Novel power system inertia demand evaluation method and system

Abstract

The invention relates to a novel power system inertia demand assessment method and system, belongs to the technical field of novel power system planning, and solves the problems that inertia demand is difficult to assess accurately and planning is unreliable due to new energy access in the prior art. Comprising the following steps: constructing a unit combination optimization model of a novel power system, and constructing an input data set and an output data set based on obtained historical data of a day-ahead new energy output and load prediction curve; constructing a frequency characteristic prediction model, and training the frequency characteristic prediction model based on the sample data set to obtain an optimal frequency characteristic prediction model; obtaining a day-ahead new energy output and load prediction curve, and adjusting the scheduling plan based on the unit combination optimization model, the optimal frequency characteristic prediction model and the inertia adjustment constraint to obtain an inertia demand evaluation result; the inertia demand evaluation result comprises a system total inertia level result and a corresponding scheduling plan.

Description

Technical Field

[0001] The present invention relates to the technical field of power system planning, and in particular to a novel power system inertia demand assessment method and system. Background Art

[0002] With the rapid development of renewable energy power generation, by the end of 2023, the installed capacity of renewable energy has reached 1.57 billion kilowatts, accounting for more than 50% of the total installed capacity for the first time. The penetration of renewable energy units has brought an impact on the frequency security of traditional power systems. Accurate inertia demand assessment is the prerequisite for ensuring the adequacy of system inertia and preventing frequency security incidents, and can provide a reference for system resource planning and the formulation of safe operation strategies. However, as the proportion of renewable energy increases, traditional inertia resources will be insufficient to support system operation. Therefore, renewable energy inertia support based on virtual inertia control has been widely studied. However, the inertia support capacity of renewable energy is affected by the grid connection mode and changes in natural conditions, and it is difficult to characterize it through accurate mathematical modeling, which brings challenges to the inertia demand assessment of new power systems. Research on accurate inertia demand assessment considering differences in renewable energy grid connection modes and time-varying operating conditions is of great significance to the construction of new power systems.

[0003] Therefore, an accurate system inertia demand assessment method is urgently needed to solve the problem of difficult to accurately assess inertia demand caused by the access of new energy. Summary of the invention

[0004] In view of the above analysis, the embodiments of the present invention aim to provide a novel method and system for evaluating inertia demand of a power system, so as to solve the problem that it is difficult to accurately evaluate inertia demand caused by the access of new energy sources.

[0005] On the one hand, an embodiment of the present invention provides a novel method for evaluating inertia demand of a power system, comprising the following steps:

[0006] Construct a unit combination optimization model for a new power system, and then construct an input data set and an output data set based on the historical data of the new energy output and load forecast curve obtained the day before; wherein each element in the input data set is a one-dimensional sequence of the dispatch plan of the new power system, and each element in the output data set is the frequency characteristic data corresponding to each dispatch plan;

[0007] Constructing a frequency feature prediction model, and training the frequency feature prediction model based on a sample data set to obtain an optimal frequency feature prediction model; wherein the sample data set includes an input data set and an output data set;

[0008] The day-ahead renewable energy output and load forecast curves are obtained, and then the dispatch plan is adjusted based on the unit combination optimization model, the optimal frequency characteristic prediction model, and the inertia adjustment constraint to obtain the inertia demand assessment result; among them, the inertia demand assessment result includes the system total inertia level result and the corresponding dispatch plan.

[0009] Furthermore, the scheduling plan includes the output, startup mode, inertia constant of each synchronous unit in each period within the scheduling cycle, the output of each new energy unit, the virtual inertia constant of each VSG, the time constant of each speed regulator, the total inertia level of the system, the synchronous inertia level of the system, the virtual inertia level of the system and the total primary frequency regulation capability of the speed regulator; the frequency characteristic data includes the maximum frequency change and the maximum frequency change rate; wherein the maximum frequency change is the difference between the system steady-state frequency and the lowest point frequency.

[0010] Furthermore, the input and output datasets are constructed in the following way:

[0011] Based on the historical data of the new energy output and load forecast curve obtained on the previous day, the unit combination optimization model is adopted to obtain the startup mode of each synchronous unit, the output of each synchronous unit and the actual output of each new energy unit in each period within the dispatch cycle;

[0012] Based on the startup mode of each synchronous unit in each period of the scheduling cycle, the output of each synchronous unit and the actual output of each new energy unit, as well as the virtual inertia constant of each VSG, the inertia constant of each synchronous unit and the primary frequency regulation capacity of each speed regulator, the total system inertia level, the system synchronous inertia level, the system virtual inertia level and the total primary frequency regulation capacity of the speed regulator in each period of each scheduling cycle are obtained, and then each scheduling plan is obtained, and each scheduling plan is respectively constituted into a one-dimensional sequence to obtain an input data set;

[0013] Based on the different disturbance scenarios and scheduling plans, the maximum frequency change and maximum frequency change rate corresponding to each scheduling plan under different disturbance scenarios are obtained through transient simulation, and the output data set is composed of frequency characteristic data.

