Power system inertia safety area evaluation method based on BP neural network algorithm

Abstract

The invention discloses an electric power system inertia safe area evaluation method based on a BP neural network algorithm, and the method specifically comprises the following steps: S1, building an electric power system model, obtaining a frequency deviation and a frequency change rate under different new energy permeability, carrying out the normalization processing, and obtaining the equivalent inertia of an electric power system area; s2, calculating the critical inertia of the system, and comparing the regional equivalent inertia with the critical inertia of the system to obtain an inertia safety region; and S3, constructing an inertia safety area evaluation model based on a BP neural network algorithm, training the model by using the training set to obtain an optimal inertia safety area evaluation model, inputting to-be-tested data into the optimal inertia safety area evaluation model, and outputting whether the power system is an inertia safety area. The invention provides a BP neural network algorithm-based power system inertia safety area evaluation method, and solves the problems that the existing power system inertia evaluation method is complex and the obtained result is not visual.

Description

Technical Field

[0001] The invention belongs to the technical field of power system evaluation, and in particular relates to a power system inertia safety zone evaluation method based on a BP neural network algorithm. Background Art

[0002] Compared with the traditional power system dominated by synchronous power supply, low inertia has become one of the main features of the new power system. Therefore, in the low inertia scenario, the huge active impact caused by DC interlocking, new energy cluster disconnection, etc. is easy to trigger the action of automatic safety line devices such as low-frequency load reduction, and the system has a great safety risk. In recent years, accidents caused by insufficient inertia have occurred frequently, sounding the alarm for the safe and stable operation of the power grid under the access of a high proportion of new energy and power electronic equipment, and causing the power grid company to re-examine the current planning and operation guidelines / procedures.

[0003] Evaluating the system inertia helps to clarify the changing trend of system inertia under the future high proportion of new energy access, identify the system inertia danger zone in advance, and then prepare countermeasures to ensure the system frequency safety. Therefore, evaluating the inertia safety zone and danger zone in the power system has become a key scientific problem that needs to be solved under the construction of a new power system, and has important research value. At the system planning level, combined with the system inertia safety constraints, evaluating whether the inertia of the area has reached a safe level can provide a reference for the power grid planning department and provide a basis for determining the safe access ratio and control method of new energy, and the DC power carrying level; at the dispatching and operation level, identifying the inertia safety zone and danger zone will help operators to grasp the system inertia level under different operating scenarios in the future, determine the minimum startup scale of the synchronous power supply under different operating scenarios, and provide a basis for refined inertia control. Summary of the invention

[0004] The purpose of the present invention is to provide a power system inertia safety zone assessment method based on BP neural network algorithm, which solves the problem that the existing power system inertia assessment method is complicated and the obtained results are not intuitive.

[0005] The technical solution adopted by the present invention is a method for evaluating the inertia safety zone of a power system based on a BP neural network algorithm, and the specific steps are as follows:

[0006] S1. Establish a power system model, set different faults or disturbances under different renewable energy penetration rates, obtain the frequency deviation and frequency change rate at the generator connecting to the bus, and normalize the frequency deviation and frequency change rate to obtain the regional equivalent inertia of the power system;

[0007] S2. Based on the maximum frequency change rate and the maximum frequency deviation collected, the critical inertia of the system is obtained, and the area where the regional equivalent inertia is higher than the critical inertia of the system is determined as the inertia safety area, and the area where the regional equivalent inertia is lower than the critical inertia of the system is determined as the inertia danger area;

[0008] S3. Construct an inertia safety zone assessment model based on the BP neural network algorithm. Randomly divide the simulation data sets obtained under different scenarios into training sets and test sets in a ratio of 7:3. Use the training set to train the inertia safety zone assessment model based on the BP neural network algorithm to obtain the optimal inertia safety zone assessment model. Input the test data into the optimal inertia safety zone assessment model to output whether the power system is in the inertia safety zone.

[0009] The present invention is also characterized in that:

[0010] The specific method of setting different faults or disturbances in S1 and obtaining the frequency deviation and frequency change rate is as follows:

[0011] Set different proportions of machine shedding and load shedding faults, record the frequency curve, and obtain the system frequency deviation by subtracting it from the rated frequency of the power system;

[0012] The obtained frequency curve is differentiated to obtain the frequency change rate.

[0013] The method for calculating the system area equivalent inertia in S1 is as follows:

[0014] When the penetration rate of new energy is 0, a traditional power system model containing only the output of synchronous generators is established, and the traditional power system model is simulated in the time domain. Different faults or disturbances are set to obtain the traditional frequency deviation and frequency change rate, and the relationship between the equivalent generator inertia constant, system frequency deviation, frequency change rate and shortfall power is fitted, thereby obtaining the calculation formula of the equivalent generator inertia constant, that is, the equivalent inertia when the penetration rate of new energy is 0 is obtained;

[0015] When the renewable energy penetration rate is not 0, renewable energy units of equal capacity are used to replace synchronous generators, and a new power system model including renewable energy units is established. The new power system model is simulated in the time domain, and different faults or disturbances are set to obtain the frequency deviation and frequency change rate at the generator connected to the bus. The frequency deviation and frequency change rate are normalized, and the system area equivalent inertia is obtained according to the calculation formula of the equivalent generator inertia constant.

