A method for constructing a system frequency minimum point prediction model for random production simulation requirements

Abstract

A method for constructing a system frequency minimum point prediction model for random production simulation needs is applied to the rapid frequency response backup planning problem. Firstly, based on the thermal power multi-machine aggregation method, through the expansion of hydropower and gas units, the aggregation of multi-type unit parameters is realized, and a system frequency dynamic process analysis model for frequency response multi-resources is established; the unit speed governor characteristics are used to characterize the dynamic change process of the unit frequency, and the practical simplification analysis of the model from high order to low order is realized; the system operating condition is used as the zero state input, and the frequency dynamic tracking from the default system steady state to the initial state of the component is completed, so as to avoid misjudgment of the stable situation and better ensure the safety of system operation. Through analytical simplification and consideration of the initial operating state of the component, the model constructed by the present invention can take into account both calculation accuracy and calculation speed, and is suitable for the rapid frequency response backup planning problem.

Description

Technical Field

[0001] The invention belongs to the field of power grid frequency stability control and relates to a method for constructing a system frequency minimum point prediction model oriented to random production simulation requirements. Background Art

[0002] Under the "dual carbon" goal, the amount of new energy grid-connected in the power system continues to increase, and the characteristics of high penetration and high power electronics are more prominent. The system frequency response problem is not prominent when the proportion of thermal power is high, but because the new energy units do not have the inertia of traditional units, the high-proportion new energy system as a whole presents low inertia and low frequency response characteristics, and the severity of the accident has escalated sharply. The traditional fixed-coefficient reserve plan is relatively lagging behind in the new frequency stability situation. In response to the "dual low" problem under the "dual high" power system, it is urgent to improve the system's rapid frequency response capability. It is urgent to reasonably evaluate the demand for rapid frequency response reserves under the new situation and accurately grasp the dynamic characteristics of the system frequency. The addition of rapid frequency response reserves can accurately improve the system's frequency regulation capabilities, and tailor-made planning plans can ensure the stable operation of the system.

[0003] At present, the backup planning scheme is obtained based on random production simulation to obtain the state of the unit and the collection of lost loads, and through system reliability analysis, the economic benefits are measured by the backup capacity to make up for the power shortage. The fault data set of hundreds or thousands of years is a simple algebraic addition and subtraction in calculating the power shortage. Adding a fast frequency response backup type means shifting from focusing on static power shortage to focusing on the frequency dynamic process. Since the frequency drop trend of the system is most significant within seconds, the fast frequency response capability must be considered in the planning scheme. The frequency dynamic process must be included in the random production simulation system state to clarify the interception capability of the system resource capacity and response speed to the power loss event. In addition, in the recovery process after the power system suffers a power disturbance, the lowest point of the frequency directly reflects the severity of the accident, and is also the basis for control decision-making in actual engineering. Therefore, the new random production simulation should include the system frequency dynamic process analysis, and the lowest point in the frequency dynamic process after the system is disturbed should be used as one of the bases for whether the planning scheme is feasible.

[0004] The existing models all default to the initial operating condition rating, and the system frequency difference, as the main indicator of the operating condition, defaults to zero. However, in actual operation, the lowest point of the system frequency depends on the initial operating condition of the system. If the initial value in the calculation process is still the rated frequency, and the result is used as the system stability criterion, it may cause misjudgment of the system safety and stability situation (for example, when the system frequency is already running at a low level, a small-scale fault can cause the system frequency to become unstable), thereby threatening the safety of system operation. In addition, since the traditional analytical method can only characterize the state of system frequency change after the disturbance, it is impossible to characterize the deployment of means with response delays. Whether there is a delay from the disturbance to the deployment of frequency response means will lead to completely different frequency response results. Therefore, it is necessary to add the system initial state (under non-standard frequency) on the basis of the original zero-state model to construct a full response model. Summary of the invention

[0005] In view of the above problems, the present invention proposes a system frequency minimum point prediction model for random production simulation needs, which is applied to the problem of fast frequency response backup planning. The traditional frequency minimum point prediction model defaults to the initial operating condition rating, which may easily lead to misjudgment of the system safety and stability situation and threaten the safety of system operation, as well as the problem that the means with response delay cannot be characterized.