[0014] Furthermore, the unit combination optimization model is to obtain the startup mode of each synchronous unit, the output of each synchronous unit and the actual output of each new energy unit in each period within the dispatching cycle under the operating constraints of the new power system with the minimum operating cost as the objective function; wherein the objective function F is expressed as:

[0015] F=min[f1+f2+f3+f4]

[0016] In the formula, f1 represents the power cost of the synchronous unit, f2 represents the total startup cost of the synchronous unit, f3 represents the total shutdown cost of the synchronous unit, and f4 represents the penalty cost for abandoning the consumption of new energy.

[0017] Furthermore, the synchronous unit power cost f1, the synchronous unit switch startup cost f2, the synchronous unit switch shutdown cost f3, and the penalty cost f4 for abandoning new energy consumption are respectively expressed as:

[0018]

[0019]

[0020]

[0021]

[0022] Where, T represents the total number of time periods; N represents the total number of synchronous units in the new power system; a i 、b i 、c i represent the first, second and third power generation cost coefficients of synchronous unit i respectively; P i,t represents the output of synchronous unit i in period t; They represent the startup cost and shutdown cost of synchronous unit i respectively; u i,t 、u i,t-1 They represent the startup mode of synchronous unit i in period t and period t-1 respectively, with a value of 1 indicating startup and a value of 0 indicating shutdown; J represents the total number of new energy units in the new power system; c j They represent the power generation cost coefficient of the new energy unit j respectively; represents the predicted output of new energy unit j during period t on a typical day; P j,t Represents the actual output of new energy unit j during period t.

[0023] Furthermore, the training of the frequency feature prediction model based on the sample data set includes:

[0024] Set the optimizer, training batch, learning rate, number of training iterations, and conservative loss function during training.

[0025] The sample data set is divided based on training batches, and the frequency feature prediction model training of each training batch is performed in sequence based on the divided sample data set. When the stopping condition is met, the optimal frequency feature prediction model is obtained.

[0026] Furthermore, the conservative loss function L total It is expressed as:

[0027] L total =L Δf +L RoCoF

[0028] Where, L ΔfRepresents the maximum frequency change loss function, L RoCoF Represents the maximum frequency change rate loss function.

[0029] Furthermore, the inertia requirement assessment result is obtained by:

[0030] S31. Based on the recent renewable energy output and load forecast curve, the unit combination optimization model is used to obtain the basic dispatch plan;

[0031] S32, forming a one-dimensional sequence of the basic scheduling plan and inputting it into an optimal frequency characteristic prediction model to obtain frequency characteristic data;

[0032] S33. Determine whether the frequency characteristic data meets the frequency safety conditions. If so, take the total system inertia level of each time period in the current basic scheduling plan as the total system inertia level result, thereby obtaining the inertia demand assessment result; otherwise, adjust the current basic scheduling plan based on the inertia adjustment constraint and the unit combination optimization model to obtain the adjusted basic scheduling plan, and repeat steps S32 to S33.

[0033] Furthermore, the inertia adjustment constraint is expressed as:

[0034]

[0035] In the formula, represents the startup mode of synchronous unit i in time period t in the pth iteration, It represents the startup mode of synchronous unit i in time period t in the p+1th iteration.

[0036] On the other hand, an embodiment of the present invention provides a novel power system inertia demand assessment system, comprising:

[0037] A data set construction module is used to construct a unit combination optimization model for a new power system, and then construct an input data set and an output data set based on the historical data of the new energy output and load forecast curve obtained a day ago; wherein each element in the input data set is a one-dimensional sequence of the dispatch plan of the new power system, and each element in the output data set is the frequency characteristic data corresponding to each dispatch plan;

[0038] A frequency feature prediction model construction and training model is used to construct a frequency feature prediction model and train the frequency feature prediction model based on a sample data set to obtain an optimal frequency feature prediction model; wherein the sample data set includes an input data set and an output data set;

[0039] The inertia demand assessment module is used to obtain the day-ahead renewable energy output and load forecast curves, and then adjust the dispatch plan based on the unit combination optimization model, the optimal frequency characteristic prediction model, and the inertia adjustment constraint to obtain the inertia demand assessment result; among them, the inertia demand assessment result includes the system total inertia level result and the corresponding dispatch plan.

[0040] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0041] The present invention provides a novel power system inertia demand assessment method and system. The method and system construct a unit combination optimization model of the novel power system, and then construct an input data set and an output data set based on the historical data of the acquired new energy output and load forecast curve on the previous day. The constructed frequency characteristic prediction model is trained based on the sample data set composed of the input data set and the output data set to obtain the optimal frequency characteristic prediction model. Based on the acquired new energy output and load forecast curve on the previous day, the scheduling plan is adjusted according to the unit combination optimization model, the optimal frequency characteristic prediction model and the inertia adjustment constraint to obtain the inertia demand assessment result, effectively assess the system inertia level, adapt to different operating conditions and expected disturbance situations, solve the problem that the inertia demand caused by the access of new energy is difficult to accurately assess, and is easy to solve and has strong adaptability. The system synchronous inertia redundancy is reduced, the inertia support capacity of the system is improved, and a reference value is provided for the formulation of the system's previous day scheduling plan.

