Power system risk alarm method and device, storage medium and product

Abstract

The invention discloses an electric power system risk alarm method and device, a storage medium and a product. The method comprises the steps that data samples are constructed at all moments; inputting the active power of the generator at the current moment, the reactive power of the generator at the current moment, the active power of the load at the current moment, the reactive power of the load at the current moment, the topological state at the current moment and the meteorological data at the next moment into a preset topological prediction model, and calculating the topological state at the next moment, the method comprises the steps of inputting historical data samples at multiple moments, the topological state at the next moment and meteorological data at the next moment into a preset operation state prediction model for operation to obtain a bus voltage at the next moment and a load rate at the next moment, and for the next moment, calculating a risk value of a power system according to the topological state, the bus voltage and the load rate. And performing alarm operation on the power system according to the risk value. And the risk prediction accuracy of the power system is improved.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of power system reliability analysis, and in particular to a power system risk alarm method, device, storage medium and product. Background Art

[0002] As the scale of power systems continues to expand and become more complex, power systems are affected by a variety of factors during operation, such as load fluctuations, equipment failures, weather changes, etc. These factors may cause abnormal operation or even failure of the power system, affecting the reliability and economy of power supply. Therefore, accurately predicting the state of the power system is one of the factors to improve the safety of the power system.

[0003] Traditional methods for judging the operating status of power systems mainly rely on rule-based system monitoring and manual analysis. These methods have low adaptability and analytical capabilities for complex factors such as load fluctuations, equipment failures, and weather changes. Therefore, the accuracy of predicting power system risks is low. Summary of the invention

[0004] The present invention provides a power system risk warning method, equipment, storage medium and product, which are used to improve the accuracy of power system risk prediction.

[0005] In a first aspect, an embodiment of the present invention provides a power system risk warning method, comprising:

[0006] Constructing data samples at various times; each of the data samples includes bus voltage, generator active power, generator reactive power, load active power, load reactive power, topological state of the power system, load rate of transmission lines and transformers, and meteorological data;

[0007] Input the generator active power at the current moment, the generator reactive power at the current moment, the load active power at the current moment, the load reactive power at the current moment, the topological state at the current moment, and the meteorological data at the next moment into a preset topological prediction model to calculate the topological state at the next moment;

[0008] Input the data samples at multiple historical moments, the topological state at the next moment, and the meteorological data at the next moment into a preset operation state prediction model for calculation, so as to obtain the bus voltage at the next moment and the load rate at the next moment;

[0009] For the next moment, calculating the risk value of the power system according to the topological state, the bus voltage and the load rate;

[0010] An alarm operation is performed on the power system according to the risk value.

[0011] In a second aspect, an embodiment of the present invention provides a power system risk warning device, comprising:

[0012] A data sample construction module is used to construct data samples at various times; each of the data samples includes bus voltage, generator active power, generator reactive power, load active power, load reactive power, topological state of the power system, load rate of transmission lines and transformers, and meteorological data;

[0013] A topology prediction module, used for inputting the generator active power at the current moment, the generator reactive power at the current moment, the load active power at the current moment, the load reactive power at the current moment, the topology state at the current moment and the meteorological data at the next moment into a preset topology prediction model, and calculating the topology state at the next moment;

[0014] A motion state prediction module, used for inputting the data samples at multiple moments in history, the topological state at the next moment and the meteorological data at the next moment into a preset operation state prediction model for calculation, so as to obtain the bus voltage at the next moment and the load rate at the next moment;

[0015] A risk value calculation module, used for calculating the risk value of the power system according to the topological state, the bus voltage and the load rate at the next moment;

[0016] The power system alarm module is used to perform an alarm operation on the power system according to the risk value.

[0017] In a third aspect, an embodiment of the present invention further provides a computer device, the computer device comprising:

[0018] one or more processors;

[0019] A storage device for storing one or more programs;

[0020] When the one or more programs are executed by the one or more processors, the one or more processors implement the power system risk warning method provided in the first aspect of the present invention.

[0021] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the power system risk warning method provided in the first aspect of the present invention.