[0016] The calculation formula of equivalent generator inertia constant is:

[0017] When a power shortage occurs due to a fault disturbance, the frequency response process of the system is expressed by the equivalent rotor motion equation, that is, the equivalent generator inertia constant calculation formula:

[0018]

[0019] Where: H sys is the equivalent generator inertia constant, Δf(t) is the system frequency deviation, ΔP G (t), ΔP L (t) are the change of generator set output and load respectively, and the difference between them is the shortfall power; D is the damping coefficient of the system;

[0020] Because of the delayed frequency control measures of the power system, the extreme value of the frequency change rate is immediately after the disturbance occurs, when Δf(t) = 0, dΔf / dt| max =dΔf / dt| t＝0 , from formula (1) we can get:

[0021]

[0022] Where: P loss =|ΔP G (t)-ΔP L (t)| represents the maximum active unbalanced power. It can be seen from formula (2) that in P loss Under certain conditions, the larger the system equivalent inertia, the slower the frequency decreases.

[0023] The specific method for collecting the frequency and frequency change rate data at the generator connecting to the bus and normalizing them is:

[0024] Set a variety of load disturbances and equivalent generator inertia constant combinations, collect the frequency and frequency change rate data at the equivalent generator connected to the bus after the disturbance and normalize them, and use the maximum and minimum value normalization method to process the frequency and frequency change rate data, that is:

[0025]

[0026] In the formula, f ij and RoCoF ij are the normalized results of the frequency and frequency change rate data of the jth second of the i-th group; x ij and ij They are the original data of frequency and frequency change rate of the j-th second of the i-th group, 1≤i≤501, 1≤j≤501.

[0027] The method for obtaining the critical inertia of the system in S2 is as follows:

[0028] The initial frequency change rate of the power system represents the system's anti-disturbance level at the moment the disturbance occurs. When a disturbance occurs in the power system, the initial frequency change rate of the power system is calculated according to the initial value theorem:

[0029]

[0030] Among them, RoCoF max is the maximum frequency change rate;

[0031] When the system frequency change rate is zero, that is, when the system frequency reaches its peak value, the maximum frequency deviation of the system is expressed as:

[0032]

[0033] Where Δf max is the maximum frequency deviation expression; t peak The time when the system frequency reaches the peak value;

[0034] The frequency change rate of the system at the moment of disturbance is the largest, so RoCoF is selected. max The frequency dynamic constraint condition is limited to a certain limit value, and the minimum inertia requirement of the system is determined by obtaining the limit value and disturbance amount at the moment of disturbance; based on RoCoF max The minimum inertia requirement of the constrained system is Its expression is:

[0035]

[0036] Among them, ΔP max is the maximum disturbance of the system when the disturbance occurs instantly;

[0037] When the frequency drops to the lowest point, time t peak When the system frequency reaches the maximum deviation, the maximum frequency deviation Δf is selected. max Not exceeding a certain limit is a frequency dynamic constraint condition. The minimum inertia requirement under the maximum frequency deviation constraint is expressed as:

[0038] The critical inertia requirement is evaluated based on the frequency change rate and frequency deviation constraints. The critical inertia of the system is the maximum value of the inertia required under the two constraints:

[0039]

[0040] Where: H min is the critical inertia of the system.

[0041] The specific methods of S3 are:

[0042] Construct an inertia safety zone assessment model based on the BP neural network algorithm, select regional power disturbance, system frequency deviation, system frequency change rate, system maximum frequency deviation, and system maximum frequency change rate as the model input data, and select whether it is an inertia safety zone as the model output;

[0043] The output y of the neuron model of the BP neural network k The mathematical expression is:

[0044]

[0045] In the formula, y k is the output of the neuron model; x i is the input signal applied to the input terminal; m ij is the connection weight coefficient between the i-th variable and the j-th neuron; θ j is the threshold of the neuron; σ is the mapping relationship of the response function of the neuron; Σ is the spatial accumulation of the postsynaptic signal;

[0046] The BP neural network model consists of an input layer, a hidden layer, and an output layer. The network is trained and optimized through the back propagation algorithm, and the mapping relationship between input and output is approximated by adjusting the connection weights. During training, the network first transmits the input signal from the input layer to the output layer through forward propagation and calculates the output value of the network. Then, the output error is calculated using the back propagation algorithm. The total calculation expression is:

[0047]

[0048] Where E is the sample error, T i is the expected output, y k is the output of the neuron model;

[0049] The error is propagated back from the output layer to the hidden layer and input layer. According to the error, the weight of each connection is adjusted through optimization algorithms such as gradient descent to reduce the error and improve the performance of the network. The optimized and corrected input weight calculation formula can be expressed as:

[0050]

[0051] In the formula, Δm ij is the connection weight coefficient, η is the learning rate, and E is the sample error;

[0052] The output weight calculation formula can be expressed as:

[0053]

[0054] In the formula, Δn jk is the connection weight coefficient;

[0055] The inertia safety zone assessment model based on the BP neural network algorithm is trained using the training set. After each iterative training, the model performance is evaluated using the test set until the error index is met and the optimal model parameters are obtained, thus obtaining the inertia safety zone assessment model based on the BP neural network algorithm.