[0006] The technical solution adopted by the present invention is:

[0007] A method for constructing a system frequency minimum point prediction model for random production simulation needs, comprising:

[0008] Step 1: Frequency response model parameter aggregation

[0009] The "weighted average"-based aggregate system frequency response (ASFR) calculation method can scientifically simplify different types of multi-machine thermal power units into a single unit and reduce the model order. This method can be extended from thermal power to multi-resource units such as hydropower and gas, and can provide ideas for the construction of a multi-resource system frequency response (MR-SFR) model in terms of parameter aggregation.

[0010] Adjustment coefficient 1 / R of multi-machine hydropower ASFR model h It can be defined as:

[0011]

[0012] Where 1 / R hi is the adjustment coefficient of the i-th unit, K hi is the unit capacity ratio of the i-th unit, N is the total number of online units, k hiis the equivalent gain parameter of the i-th unit.

[0013] Normalized gain of a single-machine branch (single-machine ratio) λ hi for:

[0014]

[0015] Among them, the sum of the standardized gains of each unit conforms to the following formula:

[0016]

[0017] The model convergence time constant can be defined as:

[0018]

[0019] Where: Y represents the model time constant set, T w Represents the inertia time constant of the hydropower unit, T y represents the servomotor response time constant, T RS Represents the governor reset time, T RH Represents the transient droop time constant of the speed regulator, Y i Represents a certain type of time constant value in the i-th unit set.

[0020] The gas turbine unit polymerization process and method are the same as above.

[0021] Step 2: Build a low-order frequency response model

[0022] The SFR open-loop model of the gas turbine unit constructed by describing the governor characteristics with the first-order inertia link has low order and high precision. The primary frequency regulation of the system is the process in which the valve moves after the unit governor senses the system frequency difference and then changes the output power of the prime mover. This part can be processed in the open loop of the frequency response model based on the governor characteristics. Consider applying this method to thermal power units and hydropower units.

[0023] The water, fire and gas units are all fitted with the governor characteristics, and the fitting parameters are solved based on the least squares method to achieve the first-order inertia characterization of the unit performance. In addition, the energy storage unit participates in the frequency regulation and is described by the first-order inertia link. The MR-SFR open-loop structure characterized by the governor characteristics is obtained. By connecting the governors of various resource units in parallel, the frequency response process of the system resource side can be characterized (such as Figure 1 As shown), where ΔP is defined Gi Represents the active power of each resource type unit, and its calculation method is expressed in the complex frequency domain as follows:

[0024]

[0025] Where: K iis the power-frequency static characteristic coefficient after fitting of the same type of resources; T i The response time constant after fitting for the same type of resources; Δ f is the frequency offset of the entire system, and s is the complex frequency independent variable of the complex frequency domain expression.

[0026] After the power disturbance occurs, the system frequency change law conforms to the following formula:

[0027]

[0028] In the formula, It represents the analytical expression of the system frequency f to obtain its differential with respect to time t; Δf(t) is the approximate expression of the frequency after differentiation; H is the inertia time constant; Δ P L is the load disturbance of the system; m Δf is the initial frequency attenuation slope after the system suffers power disturbance.

[0029] Substituting equation (6) into equation (5) and using the inverse Laplace transform, we can obtain the system output active power ΔP G Time domain expression, where e is the natural exponential:

[0030]

[0031] Combined with the MR-SFR open-loop structure characterized by the speed regulator characteristics, the SFR time domain expression after the system is subjected to power disturbance can be obtained:

[0032]

[0033] Step 3: Build a full response model

[0034] All system components, including load-side resources, can be regarded as sufficiently large inductive components, and their energy charging and discharging are complex non-zero state response processes. Although the dynamic process obtained after the simplification of the SFR model retains a high degree of accuracy, if the initial state of the system components is not taken into account when the system suffers a power disturbance, the default initial frequency deviation of the system is zero, which will cause a large misjudgment of the actual frequency trajectory of the system. A full-response SFR model that takes into account the initial operating conditions should be established to accurately characterize the load-side resource response process with delays.

[0035] Based on the output power equation of the generator set in equation (5) in step 1, the Laplace inverse transform is used to obtain:

[0036]

[0037] In the process of transforming equation (9) from the time domain to the complex frequency domain, if the initial energy state P of the system prime mover is taken into account Gi (0- ), the active power of each resource type unit can be obtained Δ P Gi Complex frequency domain expression:

[0038]

[0039] The additional term in equation (10) compared to equation (5) is the addition of the initial state on the basis of the zero-state response, thus causing a change in the power component.

