A data-model hybrid driven power system security assessment method and system

Abstract

The application provides a data-model hybrid driving power system safety evaluation method and system, and belongs to the field of power system safety evaluation. The power system safety evaluation method comprises the following steps: collecting a multi-time scale data set of a power grid dispatching platform; predicting the day-ahead load of the power system according to historical load data to obtain day-ahead load prediction data; performing power flow calculation on the power system according to the multi-time scale data set and the day-ahead load prediction data; determining the day-ahead safety evaluation index value of the power system based on a safety evaluation model according to the day-ahead fault type, the day-ahead power grid topology structure and the day-ahead power flow calculation result; performing power flow calculation on the power system according to real-time load; and determining the real-time safety evaluation index value of the power system based on the safety evaluation model according to the current fault type, the current power grid topology structure and the real-time power flow calculation result. The efficiency of power system safety evaluation is improved.

Description

Technical Field

[0001] This invention relates to the field of power system security assessment, and in particular to a data-model hybrid driven method and system for power system security assessment. Background Technology

[0002] Against the backdrop of a low-carbon transformation of the energy structure, building a new power system dominated by new energy sources will become an important means to achieve the goal of "carbon peaking and carbon neutrality," transforming the power system from a deterministic system to a highly uncertain system. On the power generation side, a high proportion of renewable energy becomes a major characteristic of the new power system. Unlike conventional hydropower and thermal power, renewable energy generation is affected by meteorological conditions and environmental factors, exhibiting intermittent and fluctuating output. Large-scale renewable energy integration introduces significant uncertainty into power system operation. On the load side, with the widespread integration of electric vehicles, increasingly frequent supply and demand interactions, and the development of photovoltaic and energy storage on the user side, loads exhibit proactive and complex characteristics. On the grid side, the transmission network is affected by random source loads, resulting in large-scale power flow fluctuations. The impact of random disturbances or faults on the power system mainly manifests as branch overloads and node voltage exceedances, which can impact the grid. While random power flow results directly provide data on branch power flow and node voltage, the system's safety level cannot be judged solely from these two aspects of data. To ensure the safe and stable operation of the system, it has become increasingly urgent to accelerate the establishment of power grid analysis and calculation methods, operation safety assessment methods and related standards that are adapted to this new environment.

[0003] The massive influx of new energy sources and novel power electronic devices has driven the construction of new power systems, but it has also greatly increased the complexity of energy and information flows in the system, leading to a surge in the calculation time for power system security assessment. This has a significant impact on large-scale actual power system dispatch. Summary of the Invention

[0004] The purpose of this invention is to provide a data-model hybrid driven method and system for power system security assessment, which can improve the efficiency of power system security assessment.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A data-model hybrid-driven method for power system security assessment includes:

[0007] Collect multi-timescale datasets from the power grid dispatching platform; the multi-timescale datasets include planned data and daily incidental fault data; the planned data includes the power grid topology at the monthly timescale, the unit on / off status at the weekly timescale, and the power grid operation data at the daily timescale; the daily incidental fault data includes the power grid topology, power grid operation data, and fault type at the time of the fault.

[0008] Obtain historical load data of the power system;

[0009] Based on the historical load data, a Bayesian network is used to predict the day-ahead load of the power system, resulting in day-ahead load prediction data.

[0010] Based on the multi-timescale dataset and the day-ahead load forecast data, power flow calculations are performed on the power system to obtain day-ahead power flow calculation results; the day-ahead power flow calculation results include active power, reactive power, and voltage distribution data of the power system;

[0011] Obtain the day-ahead fault type and day-ahead grid topology;

[0012] Based on the day-ahead fault type, the day-ahead grid topology, and the day-ahead power flow calculation results, the day-ahead security assessment index value of the power system is determined based on the security assessment model. The security assessment model is obtained by training XGBoost with a training sample set in advance. The training sample set includes multiple sets of feature values ​​and the security assessment index values ​​corresponding to each set of feature values. Each set of feature values ​​includes historical fault types, historical grid topology, and historical power flow calculation results.

[0013] Obtain the real-time load of the power system;

[0014] Based on the real-time load, power flow calculation is performed on the power system to obtain the real-time power flow calculation results;

[0015] Obtain the current fault type and the current power grid topology;

[0016] Based on the current fault type, the current power grid topology, and the real-time power flow calculation results, the real-time security assessment index value of the power system is determined according to the security assessment model.

[0017] To achieve the above objectives, the present invention also provides the following solution:

[0018] A data-model hybrid-driven power system security assessment system, comprising:

[0019] A multi-scale data acquisition unit is used to acquire multi-timescale datasets from the power grid dispatching platform. The multi-timescale datasets include planned data and daily incidental fault data. The planned data includes the power grid topology at the monthly time scale, the unit on / off status at the weekly time scale, and the power grid operation data at the daily time scale. The daily incidental fault data includes the power grid topology, power grid operation data, and fault type at the time of the fault.

[0020] The historical load acquisition unit is used to acquire historical load data of the power system.

[0021] The day-ahead load forecasting unit is connected to the historical load acquisition unit and is used to forecast the day-ahead load of the power system using a Bayesian network based on the historical load data to obtain day-ahead load forecasting data.

[0022] The day-ahead power flow calculation unit is connected to the multi-scale data acquisition unit and the day-ahead load forecasting unit, and is used to perform power flow calculation on the power system based on the multi-timescale dataset and the day-ahead load forecasting data to obtain the day-ahead power flow calculation results; the day-ahead power flow calculation results include the active power, reactive power and voltage distribution data of the power system.

[0023] The day-ahead topology acquisition unit is used to acquire day-ahead fault types and day-ahead grid topology.

[0024] The day-ahead security assessment unit, connected to the day-ahead topology acquisition unit and the day-ahead power flow calculation unit, is used to determine the day-ahead security assessment index value of the power system based on the day-ahead fault type, the day-ahead grid topology, and the day-ahead power flow calculation results, using a security assessment model. The security assessment model is obtained by pre-training XGBoost using a training sample set. The training sample set includes multiple sets of feature values ​​and the security assessment index values ​​corresponding to each set of feature values. Each set of feature values ​​includes historical fault types, historical grid topology, and historical power flow calculation results.