[0022] In a fifth aspect, an embodiment of the present invention further provides a computer program product, the computer program product comprising a computer program, and when the computer program is executed by a processor, the power system risk warning method provided in the first aspect of the present invention is implemented.

[0023] In the embodiment of the present invention, data samples are constructed at each moment; each data sample includes bus voltage, generator active power, generator reactive power, load active power, load reactive power, topological state of the power system, load rate and meteorological data of transmission lines and transformers, and the current generator active power, the current generator reactive power, the current load active power, the current load reactive power, the current topological state and the next meteorological data are input into a preset topological prediction model to calculate the topological state at the next moment, and the data samples of multiple moments in history, the topological state at the next moment and the meteorological data at the next moment are input into a preset operation state prediction model to calculate, and the bus voltage at the next moment and the load rate at the next moment are obtained. For the next moment, the risk value of the power system is calculated according to the topological state, the bus voltage and the load rate, and the power system is alarmed according to the risk value. Through the prediction of the topological state, the operation state and the risk value, potential faults can be discovered in advance and early warning can be given, thereby reducing the risk of power outages and accidents. At the same time, the combination of meteorological data enhances the ability of the power system to respond to emergencies. It comprehensively considers data from multiple aspects and uses topology prediction models and operating status prediction models to predict the risks of the power system in turn, thereby strengthening the correlation between multiple data and improving the accuracy of power system risk prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A flow chart of a power system risk warning method provided in Embodiment 1 of the present invention;

[0025] Figure 2 A schematic diagram of the structure of a network of a topology prediction model provided in the first embodiment of the present invention;

[0026] Figure 3 A schematic diagram of the structure of an operation status prediction model provided in the first embodiment of the present invention;

[0027] Figure 4 A schematic diagram of a structure of a power system risk alarm provided in the first embodiment of the present invention;

[0028] Figure 5 A structural block diagram of a power system risk warning device provided in Embodiment 2 of the present invention;

[0029] Figure 6 A schematic diagram of the structure of a computer device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0031] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can cover sequential implementations other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0032] Embodiment 1

[0033] See also Figure 1 , shows a flow chart of a power system risk warning method provided by the first embodiment of the present invention, the method can be executed by a power system risk warning device, the power system risk warning device can be implemented in the form of hardware and / or software, and the power system risk warning device can be configured in a computer device. Figure 1 As shown, the method includes:

[0034] Step 101: Construct data samples at each moment.

[0035] In this embodiment, data samples of the power system at various moments are collected and accumulated, and each moment is the current moment, the future moment, and the historical moment. These data samples include bus voltage, generator active power, generator reactive power, load active power, load reactive power, topological state of the power system, load rate of transmission lines and transformers, and meteorological data. These data samples provide a basis for subsequent power system prediction and risk assessment. Through data samples at multiple moments, the dynamic changes and potential risks of the power system can be better reflected to ensure that the state of the power system at each moment can be fully recorded and accurately analyzed, providing multi-dimensional reference information for risk assessment, thereby effectively predicting the future state of the power system and discovering potential operating risks in advance.

[0036] Exemplarily, data samples are collected from monitoring power systems and data acquisition databases. Meteorological data in the data samples are from a meteorological information center. Meteorological data include weather forecasts, typhoon forecasts, and real-time typhoon information. The time resolutions for normal weather and typhoons are 48h hourly forecasts and 24h hourly forecasts, respectively. The spatial resolution of the weather forecast is 5km. In particular, the time resolution of wind-related forecast data is 5 minutes to 3 hours; a large number of data samples are required for risk assessment of the power system, so more virtual data samples can be generated, and Gaussian distribution is chosen to generate a large number of virtual simulation data samples.

[0037] Step 102: input the current generator active power, the current generator reactive power, the current load active power, the current load reactive power, the current topological state and the next moment's meteorological data into a preset topological prediction model to calculate the next moment's topological state.