[0042] In the present invention, the above-mentioned technical solutions can also be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can become obvious from the description, or can be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The drawings are only for the purpose of illustrating particular embodiments and are not to be considered limiting of the present invention. Like reference symbols denote like components throughout the drawings.

[0044] Figure 1 A schematic flow chart of a novel power system inertia demand assessment method provided in Example 1 of the present invention;

[0045] Figure 2 A schematic diagram of prediction results of the frequency characteristic prediction model provided in Example 3 of the present invention;

[0046] Figure 3 A schematic diagram of adjusting the system inertia requirement level provided in Example 3 of the present invention;

[0047] Figure 4A schematic diagram of iterative changes of the lowest point of the system frequency in each time period provided in Embodiment 3 of the present invention;

[0048] Figure 5 A schematic diagram of inertia requirement levels under different disturbance sizes provided in Example 3 of the present invention. DETAILED DESCRIPTION

[0049] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0050] Example 1

[0051] A specific embodiment of the present invention discloses a novel method for evaluating inertia demand of a power system, such as Figure 1 As shown, the following steps are included:

[0052] S1. Construct a unit combination optimization model for a new power system, and then construct an input data set and an output data set based on the historical data of the new energy output and load forecast curve obtained a day ago; wherein each element in the input data set is a one-dimensional sequence consisting of a dispatch plan of the new power system, and each element in the output data set is frequency characteristic data corresponding to each dispatch plan.

[0053] Specifically, the scheduling plan includes the output, startup mode, inertia constant of each synchronous unit in each period within the scheduling cycle, the output of each new energy unit, the virtual inertia constant of each VSG, the time constant of each speed regulator, the total inertia level of the system, the synchronous inertia level of the system, the virtual inertia level of the system and the total primary frequency regulation capability of the speed regulator; the frequency characteristic data includes the maximum frequency change and the maximum frequency change rate; among them, the maximum frequency change is the difference between the steady-state frequency of the system and the lowest point frequency.

[0054] More specifically, the scheduling period of the scheduling plan is one day.

[0055] During implementation, the unit combination optimization model is to obtain the startup mode of each synchronous unit, the output of each synchronous unit and the actual output of each new energy unit in each period within the dispatching cycle under the operating constraints of the new power system and with the minimum operating cost as the objective function; wherein the objective function F is expressed as:

[0056] F=min[f1+f2+f3+f4]

[0057] In the formula, f1 represents the power cost of the synchronous unit, f2 represents the total startup cost of the synchronous unit, f3 represents the total shutdown cost of the synchronous unit, and f4 represents the penalty cost for abandoning the consumption of new energy.

[0058] In specific implementation, the synchronous unit power cost f1, the synchronous unit switch startup cost f2, the synchronous unit switch shutdown cost f3, and the penalty cost f4 for abandoning new energy consumption are respectively expressed as:

[0059]

[0060]

[0061]

[0062]

[0063] Where, T represents the total number of time periods; N represents the total number of synchronous units in the new power system; a i 、b i 、c i represent the first, second and third power generation cost coefficients of synchronous unit i respectively; P i,t represents the output of synchronous unit i in period t; They represent the startup cost and shutdown cost of synchronous unit i respectively; u i,t 、u i,t-1 They represent the startup mode of synchronous unit i in period t and period t-1 respectively, with a value of 1 indicating startup and a value of 0 indicating shutdown; J represents the total number of new energy units in the new power system; c j They represent the power generation cost coefficient of the new energy unit j respectively; represents the predicted output of new energy unit j during period t on a typical day; P j,t Represents the actual output of new energy unit j during period t.

[0064] In specific implementation, the operating constraints of the new power system are expressed as:

[0065] P l,t =(θ m,t -θ n,t )x l

[0066] -P l max ≤P i,t ≤P l max

[0067]

[0068] u i,t P i min ≤P i,t ≤u i,t P i max

[0069]

[0070]

[0071] P i,t -P i,t-1 ≤UR i ·u i,t-1 +P i min (u i,t -u i,t-1 )+P i max (1-u i,t )

[0072] P i,t-1 -P i,t ≤DR i ·u i,t +P i min (u i,t-1 -u i,t )+P i max (1-u i,t-1 )

[0073] Where P l,t represents the transmission power of transmission line l during period t, θ m,t ,θ n,t They represent the phase angles of nodes m and n at both ends of the transmission line l during period t, and x l represents the reactance of the transmission line l, P l max Indicates the maximum transmission power capacity of transmission line l, P i min , P i max They represent the lower and upper limits of the synchronous unit output respectively. They represent the time that synchronous unit i has been continuously started and stopped in the t-1 period, T i on 、T i off They represent the time constraints for the continuous startup and shutdown of synchronous unit i, P i,t-1 represents the output of synchronous unit i in period t-1, UR i ,DR i They represent the upward and downward climbing power limits of synchronous unit i respectively.

[0074] It can be understood that the parameters required in the unit combination optimization model can be obtained through the current renewable energy output and load forecast curve and the parameters of each unit.