[0040] The known rotor motion equation is:

[0041]

[0042] In the formula, Δ P a It is the acceleration power, that is, the difference between mechanical power and load power; D is the system damping coefficient.

[0043] To analyze the frequency component, we need to perform inverse Laplace transform on equation (11) to obtain:

[0044]

[0045] When transforming equation (12) from the time domain to the complex frequency domain, if the initial state of the system rotor kinetic energy is taken into account, that is, Δ f(0 - ), we get a new Δ The complex frequency domain expression of f is:

[0046]

[0047] Compared with equation (11), equation (13) has an additional term, which is the addition of the initial state on the basis of the zero-state response, thus causing a change in the rotor description.

[0048] According to the superposition theorem, the full response model consists of the power component and the frequency component initial state (P Gi (0 - ), Δ f(0 - )) and the zero input response caused by external excitation with power and frequency components ( Δ P L , Δ f) The zero-state response induced by these two parts, combined with the MR-SFR open-loop structure characterized by the speed regulator characteristics obtained in step 1, can obtain the system full-response open-loop structure described based on the first-order inertia link.

[0049] Previously, the time domain analytical expressions of the zero-state response component in the full response model have been described by equations (9) and (12). For the analytical calculation of the zero-input response component of MR-SFR, since the load damping has little effect on the system frequency change before the lowest frequency point appears, the effect of load damping on the system frequency is ignored in the calculation process with reference to similar studies.

[0050] Only P Gi (0 - )According to the full response process, the complex frequency domain expression can be obtained:

[0051]

[0052] By using the inverse Laplace transform, we can get the initial state P of the system considering only the energy of the prime mover. Gi (0 - )’s zero input response time domain expression:

[0053]

[0054] When only considering Δ f(0 - ), the complex frequency domain expression can be obtained according to the full response process:

[0055]

[0056] By using the inverse Laplace transform, we can obtain the initial state of the system rotor kinetic energy. Δ f(0 - )’s zero input response time domain expression:

[0057]

[0058] By superimposing equation (15) and equation (17), we can obtain the zero-input time-domain analytical formula considering the initial state of system energy:

[0059]

[0060] By superimposing the zero-state time-domain analysis of equation (8) with equation (18), we can obtain the time-domain analysis of the system's full response:

[0061]

[0062] Step 4: Calculate the lowest frequency point of the system

[0063] Through step 1, system parameter aggregation can be used to simplify the single machine of the same type of resources. On this basis, the model order is further reduced by fitting the first-order inertia link in step 2. Step 3 considers the full response state of the system on the basis of the simplified model to make the model more in line with the actual system operating conditions. The analytical model obtained through the above three steps can provide a model basis for calculating the lowest frequency point in this step. The above processes are all based on solving the severity of the frequency drop and realizing the dynamic analysis of the frequency. The lowest frequency point directly reflects the degree of threat to the system, so it is necessary to further analyze and solve the established full response model.

[0064] Let the speed governor fitting parameters of the i-th unit be Total response component parameters of the i-th unit Let t = t min hour When the system frequency reaches the lowest point, equation (19) can be rearranged as:

[0065]

[0066] Then we have:

[0067]

[0068] When P in formula (21) Gi (0 - ), Δ f(0 - ) are all 0, t min The expression can be organized as:

[0069]

[0070] The above equation is the time to reach the minimum frequency without considering the initial state. It can be seen that when the initial state is not considered, t min With power disturbance Δ P L It has nothing to do with the characteristics of the system itself. However, after considering the initial state, t min It is not only related to the characteristics of the unit itself, but also to the power disturbance. Δ P L And the initial state Δ f(0 - ) and P Gi (0 - ). Among them, δ(t) is a unit impulse function, so this component is zero when t>0, that is, t min and Δ f(0 - ) item is irrelevant, t min Only with P Gi (0 - ) is related to the characteristics of the unit itself.