[0038] The topological state refers to the connection relationship and operating status between various devices (such as generators, transformers, lines, etc.) in the power system, including the switch status and fault conditions of the devices. The topological state determines the path of power flow, the stability of the system, and the response to faults. Changes in the topological state usually mean adjustments to the structure of the power system, which may be caused by equipment failure, scheduling or other factors, affecting the safety and reliability of the power system. The topological state of the present invention includes the topological state of the power system and the probability of the topological state.

[0039] In this embodiment, the key variables such as the current generator active power, the current generator reactive power, the current load active power, the current load reactive power and the topological state of the power system are input into the preset model topological prediction model, and the meteorological data of the next moment are combined for calculation, so that the potential topological state can be identified in advance, providing basic data for further risk assessment. The foresight and responsiveness of the power system are improved, thereby optimizing the operational stability and safety of the power system.

[0040] The topology prediction model uses a random neural network. The topology prediction model can be a Bayesian neural network. The Bayesian neural network combines the advantages of Bayesian reasoning and neural networks, and can effectively learn and predict under conditions of large uncertainty. The Bayesian neural network introduces probability distribution to represent the uncertainty of weights, making the Bayesian neural network more robust when facing noisy data or limited samples, and suitable for tasks that require high reliability and risk assessment.

[0041] An example of an improvement from artificial neural networks to Bayesian neural networks. In essence, artificial neural networks can describe any complex mapping y=f(x) between input and output. Traditional artificial neural networks consist of an input layer l0, a series of hidden layers l i , i = 0, 1, ..., n-1, the output layer is l n .

[0042] l0=x;

[0043] l i =s i (w i *l i-1 +b i );

[0044] y=l n ;

[0045] In the formula, w i represents the weight, b i represents the bias, s i represents the activation function, x represents the input, l0 represents the input layer, l i represents the hidden layer, y represents the output, l n Represents the output layer.

[0046] Since the parameters of the artificial neural network are fixed (i.e., the weight W i and bias b i ), that is, each parameter has a fixed value, so given the same input, its output will not change. However, weather-related power system failures are probabilistic random events, and artificial neural networks lack the ability to simulate such problems. Therefore, random neural networks are introduced into the modeling of probabilistic random problems. Random neural networks (such as Bayesian neural networks) are artificial neural networks that introduce random components into the network. Random neural networks can be implemented by giving the network random parameters to simulate multiple possible models. This process can be summarized as follows:

[0047] θ~p(θ); y=f θ (x)+∈

[0048] In the above formula, θ~p(θ) means that the network parameter θ obeys the probability distribution p(θ), ∈ means random noise, i.e., random disturbance, y means the output of the random neural network, and f θ (x) represents the output data processed by the random neural network. A Bayesian neural network can then be defined as any random neural network trained using Bayesian inference.

[0049] In one embodiment of the present invention, step 102 may include the following steps:

[0050] Step 1021: Set a posterior distribution for the preset topology prediction model.

[0051] In this embodiment, the purpose of setting the posterior distribution is to model the uncertainty of the topological state of the power system through the Bayesian reasoning method, and obtain the probability distribution of the future topological state of the power system. The posterior distribution provides a framework that can calculate the probability distribution of the topological state of the power system at the next moment based on the known sample data at the current moment, combined with the prior knowledge of the power system (such as historical topological state, meteorological data, etc.). This is not just to obtain a predicted value, but to understand the possibility and risk of topological changes from a probabilistic perspective. Therefore, setting the posterior distribution helps the model take into account a variety of possible topological states of the power system when predicting, and introduces uncertainty and deviation in the reasoning process, further improving the reliability and robustness of the prediction. The topological model includes an input layer, a hidden layer, and an output layer.

[0052] In one embodiment of the present invention, step 1021 includes the following steps:

[0053] Step 10211: construct a posterior distribution function for the preset topology prediction model according to Bayes' theorem.