[0056] The beneficial effects of the present invention are:

[0057] The power system inertia safety zone assessment method based on the BP neural network algorithm provided by the present invention establishes a power system model, obtains the frequency deviation and frequency change rate under different new energy penetration rates, obtains the power system regional equivalent inertia, and compares it with the system critical inertia, thereby screening out the inertia safety zone and the inertia danger zone, and constructs an inertia safety zone assessment model based on the BP neural network algorithm, trains and optimizes the model, so that the inertia safety zone can be output simply and quickly. The entire method process can calculate the power system equivalent inertia and critical inertia, assess the power system inertia safety zone, find out the weak links in the power system inertia, and provide a theoretical basis for the subsequent inertia improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is the overall flow chart of the inertia safety zone evaluation method of the present invention;

[0059] Figure 2 It is a specific flow chart of the system inertia safety zone determination method based on the dual frequency constraint index of the evaluation method of the present invention;

[0060] Figure 3 This is a schematic diagram of the BP neural network model structure of the inertia safety zone assessment method of the present invention;

[0061] Figure 4 The inertia safety zone assessment method of the present invention does not include the IEEE39 node system topology diagram of the new energy unit;

[0062] Figure 5 The inertia safety zone assessment method of the present invention includes an IEEE39 node system topology diagram of a new energy unit. DETAILED DESCRIPTION

[0063] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0064] The power system inertia safety zone assessment method based on BP neural network algorithm, such as Figure 1 As shown, the specific steps are as follows:

[0065] S1. Establish a power system model, set different faults or disturbances under different renewable energy penetration rates, obtain the frequency deviation and frequency change rate at the generator connecting to the bus, and normalize the frequency deviation and frequency change rate to obtain the regional equivalent inertia of the power system;

[0066] The specific method of setting different faults or disturbances and obtaining the frequency deviation and frequency change rate is as follows: setting different proportions of machine shedding and load shedding faults, recording the frequency curve, and obtaining the system frequency deviation by subtracting it from the rated frequency of the power system;

[0067] The obtained frequency curve is differentiated to obtain the frequency change rate.

[0068] Further, such as Figure 2 As shown in the figure, the method for calculating the system area equivalent inertia is as follows:

[0069] When the penetration rate of new energy is 0, a traditional power system model containing only the output of synchronous generators is established, and the traditional power system model is simulated in the time domain. Different faults or disturbances are set to obtain the traditional frequency deviation and frequency change rate, and the relationship between the equivalent generator inertia constant, system frequency deviation, frequency change rate and shortfall power is fitted, thereby obtaining the calculation formula of the equivalent generator inertia constant, that is, the equivalent inertia when the penetration rate of new energy is 0 is obtained;

[0070] Among them, the calculation formula of the equivalent generator inertia constant is:

[0071] When a power shortage occurs due to a fault disturbance, the frequency response process of the system is expressed by the equivalent rotor motion equation, that is, the equivalent generator inertia constant calculation formula:

[0072]

[0073] Where: H sys is the equivalent generator inertia constant, Δf(t) is the system frequency deviation, ΔP G (t), ΔP L (t) are the change of generator set output and load respectively, and the difference between them is the shortfall power; D is the damping coefficient of the system;

[0074] Because of the delayed frequency control measures of the power system, the extreme value of the frequency change rate is immediately after the disturbance occurs, when Δf(t) = 0, dΔf / dt| max =dΔf / dt| t＝0 , from formula (1) we can get:

[0075]

[0076] Where: P loss =|ΔP G (t)-ΔP L (t)| represents the maximum active unbalanced power. It can be seen from formula (2) that in P loss Under certain conditions, the larger the system equivalent inertia, the slower the frequency decreases.

[0077] When the renewable energy penetration rate is not 0, renewable energy units of equal capacity are used to replace synchronous generators, and a new power system model including renewable energy units is established. The new power system model is simulated in the time domain, and different faults or disturbances are set to obtain the frequency deviation and frequency change rate at the generator connected to the bus. The frequency deviation and frequency change rate are normalized, and the system area equivalent inertia is obtained according to the calculation formula of the equivalent generator inertia constant.

[0078] Furthermore, the specific method for collecting the frequency and frequency change rate data at the generator connected to the bus and normalizing them is as follows:

[0079] Set a variety of load disturbances and equivalent generator inertia constant combinations, collect the frequency and frequency change rate data at the equivalent generator connected to the bus after the disturbance and normalize them, and use the maximum and minimum value normalization method to process the frequency and frequency change rate data, that is:

[0080]

[0081] In the formula, f ij and RoCoF ij are the normalized results of the frequency and frequency change rate data of the jth second of the i-th group; x ij and ij They are the original data of frequency and frequency change rate of the j-th second of the i-th group, 1≤i≤501, 1≤j≤501.

