An Optimization Method for Designing a Frequency Controller to Improve System Frequency Stability

Abstract

The present invention discloses a novel design optimization method for a frequency controller to improve the frequency stability of a system. First, based on the frequency division control theory and combined with a polynomial model, a frequency controller is designed. Then, according to the structural characteristics of the frequency controller, a parameter tuning model of the frequency controller is constructed. Based on this parameter tuning model and combined with the asynchronous advantage actor-critic algorithm, the parameter tuning model is transformed into a Markov decision process, and then iterative optimization is carried out. Finally, a self-updating strategy for the controller parameters is obtained. Furthermore, the online self-calibration of the frequency controller can be realized according to different real-time states of the system, ensuring that the frequency controller can still have a better effect of improving frequency stability in variable working condition scenarios, and guaranteeing the frequency stability of the system under complex working conditions.

Description

Technical Field

[0001] The present invention belongs to the field of power grid frequency stability control, and more specifically, relates to an optimization method for designing a frequency controller to improve system frequency stability. Background Art

[0002] Driven by the dual-carbon goal, green and efficient development will become the future direction of China's energy system construction. Building a new power system with a high proportion of renewable energy access and promoting the optimization of the energy industry structure, energy cleaning and low-carbonization have become the top priorities for future power development. Hydropower will play an important role in the process of carbon reduction in the energy field. The reason is that: as the main clean energy in China, hydropower has the advantages of environmental protection, priority grid connection and low cost. However, as the proportion of hydropower in the system gradually increases, the operating characteristics of the power grid have changed significantly. In addition, the seasonal characteristics of hydropower are obvious, with the changes in the output of hydropower during the wet and dry seasons, the uncertain fluctuations of the load, and the frequent occurrence of natural disasters on the transmission channels, making the transmission power, the operating mode and the safety and stability characteristics of the high-hydropower proportion system complex and variable, showing high-dimensional, time-varying and non-linear characteristics, with prominent transient frequency problems, a significant decline in the power grid's ability to resist power disturbances, a great threat to the safe and stable operation of the power grid, and a serious restriction on the transmission of clean energy.

[0003] Therefore, relevant scholars have carried out research on strategies to enhance the frequency stability in high-hydropower proportion regions. Among them, considering that the unreasonable setting of the primary frequency regulation parameters of the water turbine is the main cause of frequency instability in high-hydropower proportion systems, the optimization of the governor has become an important means to suppress ultra-low frequency oscillations, including: the particle swarm optimization method, the robust optimization method, and the gravitational search method have all been used for the optimal design of the water turbine governor parameters. However, suppressing ultra-low frequency oscillations based on the strategy of optimizing governor parameters will reduce the regulation characteristics of the governor, resulting in a decline in the tracking performance of the system and restricting the rapid regulation ability of the system. Therefore, how to design a new type of frequency controller to improve the system frequency stability without affecting the system regulation ability has become a key problem to be solved urgently. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies of the prior art and provide an optimization method for designing a frequency controller to improve system frequency stability. Through the parameter adaptive tuning strategy, appropriate parameter setting information can be given according to the real-time state of the system, so as to ensure that the proposed frequency controller can fully exert its control performance under different working conditions.

[0005] To achieve the above invention purpose, an optimization method for designing a frequency controller to improve system frequency stability according to the present invention is characterized by including the following steps:

[0006] (1). Design a frequency controller G(s) based on the frequency division control theory, specifically composed of a proportional-resonant control unit GPR (s) is connected in series with the polynomial model G M to form;

[0007] (2) Collect the frequency deviation curve f(t) of the generator in the power system, identify the oscillation mode of the frequency deviation curve, and obtain the main oscillation mode information: λ = α + jω, where λ is the main oscillation mode information, α and ω are the real part and imaginary part of the ultra-low frequency oscillation mode respectively, and j represents the imaginary symbol;

[0008] (3) Calculate the damping ratio ξ of the frequency controller in the ultra-low frequency oscillation mode;

[0009]

[0010] (4) Construct a parameter tuning model of the frequency controller according to the damping ratio ξ of the ultra-low frequency oscillation mode;

[0011] (5) Real-time collect the voltage signal V(t) and power signal P(t) of the power system;

[0012] (6) Use the voltage signal V(t) of the power system as the output of the frequency controller G(s), the power signal P(t) as the input of the frequency controller G(s), then combine the deep reinforcement learning algorithm to transform the parameter tuning model of the frequency controller into a Markov decision process, and then iterate the Markov decision process to continuously reduce its model value until the algorithm converges, and finally obtain the adaptive control parameters of the frequency controller.