[0025] The real-time load acquisition unit is used to acquire the real-time load of the power system;

[0026] A real-time power flow calculation unit, connected to the real-time load acquisition unit, is used to perform power flow calculation on the power system based on the real-time load and obtain the real-time power flow calculation result;

[0027] The current topology acquisition unit is used to acquire the current fault type and the current power grid topology.

[0028] The real-time security assessment unit, connected to the current topology acquisition unit, the real-time power flow calculation unit, and the day-ahead security assessment unit, is used to determine the real-time security assessment index value of the power system based on the security assessment model, according to the current fault type, the current power grid topology, and the real-time power flow calculation results.

[0029] According to specific embodiments provided by the present invention, the following technical effects are disclosed: Multi-timescale datasets are collected from a power grid dispatching platform; based on historical load data, a Bayesian network is used to predict the day-ahead load of the power system, obtaining day-ahead load prediction data; power flow calculations are performed on the power system based on the multi-timescale datasets and the day-ahead load prediction data, obtaining day-ahead power flow calculation results; based on the day-ahead fault type, the day-ahead grid topology, and the day-ahead power flow calculation results, a pre-trained security assessment model is used to determine the day-ahead security assessment index value of the power system; based on the current fault type, the current grid topology, and the real-time power flow calculation results, a pre-trained security assessment model is used to determine the real-time security assessment index value of the power system. By using online and offline assessment methods, the computational burden is shifted to offline training, realizing day-ahead and real-time power system security assessments, thus improving the efficiency of security assessment. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a flowchart of the data-model hybrid-driven power system security assessment method of the present invention;

[0032] Figure 2 A flowchart illustrating the data-model hybrid-driven power system security assessment process;

[0033] Figure 3 This is a circuit diagram for handling line faults;

[0034] Figure 4 A flowchart for power system security assessment;

[0035] Figure 5 The training flowchart for the XGBoost model;

[0036] Figure 6 This is a schematic diagram of the modules of the data-model hybrid-driven power system security assessment system of the present invention.

[0037] Symbol explanation:

[0038] Multi-scale data acquisition unit-1, historical load acquisition unit-2, day-ahead load prediction unit-3, day-ahead power flow calculation unit-4, day-ahead topology acquisition unit-5, day-ahead security assessment unit-6, real-time load acquisition unit-7, real-time power flow calculation unit-8, current topology acquisition unit-9, real-time security assessment unit-10. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0041] like Figure 1 and Figure 2 As shown, the data-model hybrid-driven power system security assessment method provided by this invention includes:

[0042] S1: Collect multi-timescale datasets from the power grid dispatching platform. These datasets include planned data and daily incidental fault data. The planned data includes the power grid topology at the monthly timescale, the unit on / off status at the weekly timescale, and the power grid operation data at the daily timescale. The daily incidental fault data includes the power grid topology at the time of the fault, the power grid operation data, and the fault type.

[0043] Specifically, on a monthly timescale, grid topology data after line and generator maintenance is collected from the grid dispatch platform. On a weekly timescale, the on / off status of generators during the week is collected from the grid dispatch platform. On a daily timescale, information such as grid load, active and reactive power of generator nodes is collected from the grid dispatch platform. When a grid fault occurs intermittently, the current grid topology, generator and line on / off status, and active and reactive power of generators and loads are collected in real time from the grid dispatch platform.

[0044] When a power grid fault occurs, first identify the fault type and the current power grid topology, then calculate the probability of the fault occurring: pr i = (d+1) / D s In the formula, D s d represents the number of days from the date the power system began normal operation to the current date, and d represents the number of faults that have occurred so far.

[0045] S2: Obtain historical load data of the power system.

[0046] S3: Based on the historical load data, a Bayesian network is used to predict the day-ahead load of the power system to obtain the day-ahead load prediction data.

[0047] S4: Perform power flow calculations on the power system based on the multi-timescale dataset and the day-ahead load forecast data to obtain the day-ahead power flow calculation results. The day-ahead power flow calculation results include active power, reactive power, and voltage distribution data of the power system.

[0048] Specifically, a stochastic power flow model is first established based on the multi-timescale dataset and the day-ahead load forecast data; then, the semi-invariant method is used to solve the stochastic power flow model to determine the active power, reactive power, and voltage distribution data of the power system.

[0049] S5: Obtain the day-ahead fault type and day-ahead grid topology.

[0050] S6: Based on the day-ahead fault type, the day-ahead grid topology, and the day-ahead power flow calculation results, determine the day-ahead security assessment index value of the power system using a security assessment model. The security assessment model is obtained by pre-training XGBoost using a training sample set. The training sample set includes multiple sets of feature values ​​and the corresponding security assessment index values ​​for each set of feature values. Each set of feature values ​​includes historical fault types, historical grid topology, and historical power flow calculation results.

[0051] S7: Obtain the real-time load of the power system.

[0052] S8: Perform power flow calculation on the power system based on the real-time load to obtain the real-time power flow calculation result.

[0053] S9: Get the current fault type and the current power grid topology.

[0054] S10: Based on the current fault type, the current power grid topology, and the real-time power flow calculation results, determine the real-time security assessment index value of the power system based on the security assessment model.

[0055] Furthermore, in step S3 of this invention, a day-ahead load forecasting function considering uncertainty is implemented based on a Bayesian neural network. The Bayesian network consists of a statistical model, a neural network, prior probabilities, and likelihood probabilities. The steps for day-ahead forecasting using a Bayesian network include:

[0056] (1) Network model construction. Set the model training set I, I... x The characteristic value is the historical load, I. yThis represents the predicted value. W represents the neural network parameters, with each network weight w... i ∈W satisfies the mean μ i The variance is δ i The probability distribution is given by the statistical model, and each weight is independent, with the prior probability p(W) of the weight parameters given by the statistical model.

[0057] Calculation of posterior probability based on Bayes' theorem:

[0058]

[0059] Where p(W|I) represents the posterior probability of the network parameters, p(I) y |I x W) is based on weights W and sample I. x Down, I y The conditional probability can reflect the network estimation performance.