[0082] S2. Based on the maximum frequency change rate and the maximum frequency deviation collected, the critical inertia of the system is obtained, and the area where the regional equivalent inertia is higher than the critical inertia of the system is determined as the inertia safety area, and the area where the regional equivalent inertia is lower than the critical inertia of the system is determined as the inertia danger area;

[0083] Among them, the method of obtaining the critical inertia of the system is as follows:

[0084] The initial frequency change rate of the power system represents the system's anti-disturbance level at the moment the disturbance occurs. When a disturbance occurs in the power system, the initial frequency change rate of the power system is calculated according to the initial value theorem:

[0085]

[0086] Among them, RoCoF max is the maximum frequency change rate;

[0087] When the system frequency change rate is zero, that is, when the system frequency reaches its peak value, the maximum frequency deviation of the system is expressed as:

[0088]

[0089] Where Δfmax is the maximum frequency deviation expression; t peak The time when the system frequency reaches the peak value;

[0090] The frequency change rate of the system at the moment of disturbance is the largest, so RoCoF is selected. max The frequency dynamic constraint condition is limited to a certain limit value, and the minimum inertia requirement of the system is determined by obtaining the limit value and disturbance amount at the moment of disturbance; based on RoCoF max The minimum inertia requirement of the constrained system is Its expression is:

[0091]

[0092] Among them, ΔP max is the maximum disturbance of the system when the disturbance occurs instantly;

[0093] When the frequency drops to the lowest point, time t peak When the system frequency reaches the maximum deviation, the maximum frequency deviation Δf is selected. max Not exceeding a certain limit is a frequency dynamic constraint condition. The minimum inertia requirement under the maximum frequency deviation constraint is expressed as:

[0094] The critical inertia requirement is evaluated based on the frequency change rate and frequency deviation constraints. The critical inertia of the system is the maximum value of the inertia required under the two constraints:

[0095]

[0096] Where: H min is the critical inertia of the system.

[0097] S3. Construct an inertia safety zone assessment model based on the BP neural network algorithm. The simulation data sets (including regional power disturbance, system frequency deviation, system frequency change rate, system maximum frequency deviation, and system maximum frequency change rate) obtained under different scenarios are randomly divided into a training set and a test set at a ratio of 7:3. The inertia safety zone assessment model based on the BP neural network algorithm is trained using the training set to obtain the optimal inertia safety zone assessment model. The data to be tested is input into the optimal inertia safety zone assessment model to output whether the power system is in the inertia safety zone. The specific method is as follows:

[0098] Construct an inertia safety zone assessment model based on BP neural network algorithm, such as Figure 3 As shown in the figure, the regional power disturbance, system frequency deviation, system frequency change rate, system maximum frequency deviation, and system maximum frequency change rate are selected as the input data of the model, and whether it is an inertia safety zone is selected as the model output. The output y of the BP neural network neuron model is k The mathematical expression is:

[0099]

[0100] In the formula, y k is the output of the neuron model; x i is the input signal applied to the input terminal; m ij is the connection weight coefficient between the i-th variable and the j-th neuron; θ j is the threshold of the neuron; σ is the mapping relationship of the response function of the neuron; Σ is the spatial accumulation of the postsynaptic signal;

[0101] The BP neural network model consists of an input layer, a hidden layer, and an output layer. The network is trained and optimized through the back propagation algorithm, and the mapping relationship between input and output is approximated by adjusting the connection weights. During training, the network first transmits the input signal from the input layer to the output layer through forward propagation and calculates the output value of the network. Then, the output error is calculated using the back propagation algorithm. The total calculation expression is:

[0102]

[0103] Where E is the sample error, T i is the expected output, y k is the output of the neuron model.

[0104] The error is propagated back from the output layer to the hidden layer and input layer. According to the error, the weight of each connection is adjusted through optimization algorithms such as gradient descent to reduce the error and improve the performance of the network. The optimized and corrected input weight calculation formula can be expressed as:

[0105]

[0106] In the formula, Δm ij is the connection weight coefficient, η is the learning rate, and E is the sample error.

[0107] The output weight calculation formula can be expressed as:

[0108]

[0109] In the formula, Δn jk is the connection weight coefficient.

[0110] The inertia safety zone assessment model based on the BP neural network algorithm is trained using the training set. After each iterative training, the model performance is evaluated using the test set until the error index is met and the optimal model parameters are obtained, thus obtaining the inertia safety zone assessment model based on the BP neural network algorithm.

[0111] In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments:

[0112] In the present invention, a frequency response model of a power system including a new energy unit is established based on Matlab / Simulink. The model adopts an IEEE39 node system that does not include a new energy unit. Figure 4 This is the topology diagram of the IEEE39 node system built in the present invention that does not include new energy units. The system has 10 generators, 39 buses and 46 lines. The total power of the generators is 6000MW, the rated frequency is 60Hz, and the rated capacity and inertia constant of each generator are shown in Table 1.