[0071] Among them, Ci, Bi, t min All of them are unknown quantities. The MR-SFR full response model has a total of 2N+1 unknown quantities. Therefore, a 2N+1-order equation system is established to solve:

[0072]

[0073] Integrating both ends of equation (19), we can obtain Δ The time domain expression of the full response model of f is:

[0074]

[0075] Substituting formula (6) into the above formula, we can get:

[0076]

[0077] In formula (25) Δ f(t) is determined by the power perturbation Δ P L The frequency offset caused by the system and the initial state of the rotor kinetic energy Δ f(0 - ) consists of two parts.

[0078] Solve the equations to get the result t min Substituting into equation (20), we can get the maximum frequency deviation of the system: Δ f max , and then calculate the nominal value f of the lowest frequency point after the system is subjected to power disturbance min .

[0079] f min =f0-f B Δf max (26)

[0080] Among them, f0 represents the initial frequency before the system is disturbed, f B Indicates the base frequency of the system.

[0081] In the above equations, the impact time is considered to be 0. For the characterization of the system load side response and continuous disturbance, the full response expression can be rewritten as the frequency response process after t0 when the first disturbance occurs:

[0082]

[0083] Step 5: Based on the above four steps, a system frequency minimum point prediction model for random production simulation needs is obtained. The frequency minimum point modeling and solution process is as follows: Figure 2As shown, it can provide a means of dynamic frequency tracking and a basis for assessing the severity of accidents in random production simulation processes, better grasp the frequency response characteristics of the system, and provide necessary analysis methods for rapid frequency response backup planning.

[0084] The beneficial effect of the present invention is that in the formulation of a fast frequency response backup planning scheme, random production simulation requires frequency stability analysis of a large number of operating modes, and rapid analysis and calculation are required. To this end, a system multi-resource full response frequency minimum point prediction analytical model based on the superposition theorem is proposed, which can realize the rapid analytical analysis of the frequency dynamic process. The present invention has the following characteristics: 1) Considering the full response. Taking into account the initial state of the system components, the misjudgment of the system frequency change trajectory caused by the default system standard operating conditions is avoided, and the shortcomings of the traditional analytical method that the secondary power disturbance (support) cannot be taken into account are improved, and it has better adaptability. 2) Fast calculation speed. Since the order of the aggregated simplified model is independent of the number of units and the system capacity, as the scale of the system is further expanded, its advantage in solution speed will be more prominent compared with the traditional simulation method. 3) High solution accuracy. Although it has gone through multiple steps such as parameter aggregation and first-order inertia link fitting, the calculation examples of the system under power loss and load side response support have shown its advantages in solution accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 It is the open loop schematic diagram of the system described based on the first-order inertia link;

[0086] Figure 2 Flowchart for modeling full-response MR-SFR;

[0087] Figure 3 To simulate the load loss of the power system in one year;

[0088] Figure 4 is the maximum frequency difference result of the simulation method;

[0089] Figure 5 It is the maximum frequency difference result of the full response analytical method;

[0090] Figure 6 For error distribution comparison. DETAILED DESCRIPTION

[0091] The present invention will be further described below in conjunction with the accompanying drawings and specific implementations, but they are not intended to limit the present invention.

[0092] A method for constructing a system frequency minimum point prediction model for random production simulation needs includes the following contents:

[0093] Step 1: Frequency response model parameter aggregation

[0094] The "weighted average"-based aggregate system frequency response (ASFR) calculation method can scientifically simplify different types of multi-machine thermal power units into a single unit, reducing the model order. This method can be extended from thermal power to multi-resource units such as hydropower and gas, and can provide ideas for the construction of a multi-resource system frequency response (MR-SFR) model in terms of parameter aggregation.

[0095] Adjustment coefficient 1 / R of multi-machine hydropower ASFR model h It can be defined as:

[0096]

[0097] Where 1 / R hi is the adjustment coefficient of the i-th unit, K hi is the unit capacity ratio of the i-th unit, N is the total number of online units, k hi is the equivalent gain parameter of the i-th unit.

[0098] Normalized gain of a single-machine branch (single-machine ratio) λ hi for:

[0099]

[0100] Among them, the sum of the standardized gains of each unit conforms to the following formula:

[0101]

[0102] The model convergence time constant can be defined as:

[0103]

[0104] Where: Y represents the model time constant set, T w Represents the inertia time constant of the hydropower unit, T y represents the servomotor response time constant, T RS Represents the governor reset time, T RH Represents the transient droop time constant of the speed regulator, Y i Represents a certain type of time constant value in the i-th unit set. The gas turbine unit aggregation process and method are similar.