[0060] Variational inference makes the variational distribution q φ =q(W|θ) approximates p(W|I). First, calculate the KL divergence to obtain D. KL (q||p) is used to measure the distance between different probability distributions. Then, the ELBO function is calculated, and the ELBO is maximized through the backpropagation algorithm, thereby indirectly obtaining p(W|I).

[0061]

[0062]

[0063] Wherein, P(I) is the prior probability of the data, which is obtained based on historical load data through probabilistic statistical methods.

[0064] (2) Constructing the Bayesian neural network loss function L(I x,j ).

[0065]

[0066]

[0067]

[0068]

[0069] Where n represents the number of network weights, ε, γ, α, and β are all pre-defined hyperparameters, and θ i =(μ i ,δ i ), p(w) is the prediction result corresponding to the j-th sample. i ) represents the weight parameter wi The prior probability.

[0070] (3) Model Training and Evaluation. The Bayesian neural network model was trained based on historical data. Classic prediction evaluation metrics were selected to evaluate the model training results, including Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R² score, where y i Represents the actual value. This represents the predicted value, where m is the number of samples.

[0071]

[0072]

[0073]

[0074]

[0075] When performing power flow calculations in step S4, the fixed base data comes from a multi-timescale dataset, while the random data comes from day-ahead load forecast data.

[0076] In this embodiment, only the injected power of each node is considered to be independent. The stochastic power flow calculation steps are as follows:

[0077] (1) Input raw data, including data related to lines, generators, and nodes.

[0078] (2) Information on load and generator random quantities is determined using a normal distribution, and the expected value and variance need to be determined.

[0079] (3) Determine the information of the faulty line, including the faulty line number and the failure rate.

[0080] (4) Calculate the power flow using the Newton-Raphson method. Calculate the power flow information under normal operating conditions, namely, the node injection quantity W0 (including the active and reactive power injected into each node), the state variable X0 (i.e., the voltage magnitude and voltage phase angle related information of each node), and the branch power flow variable Z0 (including the branch active power P). ij and branch reactive power Q ij ).

[0081] (5) Based on the power flow results of the Newton-Raphson method, i.e., the state variable X0, calculate the Jacobian matrix J0 and the partial derivative matrix G0 of the branch power flow equation used in the last iteration of the Newton-Raphson method. Calculate the first sensitivity matrix S0 (the inverse of J0) and the second sensitivity matrix T0 based on the Jacobian matrix J0.

[0082] The branch power flow equations are:

[0083]

[0084] Among them, P ij Q is the active power of branch ij. ij V represents the reactive power of branch ij. i Let G be the voltage magnitude at node i. ij Let θ be the conductance of the line between node i and node j in the power network. ij B is the voltage phase difference between node i and node j in a power network. ij Let t be the inductance of the line between node i and node j in the power network. ij Let b be the turns ratio of the transformer branch. ij0 It is half of the capacity of branch ij.

[0085] The above equation can be expressed as Z = g(X), and the branch power flow Z = Z0 + ΔZ. Linearizing the branch power flow equation, we get:

[0086] ΔZ = G0ΔX;

[0087]

[0088] In the formula, ΔZ represents the random disturbance of the branch power flow Z.

[0089] The elements in G0 and the elements in the Jacobian matrix J0 have the following relationship:

[0090]

[0091]

[0092]

[0093]

[0094]

[0095]

[0096]

[0097]

[0098]

[0099]

[0100]

[0101]

[0102] In the formula, Vi Let V be the voltage magnitude at node i. j V represents the voltage amplitude at the node. k Let θ be the voltage magnitude at node k, k ≠ i, j. i Let θ be the voltage phase angle at node i. j Let θ be the voltage phase angle at node j. k H is the voltage phase angle at node k. ij N ij J ij L ij θ is an intermediate variable. ij Let be the voltage phase difference between node i and node j in the power system.

[0103] Linearizing the nodal power equations, we obtain ΔW≌J0ΔX, where ΔW is the random perturbation of the nodal injection quantity W, ΔX is the random perturbation of the state variable X, W=W0+ΔW, X=X0+ΔX.

[0104] Substituting ΔZ=G0·ΔX, we get ΔZ=G0S0ΔW=T0·ΔW, and then we can find the sensitivity matrix T0=G0·S0.

[0105] (6) Assuming the load distribution is a normal distribution, obtain the semi-invariants γ of the load for each order. wL (k) .

[0106] The output of new energy sources also follows a normal distribution. The semi-invariants γ of each order of new energy output can be obtained. wg (k) .

[0107]

[0108]

[0109] Where, μ L The mean load, μ g The rated power output for new energy sources, σ L Let σ be the standard deviation of the load distribution. g The standard deviation of the contribution to new energy sources, where k represents the order of the semi-invariant.

[0110] The handling of line faults is a discrete distribution, such as... Figure 3 As shown in the figure, Z g Let b0 be the impedance of the line, and b0 be half the capacity of branch ij. Add a virtual injection source to each node on both sides of the faulty branch. When the injected power is equal to the power flowing out from both ends of branch ij, the state of the system is the same as the state after branch ij is disconnected.

[0111] The ΔP of the virtual injection source can be calculated using the following formula. i ΔQ i ΔP j ΔQ j :

[0112]

[0113] Among them, T 4×4 It is a submatrix of the second sensitivity matrix T0. Elements corresponding to branches ij are extracted from the second sensitivity matrix T0 to form T. 4×4 A line fault is equivalent to ΔP. i ΔQ i ΔP j ΔQ j The binomial distribution of the two power sources, I 4×4 It is an identity matrix.

[0114] (7) A line fault is equivalent to a binomial distribution, and the central moments of each order of this binomial distribution can be calculated. q i For availability, C i ΔP for virtual injection source i ΔQ i wait.

[0115] Through the central moments α of each order v With each order of semi-invariant γ v Find the relationships between the semi-invariants of each order.

[0116] γ1=α1=μ;

[0117]

[0118]

[0119]

[0120]

[0121]

[0122] (8) Based on the additivity of semi-invariants, we obtain the semi-invariants of each order of ΔW.

[0123] According to the theorem that "the semi-invariants of the sum of independent random variables are equal to the sum of the semi-invariants of each variable", we can obtain the semi-invariants of ΔW. The semi-invariants of each order equal to the generator's injected power and the semi-invariants of load injection power sum.