[0113] Table 1 Generator parameters

[0114]

[0115]

[0116] The total inertia of the system can be calculated from Table 1. The system power shortage is set to 300MW. According to the method in step 1, the relationship between the equivalent generator inertia constant, frequency deviation, frequency change rate and power shortage can be fitted.

[0117] At each generator node, the synchronous generator set is replaced by a new energy unit of equal capacity, where the capacity of each new energy station is 180MW, and 3060MW, 2700MW and 2340MW of new energy are set respectively, with corresponding penetration rates of 40%, 45% and 50%. Figure 5 This is the IEEE39 node system topology diagram including the new energy units built in the present invention. The system power deficit is set to 300MW. The system inertia under the corresponding penetration scenario can be calculated according to the above formula.

[0118] The method of the present invention uses the system equivalent inertia time constant to dynamically characterize the system inertia demand level, and realizes the evaluation of the minimum inertia demand through the frequency index constraint. The key is to first set the system disturbance, perform simulation operation and obtain the frequency and frequency change rate, and obtain the system inertia according to the equivalent inertia time constant formula in step 1. By setting the maximum frequency change rate and frequency deviation constraints, the corresponding minimum inertia demand is obtained, and the larger value of the inertia time constant under the frequency index constraint is taken as the minimum inertia demand of the system. By comparing the system inertia and the minimum inertia demand under different penetration rate scenarios, it is concluded whether the inertia level of the system is safe.

[0119] A BP neural network model for evaluating the inertia safety zone of a new power system is established. The system's generator output, new energy output, frequency offset, frequency change rate, system power shortage, and frequency index constraint are used as input data, and whether the system is in a safe zone is used as output data. The data format of the neural network model is shown in Table 2 below.

[0120] Table 2 Neural network model data format

[0121]

[0122] The present invention simulates four penetration scenarios with power shortages of 100MW to 500MW and a step size of 10MW to obtain relevant data sets. The input data is randomly divided into training sets and test sets in proportion, and the training set is input into the inertia safety zone assessment model based on BP neural network for training. When the test error meets the requirements, the training is stopped and the training model is saved. The test set accuracy obtained in this example can reach more than 97%, which can meet the requirements of power system inertia safety zone assessment and has practical application value.

[0123] Example 1

[0124] The specific steps of the power system inertia safety zone assessment method based on BP neural network algorithm are as follows:

[0125] S1. Establish a power system model, set different faults or disturbances under different renewable energy penetration rates, obtain the frequency deviation and frequency change rate at the generator connecting to the bus, and normalize the frequency deviation and frequency change rate to obtain the regional equivalent inertia of the power system;

[0126] S2. Based on the maximum frequency change rate and the maximum frequency deviation collected, the critical inertia of the system is obtained, and the area where the regional equivalent inertia is higher than the critical inertia of the system is determined as the inertia safety area, and the area where the regional equivalent inertia is lower than the critical inertia of the system is determined as the inertia danger area;

[0127] S3. Construct an inertia safety zone assessment model based on the BP neural network algorithm. Randomly divide the simulation data sets obtained under different scenarios into training sets and test sets in a ratio of 7:3. Use the training set to train the inertia safety zone assessment model based on the BP neural network algorithm to obtain the optimal inertia safety zone assessment model. Input the test data into the optimal inertia safety zone assessment model to output whether the power system is in the inertia safety zone.

[0128] Example 2

[0129] The specific steps of the power system inertia safety zone assessment method based on BP neural network algorithm are as follows:

[0130] S1. Establish a power system model, set different faults or disturbances under different renewable energy penetration rates, obtain the frequency deviation and frequency change rate at the generator connecting to the bus, and normalize the frequency deviation and frequency change rate to obtain the regional equivalent inertia of the power system;

[0131] The specific method of setting different faults or disturbances and obtaining the frequency deviation and frequency change rate is as follows: setting different proportions of machine shedding and load shedding faults, recording the frequency curve, and obtaining the system frequency deviation by subtracting it from the rated frequency of the power system;

[0132] The obtained frequency curve is differentiated to obtain the frequency change rate.

[0133] Furthermore, the method for calculating the system regional equivalent inertia is as follows:

[0134] When the penetration rate of new energy is 0, a traditional power system model containing only the output of synchronous generators is established, and the traditional power system model is simulated in the time domain. Different faults or disturbances are set to obtain the traditional frequency deviation and frequency change rate, and the relationship between the equivalent generator inertia constant, system frequency deviation, frequency change rate and shortfall power is fitted, thereby obtaining the calculation formula of the equivalent generator inertia constant, that is, the equivalent inertia when the penetration rate of new energy is 0 is obtained;

[0135] Among them, the calculation formula of the equivalent generator inertia constant is:

[0136] When a power shortage occurs due to a fault disturbance, the frequency response process of the system is expressed by the equivalent rotor motion equation, that is, the equivalent generator inertia constant calculation formula:

[0137]

[0138] Where: H sys is the equivalent generator inertia constant, Δf(t) is the system frequency deviation, ΔP G (t), ΔP L (t) are the change of generator set output and load respectively, and the difference between them is the shortfall power; D is the damping coefficient of the system;

[0139] Because of the delayed frequency control measures of the power system, the extreme value of the frequency change rate is immediately after the disturbance occurs, when Δf(t) = 0, dΔf / dt| max =dΔf / dt| t＝0 , from formula (1) we can get:

[0140]

[0141] Where: P loss =|ΔP G (t)-ΔP L (t)| represents the maximum active unbalanced power. It can be seen from formula (2) that in P loss Under certain conditions, the larger the system equivalent inertia, the slower the frequency decreases.