[0013] The invention purpose of the present invention is realized as follows:

[0014] A frequency controller design optimization method for improving the system frequency stability in the present invention first designs a frequency controller based on the frequency division control theory combined with a polynomial model, and then constructs a parameter tuning model of the frequency controller according to the structural characteristics of the frequency controller; based on this parameter tuning model, combined with the asynchronous advantage actor-critic algorithm, the parameter tuning model is transformed into a Markov decision process, and then iteratively optimized, and finally a self-updating strategy of the controller parameters is obtained, and then the online self-correction of the frequency controller can be realized according to different real-time states of the system, ensuring that the frequency controller can still have a good effect of improving the frequency stability in the variable working condition scenario, and ensuring the frequency stability of the system under complex working conditions.

[0015] At the same time, a frequency controller design optimization method for improving the system frequency stability in the present invention also has the following beneficial effects:

[0016] (1) Applying the frequency division control theory to the design of the frequency stability controller can achieve the accurate extraction of specific main oscillation modes, thereby enhancing the oscillation suppression effect of the frequency controller on the main oscillation modes and contributing to further improving the frequency stability of the system.

[0017] (2) Introducing the polynomial model into the frequency controller structure has greater design freedom compared with the lead-lag link in the traditional controller. By reasonably assigning values to the polynomial model, a better controller structure can be found, achieving better control effects.

[0018] (3) Transforming the tuning model of the frequency controller parameters into a Markov decision process and solving it combined with the deep reinforcement learning algorithm can obtain the self-update strategy of the controller parameters, ensuring the frequency stability of the system under complex working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is the flowchart of the design optimization method of the frequency controller for improving the frequency stability of the present invention;

[0020] Figure 2 is the equivalent simplified topology diagram of a certain hydropower unit group system;

[0021] Figure 3 is the frequency deviation curve of the generator unit CTK-1# under the excitation of Fault 1;

[0022] Figure 4 is the distribution of ultra-low frequency oscillation modes under different strategies: (a) the proposed frequency controller is not configured; (b) the traditional parameter tuning method; (c) the proposed parameter tuning method; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] The following describes the specific embodiments of the present invention with reference to the drawings, so that those skilled in the art can better understand the present invention. It should be particularly noted that in the following description, when the detailed descriptions of known functions and designs may dilute the main content of the present invention, these descriptions will be omitted here.

[0024] Embodiment

[0025] In this embodiment, Figure 2 is the equivalent simplified topology of a certain hydropower unit group system, including 53 buses, 26 lines and 17 hydropower units. The total installed capacity of the system is 1100MW. All units adopt the fifth-order model and are equipped with an excitation system and a steam turbine regulating system, and are sent to the main grid through a 500kV substation. The load in the area can be ignored.

[0026] In this embodiment, as Figure 1As shown, we will elaborate on an optimization method for the design of a frequency controller to improve the frequency stability of the system, which specifically includes the following steps:

[0027] S1. Design the frequency controller G(s) based on the frequency division control theory.

[0028] The frequency controller G(s) is specifically composed of a proportional-resonant control unit G PR (s) and a polynomial model G M (s) in series;

[0029] S1.1. Considering that the main cause of the frequency stability problem in the system with a high proportion of hydropower is the existence of a main oscillation mode with a frequency lower than 0.1 Hz inside the system, based on the frequency division control theory, introduce the proportional-resonant control unit G PR (s):

[0030]

[0031] where K P is the proportional coefficient of the proportional-resonant control unit; K R is the resonance coefficient of the proportional-resonant control unit; ω0 is the resonant frequency of the proportional-resonant control unit; ω c is the cut-off frequency of the proportional-resonant control unit, and s represents the complex variable;

[0032] The proportional-resonant control unit can separate and extract the main oscillation mode from the system with a high proportion of hydropower.