[0075] When implemented, the input and output datasets are constructed in the following way:

[0076] S11. Based on the historical data of the new energy output and load forecast curve obtained on the previous day, the unit combination optimization model is adopted to obtain the startup mode of each synchronous unit, the output of each synchronous unit and the actual output of each new energy unit in each time period within each scheduling cycle.

[0077] S12. Based on the startup mode of each synchronous unit in each time period within each scheduling cycle, the output of each synchronous unit and the actual output of each new energy unit, as well as the virtual inertia constant of each VSG, the inertia constant of each synchronous unit and the primary frequency regulation capability of each speed regulator, the total system inertia level, the system synchronous inertia level, the system virtual inertia level and the total primary frequency regulation capability of the speed regulator in each time period within each scheduling cycle are obtained, and then each scheduling plan is obtained, and each scheduling plan is respectively constructed into a one-dimensional sequence to obtain an input data set.

[0078] Specifically, the system total inertia level, system synchronous inertia level, system virtual inertia level and the total primary frequency regulation capability of the speed regulator in each period are expressed as:

[0079]

[0080]

[0081] E sys,t =E vir,t +E syn,t

[0082]

[0083] In the formula, E syn,t 、E vir,t 、E sys,t , R total,t They represent the total system inertia level, system synchronous inertia level, system virtual inertia level, and the total primary frequency regulation capability of the speed regulator in period t, respectively. i,t represents the inertia constant of synchronous unit i, H j,t represents the VSG virtual inertia constant of new energy unit j, R i,t It represents the primary frequency regulation capability of the speed regulator of synchronous unit i in period t.

[0084] S13. Based on the different disturbance scenarios and the scheduling plans, the maximum frequency change and the maximum frequency change rate corresponding to the scheduling plans under the different disturbance scenarios are obtained through transient simulation, and the output data set is constituted as the frequency characteristic data.

[0085] Specifically, different scenarios are set in the power system simulation software by setting the size and position of the active power disturbance, and the frequency change curve of each scheduling plan is obtained based on transient simulation calculation, and then the maximum frequency change amount and maximum frequency change rate of each scheduling plan are obtained.

[0086] Specifically, the disturbance magnitude ranges from 5% to 10% of the load of each node.

[0087] S2. Construct a frequency feature prediction model, and train the frequency feature prediction model based on a sample data set to obtain an optimal frequency feature prediction model; wherein the sample data set includes an input data set and an output data set.

[0088] During implementation, the frequency feature prediction model structure is sequentially an input data layer, a first processing layer, a second processing layer, a third processing layer, a fully connected layer, and an output data layer; wherein each processing layer includes a convolutional layer, a batch normalization layer, and a pooling layer connected in sequence; the model input is a one-dimensional sequence consisting of a scheduling plan, and the model output is frequency feature data; wherein the processing layer is used for feature extraction, normalization, and dimensionality reduction, and the fully connected layer is used to integrate and compress the extracted features.

[0089] It can be understood that the convolution layer performs convolution calculations on the input data to capture the mechanism relationship and features between the data; the batch normalization layer normalizes the input data of different ranges to improve the training efficiency and accuracy; the pooling layer performs dimensionality reduction processing on the input data and keeps the features unchanged; the fully connected layer connects the input data layer and the output data layer to implement the mechanism captured by the network. In other words, in the convolution layer, the convolution kernel is used to slide along a single direction of the sequence data. A single process only convolves with part of the sequence and generates a local feature map, which is expressed through an activation function. Subsequently, the local features are compressed into a fixed-length vector through pooling in the pooling layer to reduce the computational complexity. Through multiple combinations of convolutional and pooling layers, deeper data features are gradually discovered, and then flattened into feature vectors to input the fully connected layer, and finally mapped to the output space to form output data.

[0090] Specifically, the number of output channels of the convolution layers in the first processing layer, the second processing layer, and the third processing layer are 64, 128, and 256 respectively, the convolution kernel length is 4, the linear activation function is used, the same padding strategy is adopted, and the convolution step size is set to 1; the pooling window size of the pooling layer is 2; the number of neurons in the fully connected layer is 128, the activation function uses the ReLU activation function, and L2 regularization is adopted.

[0091] During implementation, the frequency feature prediction model is trained based on the sample data set, including:

[0092] S21. Set the optimizer, training batch, learning rate, number of training iterations, and conservative loss function during training.

[0093] Specifically, the optimizer is Adam, the training batch is 32, the learning rate is 1e-6, and the number of training iterations is 180.

[0094] S22, dividing the sample data set based on training batches, and training the frequency feature prediction model of each training batch in turn based on the divided sample data set, and obtaining the optimal frequency feature prediction model when the stopping condition is met.

[0095] Specifically, the stopping condition is that the number of training iterations reaches a set number; wherein, completing one frequency feature prediction model training based on the divided sample data set is one iteration.

[0096] Specifically, the frequency feature prediction model training is performed in each training batch in the following manner:

[0097] The sample data set divided by the current batch is input into the frequency feature prediction model for training, and the network hyperparameters are adjusted using the back propagation method based on the learning rate and optimizer to minimize the conservative loss function.

[0098] In specific implementation, the conservative loss function L total It is expressed as:

[0099] L total =L Δf +L RoCoF

[0100] Where, L Δf Represents the maximum frequency change loss function, L RoCoF Represents the maximum frequency change rate loss function.