[0054] In this embodiment, the core purpose of constructing the posterior distribution function is to model the uncertainty of the topological state of the power system through Bayes' theorem, so as to enable more reliable prediction and risk assessment in an uncertain environment. Bayes' theorem provides a mathematical framework that combines prior knowledge (i.e., previous power system topological states and assumptions) with new observation data (such as the current power system topological state). Specifically, through the combination of prior distribution, likelihood function and marginal likelihood, Bayes' theorem can be updated to obtain a more accurate posterior distribution. By constructing a posterior distribution function for the topological prediction model, the probability of each possible topological state of the power system at the next moment can be quantified, which is usually more important for further risk assessment, warning and scheduling decisions.

[0055] Exemplarily, the posterior distribution function is expressed as:

[0056]

[0057] In the formula, p(θ|D) represents the posterior distribution function, θ represents the network parameters of the topology prediction model, D represents the data sample, p(θ) represents the prior distribution, p(D|θ) represents the likelihood function, and p(D) represents the marginal likelihood. represents the i-th data sample, represents the i-th topological state, Indicated in The probability of occurrence of the i-th topological state under the conditions of and θ, N represents the number of data samples. θ includes a local random parameter β and a global deterministic parameter α.

[0058] The Bayesian posterior distribution for complex models such as artificial neural networks is a high-dimensional, highly non-convex probability distribution. This complexity makes computing and sampling Bayesian posteriors a thorny problem using standard methods, especially because computing evidence It is difficult. To solve this problem, distributional reasoning is introduced.

[0059] Step 10212: Perform distribution inference based on the posterior distribution function to obtain an approximate posterior of the posterior distribution as the topological model.

[0060] In this embodiment, in the Bayesian reasoning process, the purpose of distribution reasoning is to approximate the specific form of the posterior distribution through certain mathematical means, so as to further perform reasoning and prediction. Since the precise calculation of the posterior distribution is usually very complicated, especially for high-dimensional systems (such as power system topology prediction models), it may not be realistic to directly calculate the posterior distribution. Therefore, it is necessary to simplify the problem through approximate reasoning methods to obtain a reasonable approximate posterior distribution.

[0061] Distributed inference is an optimization algorithm that approximates a complex posterior distribution by introducing a simple variational distribution, and then optimizes the approximate result by minimizing the difference (i.e., divergence) between the variational distribution and the true posterior distribution. In this way, an approximate posterior of the posterior distribution with sufficient accuracy can be obtained while maintaining computational feasibility, thereby supporting topological state prediction and future risk assessment.

[0062] In one embodiment of the present invention, step 10212 may include the following steps:

[0063] Step 2121: In the process of performing distribution inference on the posterior distribution function, set the variational distribution.

[0064] In this embodiment, the variational distribution is set to approximate the complex posterior distribution by selecting a simplified, easy-to-handle distribution. The key to distributional reasoning is to introduce a simple hypothetical distribution (variational distribution), and usually a certain form of distribution family (such as Gaussian distribution) is selected to approximate the true posterior distribution. This simplified variational distribution can be calculated in a shorter time and provide an acceptable approximation. By setting the variational distribution, it is possible to use an optimization algorithm to find the optimal variational distribution without the need for complex integration, and finally approach the true posterior distribution. The choice of variational distribution will directly affect the accuracy and computational efficiency of subsequent reasoning.

[0065] Step 2122: Calculate the divergence between the variational distribution and the posterior distribution function to obtain the lower bound of evidence.

[0066] In this embodiment, when the variational distribution is set, it is necessary to evaluate the difference between the variational distribution and the true posterior distribution. The divergence between the variational distribution and the posterior distribution is calculated, and through this calculation, the evidence lower bound (ELBO) is obtained, which is an important indicator for measuring the approximation degree of the posterior distribution.

[0067] Divergence is a mathematical tool to quantify the difference between two probability distributions. Common divergence measurement methods include Kullback-Leibler (KL) divergence. The smaller the divergence, the closer the variational distribution is to the true posterior distribution, and the higher the accuracy of the distribution inference result. In distribution inference, the goal is to minimize the divergence to optimize the accuracy of the variational distribution.