[0142] When the renewable energy penetration rate is not 0, renewable energy units of equal capacity are used to replace synchronous generators, and a new power system model including renewable energy units is established. The new power system model is simulated in the time domain, and different faults or disturbances are set to obtain the frequency deviation and frequency change rate at the generator connected to the bus. The frequency deviation and frequency change rate are normalized, and the system area equivalent inertia is obtained according to the calculation formula of the equivalent generator inertia constant.

[0143] Furthermore, the specific method for collecting the frequency and frequency change rate data at the generator connected to the bus and normalizing them is as follows:

[0144] Set a variety of load disturbances and equivalent generator inertia constant combinations, collect the frequency and frequency change rate data at the equivalent generator connected to the bus after the disturbance and normalize them, and use the maximum and minimum value normalization method to process the frequency and frequency change rate data, that is:

[0145]

[0146] In the formula, f ij and RoCoF ij are the normalized results of the frequency and frequency change rate data of the jth second of the i-th group; x ij and ij They are the original data of frequency and frequency change rate of the j-th second of the i-th group, 1≤i≤501, 1≤j≤501.

[0147] S2. Based on the maximum frequency change rate and the maximum frequency deviation collected, the critical inertia of the system is obtained, and the area where the regional equivalent inertia is higher than the critical inertia of the system is determined as the inertia safety area, and the area where the regional equivalent inertia is lower than the critical inertia of the system is determined as the inertia danger area;

[0148] S3. Construct an inertia safety zone assessment model based on the BP neural network algorithm. Randomly divide the simulation data sets obtained under different scenarios into training sets and test sets in a ratio of 7:3. Use the training set to train the inertia safety zone assessment model based on the BP neural network algorithm to obtain the optimal inertia safety zone assessment model. Input the test data into the optimal inertia safety zone assessment model to output whether the power system is in the inertia safety zone.

[0149] Example 3

[0150] The specific steps of the power system inertia safety zone assessment method based on BP neural network algorithm are as follows:

[0151] S1. Establish a power system model, set different faults or disturbances under different renewable energy penetration rates, obtain the frequency deviation and frequency change rate at the generator connecting to the bus, and normalize the frequency deviation and frequency change rate to obtain the regional equivalent inertia of the power system;

[0152] S2. Based on the maximum frequency change rate and the maximum frequency deviation collected, the critical inertia of the system is obtained, and the area where the regional equivalent inertia is higher than the critical inertia of the system is determined as the inertia safety area, and the area where the regional equivalent inertia is lower than the critical inertia of the system is determined as the inertia danger area;

[0153] Among them, the method of obtaining the critical inertia of the system is as follows:

[0154] The initial frequency change rate of the power system represents the system's anti-disturbance level at the moment the disturbance occurs. When a disturbance occurs in the power system, the initial frequency change rate of the power system is calculated according to the initial value theorem:

[0155]

[0156] Among them, RoCoF max is the maximum frequency change rate;

[0157] When the system frequency change rate is zero, that is, when the system frequency reaches its peak value, the maximum frequency deviation of the system is expressed as:

[0158]

[0159] Where Δf max is the maximum frequency deviation expression; t peak The time when the system frequency reaches the peak value;

[0160] The frequency change rate of the system at the moment of disturbance is the largest, so RoCoF is selected. max The frequency dynamic constraint condition is limited to a certain limit value, and the minimum inertia requirement of the system is determined by obtaining the limit value and disturbance amount at the moment of disturbance; based on RoCoF max The minimum inertia requirement of the constrained system is Its expression is:

[0161]

[0162] Among them, ΔP max is the maximum disturbance of the system when the disturbance occurs instantly;

[0163] When the frequency drops to the lowest point, time t peak When the system frequency reaches the maximum deviation, the maximum frequency deviation Δf is selected. maxNot exceeding a certain limit is a frequency dynamic constraint condition. The minimum inertia requirement under the maximum frequency deviation constraint is expressed as:

[0164] The critical inertia requirement is evaluated based on the frequency change rate and frequency deviation constraints. The critical inertia of the system is the maximum value of the inertia required under the two constraints:

[0165]

[0166] Where: H min is the critical inertia of the system.