[0101] Specifically, the maximum frequency change loss function L Δf and the maximum frequency change rate loss function L RoCoF , respectively expressed as:

[0102]

[0103] In the formula, It represents the predicted value of the maximum frequency change when the input of the frequency feature prediction model is the input data x in the sample data and the first network hyperparameter is θ1. represents the maximum frequency change rate prediction value when the input of the frequency feature prediction model is the input data x in the sample data and the first network hyperparameter is θ2, y1 and y2 represent the maximum frequency change amount data value and the maximum frequency change rate data value corresponding to the input data x in the sample data, ReLU(·) represents the activation function, α and β represent the first weight and the second weight, respectively; represents the expected value when the input data is x and the maximum frequency change is y1; wherein the input data is x and the maximum frequency change is y1, which satisfies the distribution of the divided sample data set D; Represents the expected value when the input data is x and the maximum frequency change is y2; wherein, the input data is x and the maximum frequency change is y2, which satisfies the distribution of the divided sample data set D.

[0104] It can be understood that the total loss function is formed by two loss functions, and the maximum frequency change amount and the maximum frequency change rate are predicted separately to avoid the influence of the system's single frequency modulation on the prediction.

[0105] S3. Obtain the day-ahead renewable energy output and load forecast curves, and then adjust the dispatch plan based on the unit combination optimization model, the optimal frequency characteristic prediction model, and the inertia adjustment constraint to obtain the inertia demand assessment result; wherein, the inertia demand assessment result includes the system total inertia level result and the corresponding dispatch plan.

[0106] During implementation, the inertia requirement assessment results are obtained through the following methods:

[0107] S31. Based on the recent renewable energy output and load forecast curve, the unit combination optimization model is used to obtain the basic dispatch plan;

[0108] S32, forming a one-dimensional sequence of the basic scheduling plan and inputting it into an optimal frequency characteristic prediction model to obtain frequency characteristic data;

[0109] S33. Determine whether the frequency characteristic data meets the frequency safety conditions. If so, take the total system inertia level of each time period in the current basic scheduling plan as the total system inertia level result, thereby obtaining the inertia demand assessment result; otherwise, adjust the current basic scheduling plan based on the inertia adjustment constraint and the unit combination optimization model to obtain the adjusted basic scheduling plan, and repeat steps S32 to S33.

[0110] During the specific implementation, the current basic scheduling plan is adjusted in the following ways:

[0111] Adjust the startup mode of each synchronous unit based on the inertia adjustment constraint;

[0112] Based on the adjusted startup mode of each synchronous unit, the unit commitment optimization model is adopted to obtain the adjusted basic scheduling plan.

[0113] Specifically, the inertia adjustment constraint is expressed as:

[0114]

[0115] In the formula, represents the startup mode of synchronous unit i in time period t in the pth iteration, It represents the startup mode of synchronous unit i in time period t in the p+1th iteration.

[0116] Specifically, the frequency safety condition is that the maximum frequency change rate is less than or equal to a set frequency change rate safety value and the maximum frequency change amount is less than or equal to a frequency change amount safety value.

[0117] More specifically, the frequency change rate safety value is the maximum frequency change rate that the system can bear, and the frequency change amount safety value is the maximum frequency change amount that the system can bear.

[0118] It can be understood that, by combining the inertia support characteristics of new energy stations and synchronous units, the unit combination model and inertia adjustment constraints are constructed. Its role is to adjust the synchronous unit startup mode at the corresponding moment when the frequency safety exceeds the limit in the unit combination scheme by adding this constraint, so that the new unit combination scheme takes into account the unit operation requirements while having a higher synchronous inertia level at the corresponding moment, so as to iteratively try to meet the system frequency safety. First, based on the day-ahead unit combination (the optimization of the start-stop status and output of each generator in the day-ahead scheduling plan), and combined with the results of the lower-level IC-CNN prediction model, the inertia adjustment constraints are added to the upper-level model to obtain the synchronous inertia demand of the system. Then, considering the inertia support capacity of the new energy station, the inertia level that meets the system requirements is obtained by changing the synchronous unit startup plan (that is, the plan for the start-stop status of each synchronous machine in the scheduling plan).

[0119] It should be noted that, in this embodiment, the maximum frequency change rate and the maximum frequency change amount are selected by the following derivation:

[0120] Firstly, the frequency response models of synchronous generator sets and new energy generator sets are established.

[0121] The frequency response characteristics of a synchronous generator set can be expressed by the generator swing equation:

[0122]

[0123] Where M sg is the inertia time constant in seconds; Δf is the system frequency deviation, ΔP e is the disturbance electromagnetic power, D sgis the load regulation coefficient, f is the system frequency, and t′ is the time after the self-disturbance occurs. sg Except for the nominal value, the rest are per-unit values.

[0124] Considering the speed regulator link of the synchronous generator, its frequency response characteristics are corrected to:

[0125]

[0126] In the formula, K sg 、F sg and T sg are all parameters of synchronous generator speed regulator, and s represents the Laplace operator.