[0105] Step 2: Build a low-order frequency response model

[0106] The SFR open-loop model of the gas turbine unit constructed by describing the governor characteristics with the first-order inertia link has low order and high precision. The primary frequency regulation of the system is the process in which the valve moves after the unit governor senses the system frequency difference and then changes the output power of the prime mover. This part can be processed in the open loop of the frequency response model based on the governor characteristics. Consider applying this method to thermal power units and hydropower units.

[0107] The water, fire and gas units are all fitted with the governor characteristics, and the fitting parameters are solved based on the least squares method to achieve the first-order inertia characterization of the unit performance. In addition, the energy storage unit participates in the frequency regulation and is described by the first-order inertia link, and the MR-SFR open-loop structure characterized by the governor characteristics is obtained, such as Figure 1 As shown in the figure, ΔP Gi Represents the active power of each resource type unit, and its calculation method is expressed in the complex frequency domain as follows:

[0108]

[0109] Where: K i is the power-frequency static characteristic coefficient after fitting of the same type of resources; T i The response time constant after fitting for the same type of resources; Δ f is the frequency offset of the entire system, and s is the complex frequency independent variable of the complex frequency domain expression.

[0110] After the power disturbance occurs, the system frequency change law conforms to the following formula:

[0111]

[0112] In the formula, It represents the analytical expression of the system frequency f to obtain its differential with respect to time t; Δf(t) is the approximate expression of the frequency after differentiation; H is the inertia time constant; Δ P L is the load disturbance of the system; m Δf is the initial frequency attenuation slope after the system suffers power disturbance.

[0113] Substituting equation (6) into equation (5) and using the inverse Laplace transform, we can obtain the system output active power ΔP G Time domain expression, where e is the natural exponential:

[0114]

[0115] Combined with the MR-SFR open-loop structure characterized by the speed regulator characteristics, the SFR time domain expression after the system is subjected to power disturbance can be obtained:

[0116]

[0117] Step 3: Build a full response model

[0118] All system components, including load-side resources, can be regarded as sufficiently large inductive components, and their energy charging and discharging are complex non-zero state response processes. Although the dynamic process obtained after the simplification of the SFR model retains a high degree of accuracy, if the initial state of the system components is not taken into account when the system suffers a power disturbance, the default initial frequency deviation of the system is zero, which will cause a large misjudgment of the actual frequency trajectory of the system. A full-response SFR model that takes into account the initial operating conditions should be established to accurately characterize the load-side resource response process with delays.

[0119] Based on the output power equation of the generator set in equation (5) in step 1, the Laplace inverse transform is used to obtain:

[0120]

[0121] In the process of transforming equation (9) from the time domain to the complex frequency domain, if the initial energy state P of the system prime mover is taken into account Gi (0 - ), we can get the new active power of each resource type unit Δ P Gi Complex frequency domain expression:

[0122]

[0123] The additional term in equation (10) compared to equation (5) is the addition of the initial state on the basis of the zero-state response, thus causing a change in the power component.

[0124] The known rotor motion equation is:

[0125]

[0126] In the formula, Δ P a It is the acceleration power, that is, the difference between mechanical power and load power; D is the system damping coefficient.

[0127] To analyze the frequency component, we need to perform inverse Laplace transform on equation (11) to obtain:

[0128]

[0129] When transforming equation (12) from the time domain to the complex frequency domain, if the initial state of the system rotor kinetic energy is taken into account, that is, Δ f(0 - ), we get a new Δ The complex frequency domain expression of f is:

[0130]

[0131] Compared with equation (11), equation (13) has an additional term, which is the addition of the initial state on the basis of the zero-state response, thus causing a change in the rotor description.

[0132] According to the superposition theorem, the full response model consists of the power component and the frequency component initial state (P Gi (0 - ), Δ f(0 - )) and the zero input response caused by external excitation with power and frequency components ( Δ P L , Δ f) The zero-state response induced by these two parts, combined with the MR-SFR open-loop structure characterized by the speed regulator characteristics obtained in step 1, can obtain the system full-response open-loop structure described based on the first-order inertia link.