[0124] When considering line faults, the semi-invariants that occur during a line fault must also be included.

[0125]

[0126] According to the theorem, "A k-th order semi-invariant of a random variable is equal to α times the k-th order semi-invariant of that variable." k "times", we can get:

[0127]

[0128]

[0129] In the formula, and The matrices formed by the k powers of the elements in the first sensitivity matrix S0 and the second sensitivity matrix T0 are respectively, γ x (k) and γ z (k) Let X and Z be the semi-invariants of each order, respectively.

[0130] (9) Calculate the central moments β of ΔX and ΔZ using the relationship between semi-invariants and central moments. v .

[0131] β1 = 0;

[0132] β2=γ2=σ 2 ;

[0133] β3 = γ3;

[0134]

[0135] β5 = γ5 + 10γ3γ2;

[0136]

[0137]

[0138] (10) Calculate the coefficient c based on the relationship between the central moment and the coefficients of the Gram-Charlier expansion series. v .

[0139] c0 = 1;

[0140] c1 = 0;

[0141] c2 = 0;

[0142]

[0143]

[0144]

[0145]

[0146]

[0147]

[0148] (11) According to the coefficient c of the Gram-Charlier expansion series v Determine the probability density functions and cumulative distribution functions of ΔX and ΔZ:

[0149] The probability density function and cumulative distribution function of the random variable x can be expressed as the Gram-Charlier expansion series as follows.

[0150] F(x)=Φ(x)+c1Φ'(x)+c2Φ”(x)+c3Φ”’(x)+…;

[0151]

[0152] Where F(x) is the probability distribution function of the random variable x, and f(x) is the probability density function of x. Φ(x) and Let m and σ be the distribution function and probability density function of a normal distribution with expected value m = 0 and standard deviation σ = 1, respectively.

[0153] Use Hermite polynomials to find the Gram-Charlier series expansion.

[0154] H0(x) = 1;

[0155] H1(x) = -x;

[0156] H2(x)=x 2 -1;

[0157] H3(x)=(x 3 -3x)×(-1);

[0158] H4(x)=x 4 -6x+3;

[0159] H5(x)=(x 5 -10x 3 +15x)×(-1);

[0160] H6(x)=x 6 -15x 4 +45x 2 -15.

[0161] From this, we can obtain the Gram-Charlier series expansion of the random variable distribution function:

[0162]

[0163]

[0164] In the formula, It is the standardized variable of the random variable x.

[0165] After standardizing the random variables ΔX and ΔZ, we obtain their probability density functions and cumulative distribution functions. Since the state variables X = X0 + ΔX and the branch power flow Z = Z0 + ΔZ, we shift the probability density function and cumulative distribution function of ΔX by X0 to obtain the probability density function and cumulative distribution function of X, and shift the probability density function and cumulative distribution function of ΔZ by Z0 units to obtain the probability density function and cumulative distribution function of Z.

[0166] Data-driven methods derive uncertainty variables entirely from historical data, abandoning the use of a fixed probability density function type. By calculating variables such as the mean and variance of historical data, the range of the probability density function is determined. Therefore, it can more accurately depict the distribution of uncertainty.

[0167] In this embodiment, a deviation degree measurement method is used to establish component-level severity indices, namely the power flow over-limit severity of branches and the voltage over-limit severity of nodes, to evaluate the consequences of power flow over-limit of branches and voltage over-limit of nodes.

[0168] The training process for the security assessment model includes:

[0169] S101: Obtain multiple sets of feature values. Each set of feature values ​​includes historical fault types, historical power grid topology, and historical power flow calculation results; the historical power grid topology includes multiple nodes and multiple branches.

[0170] S102: For any set of feature values, determine a set of anticipated faults based on the historical fault types and the historical power grid topology. The set of anticipated faults includes multiple anticipated faults.

[0171] This invention establishes component probabilistic models from two aspects: generators and transmission lines. The generator probabilistic model refers to the probabilistic model of a generator under normal operating conditions or fault conditions, and is divided into two types according to whether the influence of time periods is considered: a state probabilistic model within a single time period and a multi-time period state transition probabilistic model considering time sequence. Within a single time period, the generator probabilistic model generally has two types: a two-state model and a derating state model.

[0172] A two-state model refers to a generator that exists in only two states: rated operation and shutdown. In non-sequential Monte Carlo simulations, this model assumes the generator's operating state is S. i PF i To determine the failure probability, generate a random number R uniformly distributed in the interval [0, 1] for generator i. i :

[0173]

[0174] The derating state model is also known as the multi-state model. During multi-state Monte Carlo simulation sampling, random numbers R uniformly distributed in the interval [0, 1] are generated for generator i. i Perform a simulation:

[0175]

[0176] When considering multiple consecutive time periods, the generator's operating state changes with the scheduling situation in each period, which can be represented by a state transition model. Each state transition parameter represents the probability of the generator transitioning between operating, outage, and derated states, and is obtained by fitting based on historical data or short-term forecast data.

[0177] Based on historical data, the transmission line fault rate λ (times / year) and repair rate μ (times / year) are statistically analyzed. The forced outage rate FOR = λ / (λ+μ) is calculated. A random number R is drawn from a uniform distribution of (0,1). If R ≤ FOR, the line is considered faulty. The proportions of open-circuit and short-circuit faults are P0 and P, respectively. S Sampling R again further determines the fault type: when R <P S At that time, it was considered an open circuit fault (broken line).

[0178] Further considering the impact of weather factors on the failure probability of transmission lines, and based on the characteristics of series connection of lines, the failure rate λ of the entire line under severe and normal weather conditions is obtained. 1e Repair time r 1e and unavailability U 1e :

[0179] λ 1e =λ ad R+λ no (1-R);

[0180]

[0181] U 1e =U ad +U no -U ad U no ;

[0182] Where, λad The failure rate under severe weather conditions is represented by R, which is the percentage of the line length exposed to severe weather conditions, and λ. no The failure rate under normal climatic conditions, r ad For repair time under severe weather conditions, r no Repair time under normal climatic conditions, U ad Usage unavailability under severe weather conditions, U no This represents the unavailability rate under normal climatic conditions.