[0127] Renewable energy power generation uses grid-connected virtual synchronous generator (VSG) technology, which simulates the response principle of the synchronous machine and connects to the grid in a voltage source manner, and participates in system frequency modulation based on its control strategy to provide virtual inertia. Unlike the inertia of the synchronous machine, which is its inherent physical property, when new energy is connected to the grid using VSG technology, the frequency response delay is caused by the measurement of electrical quantities, signal generation and transmission. The frequency change process after the system is disturbed can be expressed by the piecewise equivalent swing equation:

[0128]

[0129] Where, τ is the delay of virtual inertia response; E syn 、E vir They are system synchronization and virtual inertia level respectively; ΔP FR (t) is the total frequency modulation power of the system at time t; ΔP e (t) is the unbalanced power at time t; D sys is the system damping coefficient; f0 is the system rated frequency, and Δf(t) is the system frequency change.

[0130] Then, the frequency change process of the power system after the disturbance is analyzed, and it is divided into three main stages according to the regulation process: (A) inertia response stage; (B) inertia response and primary frequency regulation stage; (C) secondary frequency regulation stage. In the frequency response process, the maximum rate of change of frequency (RoCoF max ) and the lowest point are the key features of the frequency safety of the reaction system. Among them, inertia plays a role in the two stages A and B, so the following two features are selected as indicators for inertia demand assessment:

[0131] Maximum frequency change rate characteristics: A large maximum frequency change rate may cause mechanical damage to rotating components due to drastic changes in speed, island operation of the system, and stabilizer oscillation. In order to leave sufficient adjustment time and space for the subsequent frequency modulation, the following conditions must be met:

[0132]

[0133] In the formula, ΔP e is the disturbance unbalanced power, t0 is the time when the disturbance occurs, RoCoF max is the maximum frequency change rate that the system can withstand, E syn is the total synchronous inertia of the system, E vir The total virtual inertia provided for new energy stations.

[0134] Maximum frequency change characteristic: the difference between the steady-state frequency and the lowest frequency. To prevent the system from triggering low-frequency load shedding or even system crash due to excessive frequency drop, the following conditions must be met:

[0135] |Δf(t nadir )|≤Δf max

[0136] Where, t nadir is the formation time of the lowest frequency point, Δf(t nadir ) is at t nadir Maximum frequency change at a given moment, Δf max is the maximum frequency variation that the system can withstand.

[0137] Compared with the prior art, this embodiment provides a new method for evaluating inertia demand of a power system, which constructs a unit combination optimization model of a new power system, and then constructs an input data set and an output data set based on the historical data of the obtained new energy output and load forecast curve of the day before, and then trains the constructed frequency characteristic prediction model based on the sample data set composed of the input data set and the output data set to obtain the optimal frequency characteristic prediction model, thereby obtaining the new energy output and load forecast curve of the day before, adjusting the scheduling plan according to the unit combination optimization model, the optimal frequency characteristic prediction model and the inertia adjustment constraint, and obtaining the inertia demand evaluation result, effectively evaluating the system inertia level, adapting to different operating conditions and expected disturbance situations, solving the problem that the inertia demand caused by the access of new energy is difficult to accurately evaluate, and the solution is convenient and has strong adaptability; reducing the system synchronous inertia redundancy, improving the inertia support capacity of the system, and providing reference value for the formulation of the system's day-ahead scheduling plan.

[0138] Example 2

[0139] A specific embodiment of the present invention discloses a novel power system inertia demand assessment system, comprising:

[0140] A data set construction module is used to construct a unit combination optimization model for a new power system, and then construct an input data set and an output data set based on the historical data of the new energy output and load forecast curve obtained a day ago; wherein each element in the input data set is a one-dimensional sequence of the dispatch plan of the new power system, and each element in the output data set is the frequency characteristic data corresponding to each dispatch plan;

[0141] A frequency feature prediction model construction and training model is used to construct a frequency feature prediction model and train the frequency feature prediction model based on a sample data set to obtain an optimal frequency feature prediction model; wherein the sample data set includes an input data set and an output data set;

[0142] The inertia demand assessment module is used to obtain the day-ahead renewable energy output and load forecast curves, and then adjust the dispatch plan based on the unit combination optimization model, the optimal frequency characteristic prediction model, and the inertia adjustment constraint to obtain the inertia demand assessment result; among them, the inertia demand assessment result includes the system total inertia level result and the corresponding dispatch plan.

[0143] The specific implementation process of the embodiment of the present invention can be found in the above method embodiment, which will not be described in detail in this embodiment.

[0144] Since the principles of this embodiment are the same as those of the above method embodiment, the present device also has the corresponding technical effects of the above method embodiment.

[0145] Example 3

[0146] In order to verify the effectiveness and correctness of the solutions proposed in Example 1 and Example 2, a specific embodiment 3 of the present invention builds an improved IEEE-39 node system for verification; wherein the IEEE-39 node system consists of 10 generators and 39 buses; wherein the generators of nodes 34, 36, and 37 are replaced by wind farms with the same installed capacity, so that the system's new energy installed capacity penetration rate reaches 40%, and the virtual inertia constants are set to 6, 5, and 8s respectively. The system's reference power is 100MW, the reference voltage is 100kV, and the rated frequency is 60Hz.