[0133] Previously, the time domain analytical expressions of the zero-state response component in the full response model have been described by equations (9) and (12). For the analytical calculation of the zero-input response component of MR-SFR, since the load damping has little effect on the system frequency change before the lowest frequency point appears, the effect of load damping on the system frequency is ignored in the calculation process with reference to similar studies.

[0134] Only P Gi (0 - )According to the full response process, the complex frequency domain expression can be obtained:

[0135]

[0136] By using the inverse Laplace transform, we can get the initial state P of the system considering only the energy of the prime mover. Gi (0 - )’s zero input response time domain expression:

[0137]

[0138] When only considering Δ f(0 - ), the complex frequency domain expression can be obtained according to the full response process:

[0139]

[0140] By using the inverse Laplace transform, we can obtain the initial state of the system rotor kinetic energy. Δ f(0 - )’s zero input response time domain expression:

[0141]

[0142] By superimposing equation (15) and equation (17), we can obtain the zero-input time-domain analytical formula considering the initial state of system energy:

[0143]

[0144] By superimposing the zero-state time-domain analysis of equation (8) with equation (18), we can obtain the time-domain analysis of the system's full response:

[0145]

[0146] Step 4: Calculate the lowest frequency point of the system

[0147] The aggregation of system parameters in step 1 and the fitting of the first-order inertia link in step 2, combined with the construction of the full response model in step 3, are all based on solving the severity of the frequency drop and realizing frequency dynamic analysis. The lowest frequency point directly reflects the degree of threat to the system, so it is necessary to further analyze and solve the established full response model.

[0148] Let the speed governor fitting parameters of the i-th unit be Total response component parameters of the i-th unit Let t = t min hour When the system frequency reaches the lowest point, equation (19) can be rearranged as:

[0149]

[0150] Then there is

[0151]

[0152] When P in formula (21) Gi (0 - ), Δ f(0 - ) are all 0, t min The expression can be organized as:

[0153]

[0154] The above equation is the time to reach the minimum frequency without considering the initial state. It can be seen that when the initial state is not considered, t min With power disturbance Δ P L It has nothing to do with the characteristics of the system itself. However, after considering the initial state, t min It is not only related to the characteristics of the unit itself, but also to the power disturbance. Δ P L And the initial state Δ f(0 - ) and P Gi (0- ). Among them, δ(t) is a unit impulse function, so this component is zero when t>0, that is, t min and Δ f(0 - ) item is irrelevant, t min Only with P Gi (0 - ) is related to the characteristics of the unit itself.

[0155] Among them, Ci, Bi, t min All of them are unknown quantities. The MR-SFR full response model has a total of 2N+1 unknown quantities. Therefore, a 2N+1-order equation system is established to solve:

[0156]

[0157] Integrating both ends of equation (19), we can obtain Δ The time domain expression of the full response model of f is:

[0158]

[0159] Substituting formula (6) into the above formula, we can get:

[0160]

[0161] In formula (25) Δ f(t) is determined by the power perturbation Δ P L The frequency offset caused by the system and the initial state of the rotor kinetic energy Δ f(0 - ) consists of two parts.

[0162] Solve the equations to get the result t min Substituting into equation (20), we can get the maximum frequency deviation of the system: Δ f max , and then calculate the nominal value f of the lowest frequency point after the system is subjected to power disturbance min .

[0163] f min =f0-f B Δf max (26)

[0164] Among them, f0 represents the initial frequency before the system is disturbed, f B Indicates the base frequency of the system.

[0165] In the above equations, the impact time is considered to be 0. For the characterization of the system load side response and continuous disturbance, the full response expression can be rewritten as the frequency response process after t0 when the first disturbance occurs:

[0166]

[0167] Step 5: Based on the above four steps, a system frequency minimum point prediction model for random production simulation needs is obtained. The frequency minimum point modeling and solution process is as follows: Figure 2 As shown, it can provide a means of dynamic frequency tracking and a basis for assessing the severity of accidents in random production simulation processes, better grasp the frequency response characteristics of the system, and provide necessary analysis methods for rapid frequency response backup planning.