[0167] S3. Construct an inertia safety zone assessment model based on the BP neural network algorithm. The simulation data sets obtained under different scenarios are randomly divided into training sets and test sets at a ratio of 7:3. The inertia safety zone assessment model based on the BP neural network algorithm is trained using the training set to obtain the optimal inertia safety zone assessment model. The data to be tested is input into the optimal inertia safety zone assessment model to output whether the power system is in the inertia safety zone. The specific method is as follows:

[0168] Construct an inertia safety zone assessment model based on BP neural network algorithm, such as Figure 3 As shown in the figure, the regional power disturbance, system frequency deviation, system frequency change rate, system maximum frequency deviation, and system maximum frequency change rate are selected as the input data of the model, and whether it is an inertia safety zone is selected as the model output. The output y of the BP neural network neuron model is k The mathematical expression is:

[0169]

[0170] In the formula, y k is the output of the neuron model; x i is the input signal applied to the input terminal; m ij is the connection weight coefficient between the i-th variable and the j-th neuron; θ j is the threshold of the neuron; σ is the mapping relationship of the response function of the neuron; Σ is the spatial accumulation of the postsynaptic signal;

[0171] The BP neural network model consists of an input layer, a hidden layer, and an output layer. The network is trained and optimized through the back propagation algorithm, and the mapping relationship between input and output is approximated by adjusting the connection weights. During training, the network first transmits the input signal from the input layer to the output layer through forward propagation and calculates the output value of the network. Then, the output error is calculated using the back propagation algorithm. The total calculation expression is:

[0172]

[0173] Where E is the sample error, Ti is the expected output, y k is the output of the neuron model.

[0174] The error is propagated back from the output layer to the hidden layer and input layer. According to the error, the weight of each connection is adjusted through optimization algorithms such as gradient descent to reduce the error and improve the performance of the network. The optimized and corrected input weight calculation formula can be expressed as:

[0175]

[0176] In the formula, Δm ij is the connection weight coefficient, η is the learning rate, and E is the sample error.

[0177] The output weight calculation formula can be expressed as:

[0178]

[0179] In the formula, Δn jk is the connection weight coefficient.

[0180] The inertia safety zone assessment model based on the BP neural network algorithm is trained using the training set. After each iterative training, the model performance is evaluated using the test set until the error index is met and the optimal model parameters are obtained, thus obtaining the inertia safety zone assessment model based on the BP neural network algorithm.

Claims

1. A power system inertia safety zone assessment method based on BP neural network algorithm, characterized in that: The specific steps are as follows: S1. Establish a power system model, set different faults or disturbances under different renewable energy penetration rates, obtain the frequency deviation and frequency change rate at the generator connecting to the bus, and normalize the frequency deviation and frequency change rate to obtain the regional equivalent inertia of the power system; S2. Based on the maximum frequency change rate and the maximum frequency deviation collected, the critical inertia of the system is obtained, and the area where the regional equivalent inertia is higher than the critical inertia of the system is determined as the inertia safety area, and the area where the regional equivalent inertia is lower than the critical inertia of the system is determined as the inertia danger area; S3. Construct an inertia safety zone assessment model based on the BP neural network algorithm. Randomly divide the simulation data sets obtained under different scenarios into training sets and test sets in a ratio of 7:

3. Use the training set to train the inertia safety zone assessment model based on the BP neural network algorithm to obtain the optimal inertia safety zone assessment model. Input the test data into the optimal inertia safety zone assessment model to output whether the power system is in the inertia safety zone.

2. The method for evaluating the inertia safety zone of a power system based on the BP neural network algorithm according to claim 1 is characterized in that: The specific method of setting different faults or disturbances in S1 to obtain the frequency deviation and frequency change rate is as follows: Set different proportions of machine shedding and load shedding faults, record the frequency curve, and obtain the system frequency deviation by subtracting it from the rated frequency of the power system; The obtained frequency curve is differentiated to obtain the frequency change rate.

3. The method for evaluating the inertia safety zone of a power system based on the BP neural network algorithm according to claim 2 is characterized in that: The method for calculating the system area equivalent inertia in S1 is as follows: When the penetration rate of new energy is 0, a traditional power system model containing only the output of synchronous generators is established, and the traditional power system model is simulated in the time domain. Different faults or disturbances are set to obtain the traditional frequency deviation and frequency change rate, and the relationship between the equivalent generator inertia constant, system frequency deviation, frequency change rate and shortfall power is fitted, thereby obtaining the calculation formula of the equivalent generator inertia constant, that is, the equivalent inertia when the penetration rate of new energy is 0 is obtained; When the renewable energy penetration rate is not 0, renewable energy units of equal capacity are used to replace synchronous generators, and a new power system model including renewable energy units is established. The new power system model is simulated in the time domain, and different faults or disturbances are set to obtain the frequency deviation and frequency change rate at the generator connected to the bus. The frequency deviation and frequency change rate are normalized, and the system area equivalent inertia is obtained according to the calculation formula of the equivalent generator inertia constant.