[0068] For example, distributed reasoning transforms the reasoning problem into an optimization problem. Distributed reasoning introduces a set of distributed is called the variational distribution, is the parameter of this variational distribution, and then the parameter Optimize to find the member of the distribution family that is closest to the posterior of interest. The commonly used closeness measure is the Kullback-Leibler divergence, and the KL divergence is used to obtain the lower bound of the evidence. The posterior distribution that is difficult to solve directly is approximated by optimizing the approximate method, where the lower bound of the evidence is expressed as:

[0069]

[0070] In the formula, represents the divergence between the variational distribution and the posterior distribution function, θ represents the network parameters of the topology prediction model, represents the variational distribution, p(θ|D) represents the posterior distribution function, p(D,θ) represents the joint distribution, and p(D) represents the marginal likelihood. express Expected value, IE q [logp(D,θ)] represents the expected value of p(D,θ), represents the lower bound of evidence, and const represents a constant;

[0071] logp(D) is replaced by a constant const because logp(D) does not depend on θ, which represents the network parameters of the topology prediction model including local random parameters β and global deterministic parameters α.

[0072] Step 2123, optimize the lower bound of the evidence based on the loss function and Bayesian point estimation to obtain an approximate posterior as the posterior distribution of the topological model.

[0073] In this embodiment, the lower bound of evidence is obtained by setting the variational distribution and calculating the divergence, and the final approximate posterior of the variational distribution is obtained by optimizing the lower bound of evidence, so as to more accurately simulate the changes in the topological state of the power system. Optimizing the lower bound of evidence usually involves the use of loss functions and Bayesian point estimates, which are two tools that help adjust the variational distribution to make it as close to the true posterior distribution as possible. The loss function measures the gap between the variational distribution and the true distribution. Through Bayesian point estimation, the power system can select a specific point estimate (such as the maximum a posteriori estimate) instead of the complete distribution, thereby providing a more reasonable estimation result in the case of incomplete information. By optimizing the loss function, the power system can obtain an efficient and reliable approximate posterior distribution in practical applications. This approximate posterior provides a core basis for topological prediction, risk assessment and alarm of the power system.

[0074] For example, by calculating the above KL divergence, the lower bound of evidence can be defined In order to obtain the approximate posterior distribution of the topological model, it is necessary to minimize KL, which is equivalent to maximizing

[0075] According to the Bayesian decomposition Decompose into parts related to β The part related to α By optimizing the local random parameter β and the global deterministic parameter α, we can maximize The following 1-2 is to solve the part related to β Part related to α

[0076] 1. Solve the part related to β

[0077] Solve the dependence of β on various parameters in the parameter space:

[0078] p(β n |y n ,β -n ,α)=h(β n )exp{η(y n ,β -n ,α) T T(β n )-a(η(y n ,β -n ,α)};

[0079] In the formula, p(β n |y n ,β -n ,α) means that given y n ,β -n, in the case of α, β n The conditional probability distribution of n represents the nth local random parameter, y n represents the nth topological state, β -n Indicates the addition of β n The local random parameters other than h(β n ) represents β n The underlying measure, T(β n ) represents β n The sufficient statistic, η(y n ,β -n ,α) represents the natural parameter, a(η(y n ,β -n ,α) represents the cumulative function.

[0080] After solving the dependency relationship between the various parameters of β in the parameter space, the local random parameter β is iteratively optimized through the loss function. The loss function is expressed as:

[0081]

[0082] In the formula, represents the loss function, y n represents the nth topological state, β -n represents the local random parameter, α represents the global deterministic parameter, η(y n ,β -n ,α) represents y n ,β -n , the natural parameter of α, represents the parameters of the variational distribution, represents η(y n ,β -n ,α)’s expected value.

[0083] 2. Solve the part related to α

[0084] Based on empirical Bayesian point estimation, given β, the optimal setting of α can be obtained by maximizing the marginal likelihood p(Y|X,α,β). Considering that p(Y|X,α,β) cannot be calculated directly, the variational expectation maximization method is used to maximize To optimize α.

[0085]

[0086] In the formula, ξ represents the step size, α represents the global deterministic parameter, represents the lower bound of evidence for optimizing α, represents the estimated value of α after optimization.