[0183] Based on the generator probability model and the transmission line probability model, a set of anticipated faults is formed, and the probability of power flow not exceeding the limit and voltage not exceeding the limit of each node under various fault conditions is calculated.

[0184] The anticipated fault set E is:

[0185]

[0186] in, For events where generator i is in a shutdown or derating state, E line For each fault event occurring in the transmission line, the probability of each fault is calculated using the formula in step S1.

[0187] In this embodiment, when the entire probabilistic safety analysis cycle is sufficiently short (e.g., less than 15 minutes), the set of anticipated faults is determined according to the N-1 rule:

[0188]

[0189]

[0190] Where prob{·} is the probability function of the event occurring; F pq (·) represents the probability distribution function of active and reactive power flow in the branch under fault E, F vθ (·) represents the probability distribution function of the node voltage magnitude and phase angle, σ pj Let σ be the standard deviation of the power flow in branch j. pj Let be the standard deviation of the voltage amplitude at node i.

[0191] S103: Based on the anticipated fault set, the historical power grid topology, and the historical power flow calculation results, determine the power flow exceedance severity of each branch and the voltage exceedance severity of each node.

[0192] Branch flow limit exceedance severity represents the percentage of the transmission capacity limit for each line. It is defined as the degree of flow limit exceedance in branch ij under all K possible fault scenarios. The flow limit exceedance severity of branch ij is determined using the following formula:

[0193]

[0194] in, For the power flow exceeding the limit severity of branch ij, Let E be the set of faults that could cause power flow out of limit in branch ij, and K be the total number of expected faults. Let P be the active power of branch ij when the anticipated fault k occurs. max P represents the maximum active power allowed to be transmitted in branch ij, obtained from the thermal stability limiting current. min Let || be the minimum active power value allowed to be transmitted in branch ij, and || be the absolute value.

[0195] The severity of a node voltage exceedance represents the percentage by which the voltage amplitude of each node deviates from its normal amplitude limit. It is defined as the degree of voltage exceedance at node i under all K possible fault scenarios. The severity of the voltage exceedance at node i is determined using the following formula:

[0196]

[0197] in, Let i be the severity of the voltage limit violation at node i. Let E be the set of faults that could cause the voltage amplitude at node i to exceed the limit, K be the total number of expected faults, and v be the set of potential fault events. H This represents the maximum allowable voltage amplitude for node i. Exceeding this maximum voltage amplitude constraint will cause voltage collapse. Let v be the voltage amplitude at node i when the anticipated fault k occurs. L This represents the minimum allowable voltage amplitude for node i. If the voltage amplitude is less than the minimum allowable voltage amplitude constraint, it will cause low-voltage instability in the system.

[0198] S104: Determine the overload severity index value of each branch based on the severity of the power flow exceeding the limit.

[0199] The power system overload risk index reflects the average level of overload risk for each branch and takes into account the impact of the maximum risk value. The branch overload severity index value is determined using the following formula:

[0200]

[0201] Among them, R OL The branch overload severity index value is given by α and β, where α + β = 1, M is the total number of branches in the power system, and Sev branch Sev is a matrix composed of the power flow exceedance severity of each branch. branch ||1 represents Sev branch Find the 1-norm, ||Sev branch || ∞ Indicates to Sevbranch Find the ∞ norm.

[0202] S105: Determine the node over-limit severity index value based on the voltage over-limit severity of each node.

[0203] The power system node exceedance severity index reflects the average level of voltage amplitude exceedance severity at each node, and takes into account the impact of component failure events with the greatest impact on the system. The node exceedance severity index value is determined using the following formula:

[0204]

[0205] Among them, R OV Here, α and β are the severity index values ​​for node exceeding limits, α + β = 1, and N is the total number of nodes in the power system. voltage The matrix consists of the voltage exceedance severity of each node, ||Sev voltage ||1 represents Sev voltage Find the 1-norm, ||Sev voltage || ∞ Indicates to Sev voltage Find the ∞ norm.

[0206] S106: Calculate the probability of each anticipated fault occurring, the probability of the power system being safe when each anticipated fault occurs, and the probability of the power system being safe when there is no fault.

[0207] The safety probability of a power system is calculated for all anticipated faults, determining the probability of branch power flow overload or node voltage exceedance. In the anticipated fault E... i After this event, the probability that the power system will not meet the safety constraints is:

[0208]

[0209] Assume event E i If the occurrence of each fault follows a Poisson distribution and the events are independent, then the probability of each fault occurring is:

[0210]

[0211] For each anticipated fault E k For each (k = 1, 2, ..., N), the probability Pr that the system satisfies the power flow safety constraints is calculated. s1 ,Pr s2 ,…,Pr sN The probability Pr is obtained when the system satisfies the power flow safety constraints under fault-free conditions. s0 .

[0212] S107: Determine the unsafe probability index value based on the probability of each anticipated fault occurring, the safety probability of the power system when each anticipated fault occurs, and the safety probability of the power system when there is no fault.

[0213] Specifically, the unsafety probability index value is calculated using the following formula:

[0214]

[0215] Among them, Pr ins Pr is the probability index value of insecurity. s0 Let K be the safety probability of the power system under fault-free conditions, and K be the total number of anticipated faults. For the anticipated fault E k The probability of occurrence, pr sk For the anticipated fault E k The probability of the power system being safe when it occurs.

[0216] S108: Determine the safety assessment index value of the corresponding group of feature values ​​based on the branch overload severity index value, the node over-limit severity index value, and the unsafe probability index value.

[0217] Specifically, the safety assessment index values ​​are determined using the following formula:

[0218] S0=w1·Pr ins +w2·R OL +w3·R OV ;

[0219] Where S0 is the safety assessment index value, w1 is the weight of the unsafe probability index, w2 is the weight of the branch overload severity index, w3 is the weight of the node limit violation severity index, and Pr ins R is the probability index of insecurity. OL R is the severity index value for branch overload. OV This represents the severity index value for node exceeding limits. Weights can be determined through methods such as subjective and objective weighting.

[0220] S109: Train XGBoost based on the feature values ​​of each group and the security assessment index values ​​of each group of feature values ​​to obtain a security assessment model.