[0147] First, the K-means clustering algorithm is used to aggregate historical new energy and load data to form a typical day, and 7,200 samples are iteratively generated through the unit combination optimization model. In the transient simulation, the disturbance is randomly selected as 5% to 10% of the total system load level and distributed to each node according to the node load rate; considering the adjustability of the virtual inertia control parameters, the VSG virtual inertia constant is randomly set to 4 to 10s. The load surge is set to occur at 2s and the simulation time is 25s. Based on the above-constructed data set, the IC-CNN prediction model is trained offline, and 180 training propagations are performed under the condition of a batch size of 32. The prediction results and simulation results obtained are shown in the figure. Figure 2 shown.

[0148] Affected by the conservative loss function, Figure 2 The predicted value of the medium frequency feature prediction model is slightly larger than the actual value. The mean square error (MSE) of the prediction results of the maximum frequency change and the maximum frequency change rate are 0.06% and 0.03% respectively, with high prediction accuracy. This shows that the deep learning algorithm based on the frequency feature prediction model can accurately learn the frequency change mechanism of the system.

[0149] Taking a sample of the test set as an example for verification. Considering the frequency security limit of the power system, RoCoF is set max The limit is 0.5Hz / s, the minimum frequency limit is 59.5Hz, and the power disturbance is set to 10% of the load in each period. The system inertia level is sufficient after the fifth iteration of the IRA-ICNN method. The iterative adjustment is as follows: Figure 3 and Figure 4 shown.

[0150] from Figure 3 It can be seen that, considering frequency safety, the system inertia level continues to increase with iteration. Further analysis of the inertia level adjustment at each moment shows that the adjustment changes are mainly concentrated in 6-11 and 16-21. This is because the wind power decreases at 6-11 due to the influence of wind speed, resulting in a decrease in its frequency regulation capability and the virtual inertia provided. The system compensates for the total inertia level shortfall by adding synchronous inertia to meet the larger inertia demand during the morning load peak. When the wind speed rises from 16-21, the wind power frequency regulation capability and the virtual inertia provided increase. At the same time, considering the increase in evening load, the total inertia level adjustment of the system is even greater.

[0151] Figure 4 The figure shows the changes of the lowest frequency point of the system under different iteration numbers. Among them, after the first iteration, the lowest frequency point of the system is seriously over the limit at 8 o'clock and 17-22 o'clock. After the fifth iteration, the lowest frequency point of the system at each moment is above the limit. At this time, the system inertia level meets the demand. This shows that the proposed inertia demand assessment method is effective.

[0152] Figure 5 The inertia demand level of the system at different time periods under different disturbance sizes set by load ratio is given. It can be seen that as the disturbance increases, the inertia demand level of the system at each moment continues to increase, and increases significantly during the peak load period. This is because the system faces greater frequency regulation pressure when facing disturbances during the peak load period, which increases the inertia demand. As the disturbance continues to increase, the system's inertia level adjustment ability approaches the upper limit.

[0153] The results show that the proposed method can accurately evaluate the system inertia demand level. Compared with the traditional method, it can reduce the system synchronous inertia redundancy and improve the system's inertia support capacity, providing a reference value for the formulation of the system's day-ahead scheduling plan.

[0154] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0155] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A novel method for evaluating inertia demand of power system, characterized in that: The following steps are involved: Construct a unit combination optimization model for a new power system, and then construct an input data set and an output data set based on the historical data of the new energy output and load forecast curve obtained the day before; wherein each element in the input data set is a one-dimensional sequence of the dispatch plan of the new power system, and each element in the output data set is the frequency characteristic data corresponding to each dispatch plan; Constructing a frequency feature prediction model, and training the frequency feature prediction model based on a sample data set to obtain an optimal frequency feature prediction model; wherein the sample data set includes an input data set and an output data set; The day-ahead renewable energy output and load forecast curves are obtained, and then the dispatch plan is adjusted based on the unit combination optimization model, the optimal frequency characteristic prediction model, and the inertia adjustment constraint to obtain the inertia demand assessment result; among them, the inertia demand assessment result includes the system total inertia level result and the corresponding dispatch plan.

2. The novel power system inertia demand assessment method according to claim 1 is characterized in that: The scheduling plan includes the output, startup mode, inertia constant of each synchronous unit in each period within the scheduling cycle, the output of each new energy unit, the virtual inertia constant of each VSG, the time constant of each speed regulator, the total inertia level of the system, the synchronous inertia level of the system, the virtual inertia level of the system and the total primary frequency regulation capability of the speed regulator; the frequency characteristic data includes the maximum frequency change and the maximum frequency change rate; wherein the maximum frequency change is the difference between the steady-state frequency of the system and the lowest point frequency.