[0087] After optimizing the local random parameter β and the global deterministic parameter α, random variational inference is introduced, and the distribution inference objective function is optimized by using a random optimization method with noisy natural gradient combined with a small batch sampling strategy. The local random parameter β and the variational distribution parameter φ are initialized with prior knowledge. Repeat the following steps: randomly extract a small batch D from D B , D is the input data of the preset topology prediction model; calculate the estimated value of the variational distribution parameter φ and the optimized α Update the global deterministic parameter α to The weighted average of the current value is calculated and the weight ω is updated. The iteration continues until the maximum iteration step is reached, and then the optimized parameters α, β, φ are returned.

[0088] Step 1022: Receive the generator active power at the current moment, the generator reactive power at the current moment, the load active power at the current moment, the load reactive power at the current moment, the topological state at the current moment, and the meteorological data at the next moment at the input layer, perform linear transformation on the generator active power at the current moment, the generator reactive power at the current moment, the load active power at the current moment, the load reactive power at the current moment, the topological state at the current moment, and the meteorological data at the next moment at the hidden layer to obtain sample features, and convert the sample features into the topological state of the power system at the next moment according to the posterior distribution at the output layer.

[0089] In this embodiment, by constructing a three-layer topology prediction model, the generator active power at the current moment, the generator reactive power at the current moment, the load active power at the current moment, the load reactive power at the current moment, the topological state of the power system at the current moment, and the meteorological data at the next moment are processed to obtain the topological state of the power system at the next moment. In the input layer, the power system obtains key parameters from the current moment, including the active and reactive power of the generator and the load, as well as the topological state of the power system and future meteorological data, which jointly affect the topological state of the power system. In the hidden layer, linear transformation is applied to process the input data to extract sample features that can reflect the characteristics of the power system. These sample features contain potential information about the transformation of the topological state of the power system. In the output layer, the posterior distribution and Bayesian reasoning technology are used to convert the obtained sample features into predictions of the future topological state of the power system and its probability of occurrence, providing a data-driven prediction mechanism for the power system, accurately identifying possible changes, reducing potential risks, and improving the stability and reliability of the power system operation.

[0090] For example, Figure 2 The network structure diagram of the topology prediction model is shown in Figure 1. The input layer X receives multiple input data x1, x2...x k(i.e., the generator active power at the current moment, the generator reactive power at the current moment, the load active power at the current moment, the load reactive power at the current moment, the topological state of the power system at the current moment, and the meteorological data at the next moment), the topological prediction model contains local random parameters β and global deterministic parameters α (i.e., α1...α m ), the global deterministic parameter α plays a role in the input data in the input layer and multiple hidden layers (i.e., linear transformation), and the local random parameter β plays a role in the output hidden layer in conjunction with the posterior distribution. The output layer Y outputs the predicted topological state of the power system at the next moment (i.e., y1...y n ).

[0091] Step 103: input the data samples of multiple historical moments, the topological state at the next moment, and the meteorological data at the next moment into a preset operation state prediction model for calculation to obtain the bus voltage at the next moment and the load rate at the next moment.

[0092] In this embodiment, by inputting historical data samples, predicted topological states and future meteorological data into the preset operation state prediction model, the bus voltage and load rate of the power system at the next moment can be accurately predicted. The historical data samples contain data samples of the power system at the current moment and in the past period of time. These data samples can provide the operation state prediction model with time series information about the operation of the power system; the topological state at the next moment reflects the layout and connection mode of each component of the power system, which directly affects the flow and distribution of electricity; and meteorological data can affect the efficiency of power demand and power generation, especially under extreme weather conditions. Combining these factors, the dynamic characteristics of the power system can be more comprehensively understood, and the stability of the power system can be predicted. By predicting the bus voltage and load rate, potential risks can be discovered in advance, and it can be judged whether the power system has the risk of overload, failure or instability, thereby providing an important basis for the optimized operation and risk management of the power system, improving the safety, reliability and efficiency of the power system, and timely warning to avoid power outages and other failures.