[0221] The safety assessment model constructed in this invention includes a Bayesian neural network for day-ahead stochastic prediction and an XGboost network for safety assessment. Day-ahead, the probability density function of the Bayesian neural network for stochastic prediction of wind and solar load, combined with other boundary conditions provided by the scheduling platform, forms the input to the XGboost network, enabling the assessment of system safety for the following day. On one hand, changes in boundary conditions such as network architecture and installed capacity can dynamically update both networks based on the scheduling platform input and actual power flow statistics, achieving adaptive updates to the parameters of the two neural networks. On the other hand, in the event of a fault, the fault type and frequency can be used to update the weight parameters of the XGboost network.

[0222] To improve the accuracy and speed of safety prediction, before training XGBoost, the feature values ​​in the training sample set, namely the historical power flow calculation results (branch active power flow distribution, mean and variance of voltage amplitude, etc.), are first subjected to z-score standardization.

[0223]

[0224] Where x′ represents the standardized data, x represents the unstandardized data, and x' represents the data before standardization. μ x is the mean of the data before standardization. σ This represents the standard deviation of the data before standardization.

[0225] The above formula is used to standardize all collected data according to their categories. Variance and mean are specific to a particular data category. For example, the standardization process for the active power flow variance in the sample is as follows: calculate the variance and mean of the active power flow variance category, and then standardize them.

[0226] This invention provides a power system security assessment that includes both day-ahead and real-time assessments. It completes online inference based on offline training, thereby improving the efficiency of the security assessment.

[0227] In the day-ahead security assessment, during offline training, the model inputs fault types and grid topology data from multiple time scales, as well as historical power flow, voltage distribution mean, and voltage distribution variance data from stochastic power flow calculations. During online inference, the security assessment model inputs fault types, grid topology, newly calculated power flow, voltage distribution mean, and voltage distribution variance based on day-ahead predicted loads, and outputs the day-ahead security assessment results.

[0228] In the real-time security assessment function, during offline training, the model inputs fault types and grid topology data across multiple time scales, as well as historical power flow, voltage distribution mean, and voltage distribution variance data from stochastic power flow calculations. During online inference, the security assessment model inputs fault types, grid topology, power flow calculated based on real-time loads, and voltage distribution mean and variance, and outputs real-time security assessment results.

[0229] Specifically, when building a security assessment model based on XGBoost, the system security assessment scenario is first designed. Based on the data required for day-ahead and real-time security assessments, a security assessment training sample set is constructed. Based on the corresponding fault set and data collection points, a sample set X = [X1,...,X2] representing the changes in stochastic power flow input and output under normal operating conditions and fault conditions is constructed. i ,...,X N ]∈R N×M Where N represents the number of samples, i.e., N security assessment scenarios, and M represents the dimension of the feature vector of each sample. Sample X i The input feature vector contains the mean {f} of the active power flow distribution of the branches. h |h∈H} and variance {σ} h |h∈H}, and the magnitude of the node voltage {v a |a∈A} and variance {ε} a |a∈A}, its dimension M is the sum of the number of branches H and the number of nodes A. Each sample X i The corresponding system insecurity probability y i This constitutes the output Y of the security assessment model.

[0230] Based on the training sample set described above, a security assessment model is constructed. The system security assessment framework is as follows: Figure 4 As shown, for a given dataset D = {(X i ,y i )}(|D|=N,X i ∈R M ,y i The ensemble model of a tree (∈R) is represented by the following equation:

[0231]

[0232] In the formula, The output of the model; T k (X i ) represents the prediction result of the k-th decision tree; This represents the mapping from the input samples of a node to the leaf nodes; each tree T k This corresponds to an independent tree structure q and the weights of the leaves. X iis the feature vector of the i-th sample; q represents the index of the leaf corresponding to the sample in the structure of each tree; Λ is the number of leaves in the tree; It is a collection space of trees.

[0233] The objective function for XGBoost training is:

[0234]

[0235] In the formula, These are model parameters; N represents the quantization error of the model on the training samples, and N is the number of training samples. This is the model complexity regularization term, used to reduce the risk of overfitting. K is the number of base learners in the model. C is a constant. Γ i (X i ) represents the mapping from a given input sample value to a leaf node.

[0236] The model complexity of a single base learner in the XGBoost algorithm is:

[0237]

[0238] In the formula, M is the base learner T k The number of leaf nodes, λ represents the L2 regularization coefficient, and γ represents the difficulty of node splitting. The L2 norm represents the weight of the leaf node.

[0239] In general, the XGBoost model is trained using an incremental training method. This means that each time, a new function (i.e., a new tree) is added to the model while retaining the original model. By integrating a series of base learners with relatively weak learning capabilities, better performance can be obtained. A smaller value indicates a better tree structure. The incremental function added in each round aims to minimize the objective function, training the k-th base learner T. k The learning objective function at that time is:

[0240]

[0241] In the formula, For T k The parameters; Ω(T) k ) for T k Model complexity; This represents the model residual from the previous iteration; For T k The output; the learning rate ε ranges from (0,1).

[0242] Based on the training sample set, CART base learners are continuously trained to fit the residuals of previous models and integrated into the XGBoost model until a preset number of base learners are trained or the model residuals are less than a set threshold.

[0243]

[0244]

[0245] In the formula, It is the model prediction value of the i-th sample in round t. Retain the model predictions from round t-1. Then, add a new function T. t (X i The training flowchart is as follows: Figure 5 As shown.

[0246] This method combines online and offline model-data-driven assessment to achieve security assessment applicable to new power systems. First, it constructs a multi-timescale security assessment dataset for new power systems, covering monthly, weekly, and daily power grid architecture and operational data, by collecting data online from the power grid dispatching platform. Then, it establishes a stochastic power flow model for the new energy power system and solves the model using a semi-invariant method. Finally, it dynamically interacts with the power grid dispatching platform to update data in real time and train the data-driven model, transforming the complex calculation process into matrix operations and shifting the computational burden to offline training. This enables day-ahead and real-time multi-scale security assessment of new energy power systems, improving the computational efficiency of new energy power system security assessment. This invention includes a multi-timescale data module, a stochastic power flow calculation module, and a data-driven multi-timescale new energy power system security assessment module.