3. The novel power system inertia demand assessment method according to claim 2 is characterized in that: The input and output datasets are constructed in the following way: Based on the historical data of the new energy output and load forecast curve obtained on the previous day, the unit combination optimization model is adopted to obtain the startup mode of each synchronous unit, the output of each synchronous unit and the actual output of each new energy unit in each period within the dispatch cycle; Based on the startup mode of each synchronous unit in each period of the scheduling cycle, the output of each synchronous unit and the actual output of each new energy unit, as well as the virtual inertia constant of each VSG, the inertia constant of each synchronous unit and the primary frequency regulation capacity of each speed regulator, the total system inertia level, the system synchronous inertia level, the system virtual inertia level and the total primary frequency regulation capacity of the speed regulator in each period of each scheduling cycle are obtained, and then each scheduling plan is obtained, and each scheduling plan is respectively constituted into a one-dimensional sequence to obtain an input data set; Based on the different disturbance scenarios and scheduling plans, the maximum frequency change and maximum frequency change rate corresponding to each scheduling plan under different disturbance scenarios are obtained through transient simulation, and the output data set is composed of frequency characteristic data.

4. The novel power system inertia demand assessment method according to claim 2 is characterized in that: The unit combination optimization model is to obtain the startup mode of each synchronous unit, the output of each synchronous unit and the actual output of each new energy unit in each period within the dispatching cycle under the operating constraints of the new power system and with the minimum operating cost as the objective function; wherein the objective function F is expressed as: F=min[f1+f2+f3+f4] In the formula, f1 represents the power cost of the synchronous unit, f2 represents the total startup cost of the synchronous unit, f3 represents the total shutdown cost of the synchronous unit, and f4 represents the penalty cost for abandoning the consumption of new energy.

5. The novel power system inertia demand assessment method according to claim 4 is characterized in that: The synchronous unit power cost f1, the synchronous unit switch startup cost f2, the synchronous unit switch shutdown cost f3, and the penalty cost f4 for abandoning new energy consumption are respectively expressed as: Where, T represents the total number of time periods; N represents the total number of synchronous units in the new power system; a i 、b i 、c i denote the first, second and third generation cost coefficients of synchronous unit i respectively; P i,t represents the output of synchronous unit i in period t; They represent the startup cost and shutdown cost of synchronous unit i respectively; u i,t 、u i,t-1 They represent the startup mode of synchronous unit i in period t and period t-1 respectively, with a value of 1 indicating startup and a value of 0 indicating shutdown; J represents the total number of new energy units in the new power system; c j They represent the power generation cost coefficient of the new energy unit j respectively; represents the predicted output of new energy unit j during period t on a typical day; P j,t Represents the actual output of new energy unit j during period t.

6. The novel power system inertia demand assessment method according to claim 1 is characterized in that: The training of the frequency feature prediction model based on the sample data set includes: Set the optimizer, training batch, learning rate, number of training iterations, and conservative loss function during training. The sample data set is divided based on training batches, and the frequency feature prediction model training of each training batch is performed in sequence based on the divided sample data set. When the stopping condition is met, the optimal frequency feature prediction model is obtained.

7. The novel power system inertia demand assessment method according to claim 6 is characterized in that: The conservative loss function L total It is expressed as: L total =L Δf +L RoCoF Where, L Δf Represents the maximum frequency change loss function, L RoCoF Represents the maximum frequency change rate loss function.

8. The novel power system inertia demand assessment method according to claim 2 is characterized in that: The inertia requirement assessment results are obtained by: S31. Based on the recent renewable energy output and load forecast curve, the unit combination optimization model is used to obtain the basic dispatch plan; S32, forming a one-dimensional sequence of the basic scheduling plan and inputting it into an optimal frequency characteristic prediction model to obtain frequency characteristic data; S33, judging whether the frequency characteristic data meets the frequency safety condition, if so, taking the system total inertia level of each time period in the current basic scheduling plan as the system total inertia level result, thereby obtaining the inertia demand assessment result; Otherwise, the current basic scheduling plan is adjusted based on the inertia adjustment constraint and the unit commitment optimization model to obtain an adjusted basic scheduling plan, and steps S32 to S33 are repeated.

9. The novel power system inertia demand assessment method according to claim 5 is characterized in that: The inertia adjustment constraint is expressed as: In the formula, represents the startup mode of synchronous unit i in time period t in the pth iteration, It represents the startup mode of synchronous unit i in time period t in the p+1th iteration.

10. A new type of power system inertia demand assessment system, characterized in that: include: A data set construction module is used to construct a unit combination optimization model for a new power system, and then construct an input data set and an output data set based on the historical data of the new energy output and load forecast curve obtained a day ago; wherein each element in the input data set is a one-dimensional sequence of the dispatch plan of the new power system, and each element in the output data set is the frequency characteristic data corresponding to each dispatch plan; A frequency feature prediction model construction and training model is used to construct a frequency feature prediction model and train the frequency feature prediction model based on a sample data set to obtain an optimal frequency feature prediction model; wherein the sample data set includes an input data set and an output data set; The inertia demand assessment module is used to obtain the day-ahead renewable energy output and load forecast curves, and then adjust the dispatch plan based on the unit combination optimization model, the optimal frequency characteristic prediction model, and the inertia adjustment constraint to obtain the inertia demand assessment result; among them, the inertia demand assessment result includes the system total inertia level result and the corresponding dispatch plan.

Images