[0168] In the specific implementation example, the simulation and analysis of the frequency minimum point of the full response MR-SFR for the random production simulation process throughout the year are used as the object to compare the solution speed. After the failure of the system random production simulation occurs, the response delay is set to 1s, and the load-side resources with a size of 50% of the system load loss are used for power support, and the frequency response support effect under the spare capacity is analyzed. Considering the power shortage caused by the unit failure, the simulation time is set to one year, the time interval is 1h, and the total time of the year is 8736h. According to Figure 3 The annual load loss simulation curve of the power system is combined with the full-response MR-SFR model to complete the frequency minimum point prediction of the multi-resource system for random production simulation needs.

[0169] Figure 4 , 5 Corresponding to Figure 3 The maximum frequency difference comparison results of the full response MR-SFR simulation and analytical method under the full load loss condition throughout the year, the frequency difference reference value is 50Hz. It can be seen from the figure that the overall lowest point distribution and value are relatively consistent. For a clearer comparison, the following are calculated: Figure 6 The distribution of absolute and relative errors between the two, the maximum absolute error is -5.6935*10 -4 Hz, the relative error rate of the simulation method as the benchmark value is 0.058%, corresponding to the 3257th sampling point, which corresponds to a system power shortage of 365.9MW; the 5814th sampling point corresponds to the system's annual maximum power shortage of 865.6MW, which is a low-probability extreme event, and the absolute error of the lowest system frequency point is -2.9105*10 -4 The relative error rate of Hz is 0.012%; the maximum relative error rate of the 3239th sampling point corresponding to the lowest point of the system frequency is 0.089%, and the absolute error is -6.29*10 -7 Hz, which corresponds to a system power shortage of 263.5kW. It can be seen that the frequency minimum point prediction error is relatively large when the system has a small power disturbance. The average absolute error of the overall frequency minimum point prediction under load loss conditions in 8736 time intervals throughout the year is -5.9263*10 -5Hz, with an average relative error of 0.014%. In terms of solution speed, in a solution environment with CPU: AMD Ryzen 3 3200G, main frequency 3.60GHz, memory 16.00GB, MATLAB R2018b was used to process 543 load loss points simulated throughout the year. The full response analytical method took 258.97s, while the original simulation method took 2017.43s, a speed increase of 7.79 times.

[0170] The above describes the preferred method of the present invention. It should be understood that the present invention is not limited to the above-mentioned specific implementation method, and the equipment and structures that are not described in detail should be understood to be implemented in a common manner in the art; any technician familiar with the art can use the above-disclosed methods and technical contents to make many possible changes and modifications to the technical methods of the present invention without departing from the scope of the technical solution of the present invention, or modify them into equivalent methods with equivalent changes, which does not affect the essential content of the present invention. Therefore, any simple modification, equivalent change and modification made to the above method according to the technical essence of the present invention without departing from the content of the technical method of the present invention still falls within the scope of protection of the technical method of the present invention.