4. The method for evaluating the inertia safety zone of a power system based on the BP neural network algorithm according to claim 3 is characterized in that: The equivalent generator inertia constant calculation formula is: When a power shortage occurs due to a fault disturbance, the frequency response process of the system is expressed by the equivalent rotor motion equation, that is, the equivalent generator inertia constant calculation formula: Where: H sys is the equivalent generator inertia constant, Δf(t) is the system frequency deviation, ΔP G (t), ΔP L (t) are the change of generator set output and load respectively, and the difference between them is the shortfall power; D is the damping coefficient of the system; Because of the delayed frequency control measures of the power system, the extreme value of the frequency change rate is immediately after the disturbance occurs, when Δf(t) = 0, dΔf / dt| max =dΔf / dt| t＝0 , from formula (1) we can get: Where: P loss =|ΔP G (t)-ΔP L (t)| represents the maximum active unbalanced power. It can be seen from formula (2) that in P loss Under certain conditions, the larger the system equivalent inertia, the slower the frequency decreases.

5. The method for evaluating the inertia safety zone of a power system based on the BP neural network algorithm according to claim 3 is characterized in that: The specific method of collecting the frequency and frequency change rate data at the generator connecting to the bus and normalizing them is: Set a variety of load disturbances and equivalent generator inertia constant combinations, collect the frequency and frequency change rate data at the equivalent generator connected to the bus after the disturbance and normalize them, and use the maximum and minimum value normalization method to process the frequency and frequency change rate data, that is: In the formula, f ij and RoCoF ij are the normalized results of the frequency and frequency change rate data of the jth second of the i-th group; x ij and ij They are the original data of frequency and frequency change rate of the j-th second of the i-th group, 1≤i≤501, 1≤j≤501.

6. The method for evaluating the inertia safety zone of a power system based on the BP neural network algorithm according to claim 5 is characterized in that: The method for obtaining the critical inertia of the system in S2 is as follows: The initial frequency change rate of the power system represents the system's anti-disturbance level at the moment the disturbance occurs. When a disturbance occurs in the power system, the initial frequency change rate of the power system is calculated according to the initial value theorem: Among them, RoCoF max is the maximum frequency change rate; When the system frequency change rate is zero, that is, when the system frequency reaches its peak value, the maximum frequency deviation of the system is expressed as: Where Δf max is the maximum frequency deviation expression; t peak The time when the system frequency reaches the peak value; The frequency change rate of the system at the moment of disturbance is the largest, so RoCoF is selected. max The frequency dynamic constraint condition is limited to a certain limit value, and the minimum inertia requirement of the system is determined by obtaining the limit value and disturbance amount at the moment of disturbance; based on RoCoF max The minimum inertia requirement of the constrained system is Its expression is: Among them, ΔP max is the maximum disturbance of the system when the disturbance occurs instantly; When the frequency drops to the lowest point, time t peak When the system frequency reaches the maximum deviation, the maximum frequency deviation Δf is selected. max Not exceeding a certain limit is a frequency dynamic constraint condition. The minimum inertia requirement under the maximum frequency deviation constraint is expressed as: The critical inertia requirement is evaluated based on the frequency change rate and frequency deviation constraints. The critical inertia of the system is the maximum value of the inertia required under the two constraints: Where: H min is the critical inertia of the system.

7. The method for evaluating the inertia safety zone of a power system based on the BP neural network algorithm according to claim 6 is characterized in that: The specific method of S3 is: Construct an inertia safety zone assessment model based on the BP neural network algorithm, select regional power disturbance, system frequency deviation, system frequency change rate, system maximum frequency deviation, and system maximum frequency change rate as the model input data, and select whether it is an inertia safety zone as the model output; The output y of the neuron model of the BP neural network k The mathematical expression is: In the formula, y k is the output of the neuron model; x i is the input signal applied to the input terminal; m ij is the connection weight coefficient between the i-th variable and the j-th neuron; θ j is the threshold of the neuron; σ is the response function mapping relationship of the neuron; ∑ is the spatial accumulation of postsynaptic signals; The BP neural network model consists of an input layer, a hidden layer, and an output layer. The network is trained and optimized through the back propagation algorithm, and the mapping relationship between input and output is approximated by adjusting the connection weights. During training, the network first transmits the input signal from the input layer to the output layer through forward propagation and calculates the output value of the network. Then, the output error is calculated using the back propagation algorithm. The total calculation expression is: Where E is the sample error, T i is the expected output, y k is the output of the neuron model; The error is propagated back from the output layer to the hidden layer and input layer. According to the error, the weight of each connection is adjusted through optimization algorithms such as gradient descent to reduce the error and improve the performance of the network. The optimized and corrected input weight calculation formula can be expressed as: In the formula, Δm ij is the connection weight coefficient, η is the learning rate, and E is the sample error; The output weight calculation formula can be expressed as: In the formula, Δn jk is the connection weight coefficient; The inertia safety zone assessment model based on the BP neural network algorithm is trained using the training set. After each iterative training, the model performance is evaluated using the test set until the error index is met and the optimal model parameters are obtained, thus obtaining the inertia safety zone assessment model based on the BP neural network algorithm.

Images