[0033] S1.2. Considering that the lead-lag link structure adopted in the traditional controller is simple and has limited characterization ability, making it difficult to realize the expression of the optimal controller structure, introduce the polynomial model G M (s):

[0034]

[0035] where a0, a1, a2, b0, b1, b2 are the variables to be tuned in the polynomial.

[0036] S1.3. Connect the proportional-resonant control unit G PR (s) and the polynomial model G M (s) in series to obtain the frequency controller structure G(s):

[0037]

[0038] S2. Collect the frequency deviation curve f(t) of the CTHC-2# generator in the system, identify the oscillation mode of the curve, and obtain the main oscillation mode information;

[0039] λ = α + jω

[0040] Where: In the formula, α and ω are the real part and the imaginary part of the ultra-low frequency oscillation mode respectively, and j represents the imaginary symbol. ξ is the damping ratio of the ultra-low frequency oscillation mode. In this embodiment, the identification result shows that: α = -0.012, ω = 0.57, and ξ = 2.1%

[0041] S3. Calculate the damping ratio ξ of the frequency controller in the ultra-low frequency oscillation mode;

[0042]

[0043] S4. Construct a parameter tuning model for the frequency controller G(s);

[0044] S4.1. Given the bandwidth B = 0.1 of the frequency controller G(s);

[0045] S4.2. To enable the frequency controller to accurately filter out the main oscillation mode in the high hydropower proportion system, the resonant frequency of the frequency controller can be set as: ω0 = ω = 0.57, thus completing the tuning of ω0;

[0046] S4.3. Considering that the modal information identified in step S2 is vulnerable to noise and there is a certain error, therefore, the proportional resonant unit G PR (s) of the frequency controller needs to reasonably set K P and ω c to be determined so that the frequency controller has a certain bandwidth B.

[0047] Therefore, we can calculate the cut-off frequency ω c of the proportional resonant control unit according to the bandwidth B of the frequency controller G(s) = Bπ; in this embodiment, the bandwidth B is set to 0.1, so ω c = 0.314 can be obtained. That is, the tuning of ω c in the frequency controller is completed;

[0048] For K P and K R , in order to amplify or reduce the input signal V(t) when passing through the frequency controller and cause signal distortion. The gain of the frequency controller can be set to 1, that is, G PR (jω) = 1. Substituting it into the expression of the proportional resonant control unit G PR (s), the relational expression: K P +K R = 1 can be obtained. Therefore, the value of K P can be tuned, and then the tuning of K R is completed.

[0049] S4.4. The remaining parameter set to be tuned for the final frequency controller is: θ = (K P, a0, a1, a2, b0, b1, b2);

[0050] S4.5. For the remaining part of the parameters to be tuned, a parameter optimization model r can be constructed as follows:

[0051]

[0052] where ξ set is the pre-designed lower limit of damping, which is set to 0.06 in this embodiment; θ max and θ min are the upper and lower limits of the parameters to be tuned. In this embodiment, θ max =(100, 10, 10, 10, 10, 10, 10); θ min =(5, 0.01, 0.01, 0.01, 0.01, 0.01, 0.01)

[0053] S5. Real-time collect the voltage signal V(t) and power signal P(t) of the CTHC-2# generator in a certain hydropower unit system;

[0054] S6. Take the voltage signal V(t) of the power system as the output of the frequency controller G(s), and the power signal P(t) as the input of the frequency controller G(s). Then, combine the deep reinforcement learning algorithm to transform the parameter tuning model of the frequency controller into a Markov decision process, and then iterate the Markov decision process to continuously reduce its model value until the algorithm converges, and finally obtain the adaptive control parameters of the frequency controller.

[0055] Figure 3 The frequency oscillation suppression effects are given when three strategies of optimizing the governor parameters, configuring a traditional controller, and the proposed frequency controller are applied to the system under the excitation of Fault 1 (a two-phase short-circuit fault occurred in the outgoing line of the CJL-220KV substation starting from 2 seconds and lasting for 100 milliseconds). It can be seen that compared with optimizing the governor parameters and configuring a traditional controller, the proposed frequency controller has a better oscillation suppression effect on the frequency oscillation curve and can make the system oscillation suppression faster.

