient regulation process of the feeder load does not need to be considered, in order to simplify the model structure and help operators intuitively grasp the power-voltage coupling static gain of the feeder load, the first-order discrete transfer function model is selected as the feeder load power-voltage characteristic model, and the model gain is the power-voltage coupling static gain of the feeder load;

[0046] (4) Apply a pseudo-random binary voltage disturbance signal to stimulate the load power-voltage coupling characteristics online in order to obtain valid data required for model identification; the pseudo-random binary signal is a three-parameter signal, and the signal parameters determine the statistical characteristics of the signal. To improve the identification effect, it is necessary to optimize the interval period Ts, sequence length Np, and time domain amplitude a.

[0047] The interval period Ts should ensure that the spectrum of the disturbance signal covers the spectrum of the load voltage-power characteristic process. The calculation formula of the interval period Ts is: Ts=KsTa, where Ta is the time constant of the application program and Ks is the interval period coefficient. According to operating experience, Ks≤0.5.

[0048] The sequence length Np should ensure that the signal duration is greater than the response time of the load voltage-power characteristic. The calculation formula of the sequence length Np is: Np = KnTP, where TP is the response time of the load voltage-power characteristic. According to field test results, TP is generally 0.5 to 0.7 seconds. Based on operating experience, the sequence length coefficient Kn is 1.1 to 1.3.

[0049] The time domain amplitude a should fully stimulate the load voltage-power characteristics while not affecting the normal operation of the system. First, the amplitude a must be within the allowable range of the national standard, the power grid company, and the voltage regulating equipment, that is, a≤AST, a≤ADSO, a≤AVT, where AST is the maximum feeder voltage offset allowed by the national standard, ADSO is the maximum feeder voltage offset allowed by the power grid company, and AVT is the maximum voltage adjustment allowed by the voltage regulating equipment. Under normal circumstances, AST and ADSO are both 0.07pu, and AVT is generally 0.1-0.15pu; secondly, it should be ensured that the power change caused by the voltage disturbance is greater than the natural power change of the load, limiting the minimum signal-to-noise ratio of the power data, that is, SNR=σ Y / σ N ≥KSNR, where σ Y is the standard deviation of power variation caused by voltage disturbance, σ N The standard deviation of the load's natural power variation; when the load voltage-power characteristic is considered in an approximately linear manner, σ Y =Kσ X , where K is the voltage-power characteristic steady-state gain of the feeder load, σ X is the standard deviation of the random binary signal; according to the mathematical characteristics of random binary signals, σ X ≈a; In summary, a≥K SNR σ N / K, where σ N and K can be calculated from historical data;

[0050] (5) Based on the disturbance signal obtained by optimization in step (4), two rounds of voltage disturbance signals are continuously applied using various voltage regulating devices, such as transformers, dynamic voltage restorers, and energy storage voltage source control. The voltage and power data during the disturbance are recorded by the measuring device and stored in the substation automation control system;

[0051] (6) Retrieving data from the backend, selecting the voltage and power data during the second round of disturbance test in step (5), and performing data preprocessing such as bad data removal, high-frequency filtering, detrending, and zero-meaning on the data;

[0052] (7) Using the data preprocessed in step (6) and the selected model structure, the generalized least squares identification algorithm is used to identify the model parameters. Taking the third-order transfer function model as an example, the specific process of model parameter identification is introduced. Considering the influence of noise, the voltage-power coupling characteristic model of the feeder load is shown in the following equations (1)-(3):

[0053] A(z -1 )P(k)=B(z -1 )V(k)+e(k) (1)

[0054] A(z-1 )=1+n1z -1 +n2z -2 +n3z -3 (2)

[0055] B(z -1 )=m1z -1 +m2z -2 +m3z -3 (3)

[0056] Where V(k) is the feeder voltage disturbance sequence (input data); P(k) is the corresponding feeder load power sequence (output data); e(k) is the noise sequence of the feeder load power; k is the current moment; m1, m2, m3, n1, n2, and n3 are all parameters of the feeder load voltage-power coupling characteristic model;

[0057] Colored noise can be represented by white noise and a suitable noise model, as shown in formula (4). The noise model selects a third-order autoregressive model, as shown in formula (5):

[0058] e(k)=v(k) / C(z -1 ) (4)

[0059] C(z -1 )=1+c1z -1 +c2z -2 +c3z -3 (5)

[0060] Where v(k) is a white noise sequence with a mean of 0; c1, c2, and c3 are all noise model parameters;

[0061] First, the noise model C(z -1 ) The original voltage and power data are filtered as shown in Equation (6) and Equation (7) respectively. Using the filtered voltage and power data, Equation (1) is organized into Equation (8).

[0062] P f (k) = C(z -1 )P(k) (6)

[0063] V f (k) = C(z -1 )V(k) (7)

[0064] A(z -1 )P f (k) = B(z -1 )V f (k)+v(k) (8)

[0065] Among them, V f (k) and Pf (k) are the filtered voltage and power data;

[0066] Then, rewrite Equation (8) into the least squares standard form, as shown in Equation (9). At this time, the noise in the model is white noise, and the traditional least squares algorithm can be used to recursively estimate the feeder load regulation model parameters θ, as shown in Equations (10)-(12):

[0067]

[0068]

[0069] K f (k) = M f (k-1)h f T (k)[h f T (k)M f (k-1)h f T (k)+1] -1 (11)

[0070] M f (k) = M f (k-1)-K f (k)h f T (k)M f (k-1) (12)

[0071] in, h f (k)=[-P f (k-1),-P f (k-2),-P f (k-3),-V f (k-1),-V f (k-2),-V f (k-3)]T; θ = [n1,n2,n3,m1,m2,m3]T; v = [v(1),v(2),…,v(L)]T; L is the sequence length; is the estimated value of the feeder load voltage-power coupling characteristic model parameter, ε is a very small real number vector; K f (k) is the gain matrix; M f (k) is the covariance matrix, and Mf(0) = aI, a is a very large real number, I is the identity matrix; k is the current moment;

[0072] Finally, the noise model is estimated based on the actual voltage and power data and the estimated feeder load voltage-power coupling characteristic model; the noise model is rewritten into the standard least squares form, as shown in Equation (13). Because the noise in the model is white noise, the traditional least squares algorithm is continued to be used for recursive identification, as shown in Equations (14)-(16).