Claims

1. A method for constructing a system frequency minimum point prediction model for random production simulation needs, characterized in that: The model building method comprises the following steps: Step 1: Frequency response model parameter aggregation The frequency response model of multi-resource system is constructed by using the aggregate system frequency response method based on "weighted average"; Adjustment coefficient 1 / R of multi-machine hydropower ASFR model h Defined as: Where 1 / R hi is the adjustment coefficient of the i-th unit, K hi is the unit capacity ratio of the i-th unit, N is the total number of online units, k hi is the equivalent gain parameter of the i-th unit; Normalized gain λ of a single branch hi for: Among them, the sum of the standardized gains of each unit conforms to the following formula: The model convergence time constant is defined as: Where: Y represents the model time constant set, T w Represents the inertia time constant of the hydropower unit, T y represents the servomotor response time constant, T RS Represents the governor reset time, T RH Represents the transient droop time constant of the speed regulator, Y i Represents a certain type of time constant value in the i-th unit set; The gas turbine unit polymerization process and method are the same as described above; Step 2: Build a low-order frequency response model The water, fire and gas units are all fitted with the governor characteristics, and the fitting parameters are solved based on the least squares method to achieve the first-order inertia characterization of the unit performance. In addition, the first-order inertia link is used to characterize the energy storage unit participating in the frequency regulation, and the MR-SFR open-loop structure characterized by the governor characteristics is obtained, where ΔP Gi Represents the active power of each resource type unit, and its calculation method is expressed in the complex frequency domain as follows: Where: K i is the power-frequency static characteristic coefficient after fitting of the same type of resources; T i is the response time constant after fitting of the same type of resources; Δf is the frequency offset of the entire system, and s is the complex frequency independent variable of the complex frequency domain expression; After the power disturbance occurs, the system frequency change law conforms to the following formula: Where Δf(t) is the frequency approximation after differentiation; H is the inertia time constant; ΔP L is the load disturbance of the system; m Δf is the initial frequency attenuation slope after the system suffers power disturbance; Substituting equation (6) into equation (5) and using the inverse Laplace transform, we can obtain the system output active power ΔP G Time domain expression, where e is the natural exponential: Combined with the MR-SFR open-loop structure characterized by the speed regulator characteristics, the SFR time domain expression after the system is subjected to power disturbance can be obtained: Step 3: Build a full response model Based on the output power equation of the generator set in equation (5) in step 1, the Laplace inverse transform is used to obtain: In the process of transforming equation (9) from the time domain to the complex frequency domain, if the initial energy state P of the system prime mover is taken into account Gi (0 - ), and obtain the new active power ΔP of each resource type unit Gi Complex frequency domain expression: The rotor motion equation is: In the formula, ΔP a is the acceleration power, i.e. the difference between mechanical power and load power; D is the system damping coefficient; To analyze the frequency components, we need to perform an inverse Laplace transform on equation (11) to obtain: When transforming equation (12) from the time domain to the complex frequency domain, if the initial state of the system rotor kinetic energy is taken into account, that is, Δf(0 - ), and obtain the new complex frequency domain expression of Δf: According to the superposition theorem, the full response model consists of the power component and the frequency component initial state (P Gi (0 - )、Δf(0 - )) and the zero input response caused by the external excitation of power and frequency components (ΔP L , Δf), combined with the MR-SFR open-loop structure characterized by the speed regulator characteristics obtained in step 2, the full response open-loop structure of the system described based on the first-order inertia link can be obtained; The time domain analytical expression of the zero-state response component in the full response model is described by equations (9) and (12), and the analytical solution of the zero-input response component of MR-SFR is obtained. The influence of load damping on the system frequency is ignored in the process of obtaining the component. Only P Gi (0 - )According to the full response process, the complex frequency domain expression can be obtained: By using the inverse Laplace transform, we can get the initial state P of the system considering only the energy of the prime mover. Gi (0 - )’s zero input response time domain expression: When only Δf(0 - ), the complex frequency domain expression can be obtained according to the full response process: By using the Laplace inverse transformation, we can get the value of considering only the initial state of the rotor kinetic energy of the system Δf(0 - )’s zero input response time domain expression: By superimposing equation (15) and equation (17), we can obtain the zero-input time-domain analytical formula considering the initial state of system energy: The zero-state time-domain analysis of equation (8) is superimposed with equation (18) to obtain the time-domain analysis of the full response of the system: Step 4: Calculate the lowest frequency point of the system Let the speed governor fitting parameters of the i-th unit be Total response component parameters of the i-th unit Let t = t min hour When the system frequency reaches the lowest point, equation (19) is: Then we have: When P in formula (21) Gi (0 - )、Δf(0 - ) are all 0, t min The expression is: The above equation is the time to reach the minimum frequency without considering the initial state; where δ(t) is a unit impulse function, so this component is zero when t>0, that is, t min With Δf(0 - ) item is irrelevant, t min Only with P Gi (0 - ) is related to the characteristics of the unit itself; among them, Ci, Bi, t min All of them are quantities to be determined, and a 2N+1-order equation system is established to solve: Integrating both ends of equation (19) yields the time domain expression of the full response model of Δf: Substituting formula (6) into the above formula, we can get: Solve the equations to get the result t min Substituting into equation (20), we can get the maximum frequency deviation Δf of the system: max , calculate the nominal value f of the lowest frequency point after the system is subjected to power disturbance min ; f min =f0-f B ·Δf max (26) Among them, f0 represents the initial frequency before the system is disturbed, f B Indicates the base frequency of the system; In the above equations, the impact time is considered to be 0. For the characterization of the system load side response and continuous disturbance, the full response expression can be rewritten as the frequency response process after t0 when the first disturbance occurs: Step 5: Based on the above four steps, the system frequency minimum point prediction model for random production simulation needs is obtained, which can provide frequency dynamic tracking means and accident severity assessment basis for the random production simulation process, and grasp the system frequency response characteristics.

Images