[0056] Figure 4 (a) The distribution and probability density function of the main oscillation modes of the system in 2187 working conditions are counted in the scenario without configuring the proposed frequency controller; Figure 4 (b) The distribution and probability density function of the main oscillation modes of the system in 2187 working conditions are counted in the scenario where the system is configured with the proposed frequency controller and the controller is tuned based on the traditional method; Figure 4(c) The distribution and probability density function of the ultra-low frequency oscillation mode of the system in 2187 working conditions are counted for the frequency controller proposed in the system configuration, and the controller parameters are tuned based on the proposed method. It can be seen that when the proposed frequency controller is not configured, the ultra-low frequency oscillation mode is located in the right half-plane of the real axis under certain working conditions, which means that the system has the risk of ultra-low frequency oscillation. By analyzing the results of the ultra-low frequency oscillation mode with the frequency controller designed based on the traditional parameter tuning method and the proposed parameter tuning method, it can be seen that the ultra-low frequency oscillation mode is located in the left half-plane of the real axis under all working conditions. In addition, compared with the traditional tuning method, the proposed parameter tuning method makes the ultra-low frequency oscillation mode move more to the left, and the system has a greater stability margin.

[0057] Although the above description of the illustrative specific embodiments of the present invention is provided for the understanding of those skilled in the art of the present technology, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art of the present technology, as long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, these changes are obvious, and all inventions created using the concept of the present invention are within the scope of protection.

Claims

1. An optimization method for designing a frequency controller to improve the frequency stability of a power grid, characterized in that Including the following steps: (1) Design a frequency controller G(s) based on frequency division control theory, specifically a proportional resonant control unit G PR (s) and the polynomial model G M (s) series composition; (2) Collect the frequency deviation curve f(t) of the generator in the power system, identify the oscillation mode of the frequency deviation curve, and obtain the main oscillation mode information: λ = α + jω, where λ is the main oscillation mode information, α and ω are the real part and the imaginary part of the ultra-low frequency oscillation mode respectively, and j represents the imaginary symbol; (3) Calculate the damping ratio ξ of the frequency controller in the ultra-low frequency oscillation mode; (4) Construct a parameter tuning model of the frequency controller according to the damping ratio ξ of the ultra-low frequency oscillation mode; (5) Collect the voltage signal V(t) and power signal P(t) of the power system in real time; (6) Use the voltage signal V(t) of the power system as the output of the frequency controller G(s), and the power signal P(t) as the input of the frequency controller G(s). Then, combine the deep reinforcement learning algorithm to transform the parameter tuning model of the frequency controller into a Markov decision process, and then iterate the Markov decision process to continuously reduce its model value until the algorithm converges, and finally obtain the adaptive control parameters of the frequency controller; Among them, the method for designing the frequency controller G(s) based on the frequency division control theory in step (1) is: (2.1) Based on the frequency division control theory, define the proportional-resonant control unit G PR (s): Among them, K P is the proportional coefficient of the proportional resonance control unit; K R is the resonance coefficient of the proportional resonance control unit; ω0 is the resonance frequency of the proportional resonance control unit; ω c is the cut-off frequency of the proportional resonance control unit, and s represents a complex variable; (2.2), Define the polynomial model G M (s): Among them, a0, a1, a2, b0, b1, b2 are variables to be tuned in the polynomial; (2.3) Connect the proportional-resonant control unit G PR (s) in series with the polynomial model G M (s) to obtain the frequency controller structure G(s): Among them, the parameter tuning model of the frequency controller G(s) is: (3.1) Given the bandwidth B of the frequency controller G(s); (3.2) Set the resonance frequency of the proportional resonance control unit to ω0 = ω to complete the tuning of ω0; (3.3) Calculate the cut-off frequency ω of the proportional-resonant control unit according to the bandwidth B of the frequency controller G(s). c = Bπ, and complete the tuning of ω. c ​ (3.4)、Since the proportional coefficient K of the proportional resonance control unit P and the resonance coefficient K R satisfy: K P + K R = 1, thus by adjusting the value of K P , the adjustment of K R is completed; Finally, the remaining set of parameters to be tuned by the frequency controller is: θ = (K P , a0, a1, a2, b0, b1, b2); (3.5) Construct a parameter tuning optimization model r of the frequency controller according to the damping ratio ξ of the ultra-low frequency oscillation mode; θ max ≤ θ ≤ θ min Among them, ξ set is the lower limit value of the pre-designed damping, θ max and θ min are the upper and lower limits of the parameters to be tuned.

Images