[0073]

[0074]

[0075] K e (k) = M e (k-1)h e T (k)[h e T (k)M e (k-1)h e T (k)+1] -1 (15)

[0076] M e (k) = M e (k-1)-K e (k)h e T (k)M e (k-1) (16)

[0077] Among them, e=[e(1),e(2),…,e(L)]T; He=[heT(1),heT(2),…,heT(L)]T; he(k)=[-e(k-1),-e(k-2),-e(k-3)]T; θe=[c1,c2,c3]T; v=[v(1),v(2),…,v(L)]T; are estimates of the noise model parameters, K e (k) is the gain matrix; M e (k) is the covariance matrix, M e (0) = 1;

[0078] In actual engineering, load power noise cannot be measured directly, so it is replaced by noise estimation, such as equations (17) and (18):

[0079]

[0080]

[0081] in, is the noise estimate; h(k) = [-P(k-1), -P(k-2), -P(k-3), -V(k-1), -V(k-2), -V(k-3)]T.

[0082] The noise model is recursively identified using equations (14)-(18), and the parameters of the feeder load voltage-power coupling characteristic model are further identified using equations (10)-(12). The feeder load voltage-power coupling characteristic model and the noise model are continuously and alternately identified until the process ends and the feeder load voltage-power coupling characteristic model is equal to the average of the parameter identification results of the last three moments.

[0083] The present invention can realize the identification of the active power-voltage coupling characteristic model and the reactive power-voltage coupling characteristic model of the feeder load through the online modeling method of the above-mentioned load voltage-power coupling characteristic model, and is not limited to the identification of the feeder load voltage-power coupling characteristics. According to the load measurement range of the measuring device, the power-voltage coupling characteristics of a single load, a substation load and the load carried by the entire substation can be identified.

[0084] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An online modeling method for a load voltage-power coupling characteristic model, characterized in that: The following steps are involved: (1) Identification of the starting load voltage-power characteristic model; (2) Analyze application requirements and determine the load voltage-power coupling characteristic model structure. If the transient regulation process of the load power needs to be considered, execute step (3); otherwise, execute step (4); (3) Select a third-order discrete transfer function model as the feeder load voltage-power coupling characteristic model. If the time constant of the model is the same as the time constant of the application, execute step (5); (4) Selecting a first-order discrete transfer function model as the feeder load voltage-power coupling characteristic model. If the time constant of the model is the same as the time constant of the application, proceed to step (5); (5) Selecting a pseudo-random binary signal that has statistical characteristics similar to white noise and is easy to generate in engineering as the voltage disturbance signal, and optimizing and determining the parameters of the pseudo-random binary signal; (6) Based on the application's requirements for model update frequency, use the voltage regulator to apply two rounds of voltage disturbance signals periodically or as needed, and use the measurement equipment to record the voltage and power data during the test; (7) preprocessing the test data recorded in step (6); (8) Using the preprocessed data and the selected model structure, the generalized least squares algorithm is used to identify the model parameters; The first-order discrete transfer function model in step (4) is as follows: Where K is the steady-state gain of the voltage-power characteristic; G(z) is the voltage-power coupling characteristic model of the feeder load; ΔP(z) and ΔV(z) are the changes in power and voltage, respectively; The parameters of the pseudo-random binary signal in step (5) include the interval period Ts, the sequence length Np and the time domain amplitude a; The calculation formula of the interval period Ts is: Ts=KsTa Where Ta is the time constant of the application, Ks is the interval period coefficient, and according to operating experience, Ks ≤ 0.

5.

2. The online modeling method of a load voltage-power coupling characteristic model according to claim 1, characterized in that: The third-order discrete transfer function model in step (3) is as follows: Where G(z) is the feeder load voltage-power coupling characteristic model; ΔP(z) and ΔV(z) are the changes in power and voltage, respectively; m1, m2, m3, n1, n2, and n3 are all model parameters.

3. The online modeling method of a load voltage-power coupling characteristic model according to claim 1, characterized in that: The calculation formula of the sequence length Np is: Np=KnTP Wherein, TP is the response time of the load voltage-power characteristic. According to the field test results, TP is 0.5-0.7s. Based on operating experience, the sequence length coefficient Kn is 1.1-1.

3.

4. The online modeling method of a load voltage-power coupling characteristic model according to claim 1, characterized in that: The time domain amplitude a is within the allowable range of the national standard, the power grid company, and the voltage regulating equipment, that is, a≤AST, a≤ADSO, a≤AVT, where AST is the maximum feeder voltage offset allowed by the national standard, ADSO is the maximum feeder voltage offset allowed by the power grid company, and AVT is the maximum voltage regulation allowed by the voltage regulating equipment. AST and ADSO are both 0.07pu, and AVT is 0.1-0.15pu.

5. The online modeling method of a load voltage-power coupling characteristic model according to claim 1, characterized in that: The voltage regulating device in step (6) includes a transformer, a dynamic voltage restorer, and an energy storage voltage source.

6. The online modeling method of a load voltage-power coupling characteristic model according to claim 1, characterized in that: The pre-processing method for the test data in step (7) includes bad data removal, high-frequency filtering, detrending and zero-meaning.

Images