[0247] This invention can directly export data such as annual inter-provincial transmission line power flow, monthly maintenance plans, weekly unit combination plans, and day-ahead market clearing results from the dispatching platform. On one hand, this data constitutes the system operating boundary conditions in stochastic power flow calculations, providing the probability density functions for uncertain sources (new energy generation, user power load, and line outages) in stochastic power flow calculations. On the other hand, combined with the output of the stochastic power flow module, this data can also be used as input to the XGboost network for training the network and evaluating system security on typical days at different time scales. This simplifies the calculation process of traditional power system static security assessment, is suitable for new power system security assessment scenarios with new energy integration, and offers fast and robust security assessment.

[0248] like Figure 6As shown, the data-model hybrid driven power system security assessment system of the present invention includes: a multi-scale data acquisition unit 1, a historical load acquisition unit 2, a day-ahead load prediction unit 3, a day-ahead power flow calculation unit 4, a day-ahead topology acquisition unit 5, a day-ahead security assessment unit 6, a real-time load acquisition unit 7, a real-time power flow calculation unit 8, a current topology acquisition unit 9, and a real-time security assessment unit 10.

[0249] The multi-scale data acquisition unit 1 is used to collect multi-timescale datasets from the power grid dispatching platform. The multi-timescale datasets include planned data and daily incidental fault data; the planned data includes the power grid topology at the monthly timescale, the unit on / off status at the weekly timescale, and the power grid operation data at the daily timescale; the daily incidental fault data includes the power grid topology, power grid operation data, and fault type at the time of the fault.

[0250] Historical load acquisition unit 2 is used to acquire historical load data of the power system.

[0251] The day-ahead load forecasting unit 3 is connected to the historical load acquisition unit 2. The day-ahead load forecasting unit 3 is used to forecast the day-ahead load of the power system using a Bayesian network based on the historical load data, and obtain day-ahead load forecasting data.

[0252] The day-ahead power flow calculation unit 4 is connected to the multi-scale data acquisition unit 1 and the day-ahead load forecasting unit 3. The day-ahead power flow calculation unit 4 is used to perform power flow calculations on the power system based on the multi-timescale dataset and the day-ahead load forecasting data to obtain the day-ahead power flow calculation results. The day-ahead power flow calculation results include the active power, reactive power, and voltage distribution data of the power system.

[0253] The day-ahead topology acquisition unit 5 is used to acquire the day-ahead fault type and the day-ahead grid topology.

[0254] The day-ahead security assessment unit 6 is connected to the day-ahead topology acquisition unit 5 and the day-ahead power flow calculation unit 4. The day-ahead security assessment unit 6 is used to determine the day-ahead security assessment index value of the power system based on the day-ahead fault type, the day-ahead grid topology and the day-ahead power flow calculation result, and a security assessment model. The security assessment model is obtained by training XGBoost in advance using a training sample set. The training sample set includes multiple sets of feature values ​​and the security assessment index values ​​corresponding to each set of feature values. Each set of feature values ​​includes historical fault types, historical grid topology and historical power flow calculation results.

[0255] The real-time load acquisition unit 7 is used to acquire the real-time load of the power system.

[0256] The real-time power flow calculation unit 8 is connected to the real-time load acquisition unit 7. The real-time power flow calculation unit 8 is used to perform power flow calculation on the power system based on the real-time load and obtain the real-time power flow calculation result.

[0257] Current topology acquisition unit 9 is used to acquire the current fault type and the current power grid topology.

[0258] The real-time security assessment unit 10 is connected to the current topology acquisition unit 9, the real-time power flow calculation unit 8, and the day-ahead security assessment unit 6. The real-time security assessment unit 10 is used to determine the real-time security assessment index value of the power system based on the security assessment model according to the current fault type, the current power grid topology, and the real-time power flow calculation results.

[0259] Compared to existing technologies, the data-model hybrid-driven power system security assessment system provided by this invention has the same beneficial effects as the data-model hybrid-driven power system security assessment method described above, and will not be repeated here.

[0260] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0261] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A data-model hybrid-driven method for power system security assessment, characterized in that, The data-model hybrid-driven power system security assessment method includes: Collect multi-timescale datasets from the power grid dispatching platform; the multi-timescale datasets include planned data and daily incidental fault data; the planned data includes the power grid topology at the monthly timescale, the unit on / off status at the weekly timescale, and the power grid operation data at the daily timescale; the daily incidental fault data includes the power grid topology, power grid operation data, and fault type at the time of the fault. Obtain historical load data of the power system; Based on the historical load data, a Bayesian network is used to predict the day-ahead load of the power system, resulting in day-ahead load prediction data. Based on the multi-timescale dataset and the day-ahead load forecast data, power flow calculations are performed on the power system to obtain day-ahead power flow calculation results; the day-ahead power flow calculation results include active power, reactive power, and voltage distribution data of the power system; Obtain the day-ahead fault type and day-ahead grid topology; Based on the day-ahead fault type, the day-ahead grid topology, and the day-ahead power flow calculation results, the day-ahead security assessment index value of the power system is determined based on the security assessment model. The security assessment model is obtained by training XGBoost with a training sample set in advance. The training sample set includes multiple sets of feature values ​​and the security assessment index values ​​corresponding to each set of feature values. Each set of feature values ​​includes historical fault types, historical grid topology, and historical power flow calculation results. Obtain the real-time load of the power system; Based on the real-time load, power flow calculation is performed on the power system to obtain the real-time power flow calculation results; Obtain the current fault type and the current power grid topology; Based on the current fault type, the current power grid topology, and the real-time power flow calculation results, the real-time security assessment index value of the power system is determined according to the security assessment model. The training process of the security assessment model includes: Multiple sets of feature values ​​are obtained; each set of feature values ​​includes historical fault types, historical power grid topology, and historical power flow calculation results; the historical power grid topology includes multiple nodes and multiple branches; For any set of feature values, a set of anticipated faults is determined based on the historical fault types and the historical power grid topology; the set of anticipated faults includes multiple anticipated faults. Based on the anticipated fault set, the historical power grid topology, and the historical power flow calculation results, determine the power flow exceedance severity of each branch and the voltage exceedance severity of each node. Determine the overload severity index value of each branch based on the severity of the power flow exceeding the limit in each branch; Determine the node limit exceedance severity index value based on the voltage exceedance severity of each node; Calculate the probability of each anticipated fault occurring, the probability of the power system being safe when each anticipated fault occurs, and the probability of the power system being safe when there is no fault. The unsafe probability index value is determined based on the probability of each anticipated fault occurring, the safety probability of the power system when each anticipated fault occurs, and the safety probability of the power system when there is no fault. ;in, Pr ins Pr is the probability index value of insecurity. s0 The probability of a fault-free power system. K The total number of anticipated failures, anticipating the malfunction E k The probability of occurrence anticipating the malfunction E k The probability of a power system's safety when it occurs; Based on the branch overload severity index value, the node over-limit severity index value, and the unsafety probability index value, determine the safety assessment index value for the corresponding set of feature values: ;in, S 0 represents the safety assessment index value. w 1 represents the weight of the insecurity probability index. w 2 represents the weight of the branch overload severity index. w 3 represents the weight of the node exceeding the limit severity index. Pr ins This is the value of the probability of insecurity index. R OL The value represents the severity index of branch circuit overload. R OV The value represents the severity index of node exceeding limits; Based on the feature values ​​of each group and the security assessment index values ​​of each group of feature values, XGBoost is trained to obtain a security assessment model.

2. The data-model hybrid-driven power system security assessment method according to claim 1, characterized in that, The step of performing power flow calculations on the power system based on the multi-timescale dataset and the day-ahead load forecast data to obtain day-ahead power flow calculation results specifically includes: Based on the multi-timescale dataset and the day-ahead load forecast data, a stochastic power flow model is established; The stochastic power flow model is solved using the semi-invariant method to determine the active power, reactive power, and voltage distribution data of the power system.

3. The data-model hybrid-driven power system security assessment method according to claim 1, characterized in that, The following formula is used to determine the branch. ij The severity of the trend exceeding limits: ； in, branch road ij The more severe the trend, the more serious it becomes. To cause branch roads ij A collection of faults that exceed the limits of current flow. E For the anticipated fault set, K The total number of anticipated failures, branch road ij Anticipating the malfunction k Active power at the time of occurrence P max branch road ij The maximum active power value allowed to be transmitted. P min branch road ij The minimum active power value allowed to be transmitted, where || is the absolute value. P ij branch road ij The active power.

4. The data-model hybrid-driven power system security assessment method according to claim 1, characterized in that, The following formula is used to determine the nodes. i Voltage over-limit severity: ； in, For nodes i The severity of voltage exceeding the limit, To cause nodes i A set of faults involving voltage amplitude exceeding limits. E For the set of anticipated failure events, K The total number of anticipated failures, For nodes i Maximum permissible voltage amplitude For nodes i Anticipating the malfunction k Voltage amplitude at the time of occurrence For nodes i Minimum permissible voltage amplitude.

5. The data-model hybrid-driven power system security assessment method according to claim 1, characterized in that, The following formula is used to determine the severity index value of branch overload: ； in, R OL The value represents the severity index of branch circuit overload. α and β These are the weighting coefficients. α + β =1, M The total number of branches in the power system. Sev branch A matrix composed of the power flow exceedance severity of each branch. Indicates to Sev branch Find the 1-norm. Indicates to Sev branch Find the ∞ norm.

6. The data-model hybrid-driven power system security assessment method according to claim 1, characterized in that, The severity index value for node exceeding limits is determined using the following formula: ； in, R OV The severity index value for node exceeding limits. α and β These are the weighting coefficients. α + β =1, N This represents the total number of nodes in the power system. Sev voltage This is a matrix composed of the severity of voltage exceedance at each node. Indicates to Sev voltage Find the 1-norm. Indicates to Sev voltage Find the ∞ norm.

7. A data-model hybrid-driven power system security assessment system, applied to the data-model hybrid-driven power system security assessment method according to any one of claims 1-6, characterized in that, The data-model hybrid-driven power system security assessment system includes: A multi-scale data acquisition unit is used to acquire multi-timescale datasets from the power grid dispatching platform. The multi-timescale datasets include planned data and daily incidental fault data. The planned data includes the power grid topology at the monthly time scale, the unit on / off status at the weekly time scale, and the power grid operation data at the daily time scale. The daily incidental fault data includes the power grid topology, power grid operation data, and fault type at the time of the fault. The historical load acquisition unit is used to acquire historical load data of the power system. The day-ahead load forecasting unit is connected to the historical load acquisition unit and is used to forecast the day-ahead load of the power system using a Bayesian network based on the historical load data to obtain day-ahead load forecasting data. The day-ahead power flow calculation unit is connected to the multi-scale data acquisition unit and the day-ahead load forecasting unit, and is used to perform power flow calculation on the power system based on the multi-timescale dataset and the day-ahead load forecasting data to obtain the day-ahead power flow calculation results; the day-ahead power flow calculation results include the active power, reactive power and voltage distribution data of the power system. The day-ahead topology acquisition unit is used to acquire day-ahead fault types and day-ahead grid topology. The day-ahead security assessment unit, connected to the day-ahead topology acquisition unit and the day-ahead power flow calculation unit, is used to determine the day-ahead security assessment index value of the power system based on the day-ahead fault type, the day-ahead grid topology, and the day-ahead power flow calculation results, using a security assessment model. The security assessment model is obtained by pre-training XGBoost using a training sample set. The training sample set includes multiple sets of feature values ​​and the security assessment index values ​​corresponding to each set of feature values. Each set of feature values ​​includes historical fault types, historical grid topology, and historical power flow calculation results. The real-time load acquisition unit is used to acquire the real-time load of the power system; A real-time power flow calculation unit, connected to the real-time load acquisition unit, is used to perform power flow calculation on the power system based on the real-time load and obtain the real-time power flow calculation result; The current topology acquisition unit is used to acquire the current fault type and the current power grid topology. The real-time security assessment unit, connected to the current topology acquisition unit, the real-time power flow calculation unit, and the day-ahead security assessment unit, is used to determine the real-time security assessment index value of the power system based on the security assessment model, according to the current fault type, the current power grid topology, and the real-time power flow calculation results.

Images