[0093] For example, Figure 3 As shown in the figure, it is a schematic diagram of the structure of the running state prediction model, which takes the data samples at multiple moments in history (i.e. ) is input into each LSTM module in the first layer (i.e., LSTMLayer1) of the double-layer long short-term memory network in time sequence, and the first time sequence feature is extracted. The first time sequence feature is input into each LSTM module in the second layer (i.e., LSTM Layer2) of the double-layer long short-term memory network, and the second time sequence feature is extracted. The topological state at the next moment and the meteorological data at the next moment are input into the Dense module in the first layer (i.e., FC Layer 1) of the double-layer fully connected network, and the first abstract feature is extracted. The first abstract feature is input into the Dense module in the second layer (i.e., FC Layer 2) of the double-layer fully connected network, and the second abstract feature is extracted. The second time sequence feature is spliced ​​with the second abstract feature to obtain the spliced ​​feature. The spliced ​​feature is calculated in the preset operation state prediction model, and the output layer outputs the bus voltage (V t+1 ) and the load rate at the next moment (i.e., p t+1 ), t+1 represents the next moment.

[0094] For the training of the operation status prediction model, multi-source heterogeneous spatiotemporal data is obtained by collecting historical data of the energy management system or simulation data of multiple platforms. Then, after data preprocessing, feature selection and data integration, the training samples of the operation status prediction model are obtained. The operation status prediction model can be trained offline or online.

[0095] Step 104: For the next moment, the risk value of the power system is calculated based on the topological state, bus voltage and load factor.

[0096] In this embodiment, the risk value of the system is calculated according to the topological state, bus voltage and load rate of the power system at the next moment to evaluate its stability and safety. This process quantifies the risks that the power system may face by comprehensively analyzing the topological state, bus voltage and load rate at the next moment. The topological state describes the connection and layout of each component (such as generators, transformers, transmission lines, etc.) in the power system, which directly affects the distribution of power flow and the stability of the system; the bus voltage reflects the distribution of voltage in the power system. Too high or too low voltage will cause equipment damage or unstable power supply; the load rate represents the load of each part of the power system. Overload will cause equipment failure or system collapse in the store. By combining the topological state, bus voltage and load rate at the next moment, calculating the risk value can identify potential hidden dangers in the power system, such as power overload, equipment failure or system instability. The calculation of the risk value can not only help identify potential risks in the current system state, but also provide a basis for subsequent alarm operations, thereby achieving early warning and preventing major failures.

[0097] Exemplarily, the risk value is expressed as:

[0098] risk(t+1,OS i )=p t+1 (Topo i )*Sev(V t+1 ,p t+1 );

[0099]

[0100] In the formula, risk(t+1,OS i ) represents the risk index, t represents the time, OS i Indicates the combination of bus voltage and load factor, Topo i represents the topological state, p t+1 (Topo i ) represents the probability of the topological state at the next moment, V t+1 represents the bus voltage at the next moment, p t+1 Indicates the load factor of the transmission line and transformer at the next moment, Sev(V t+1 ,p t+1 ) represents the severity between the bus voltage at the next moment and the load rate of the transmission line and transformer at the next moment, α i represents the i-th global deterministic parameter, β i represents the jth local random parameter, N represents the number of global deterministic parameters, L represents the number of local random parameters, Sev(V t+1 ) indicates the severity of the bus voltage at the next moment, Sev(p t+1 ) represents the severity of the load rate of the transmission line and transformer at the next moment, Risk represents the risk value, Ω t Represents a set of topological states, if represents a conditional statement, when if(V t+1 )<0.9 means when V t+1 When Sev(V t+1 ) is equal to 0.9 and V t+1 The difference between; when 0.9≤if(V t+1 )≤1.1 means when V t+1 Between 0.9 and 1.1, Sev(V t+1 ) is equal to 0; when if(V t+1 )>1.1 means when V t+1 Greater than 1.1, Sev(V t+1 ) is equal to V t+1 The difference between 1.1 and 1.1; when if(p t+1 )≤0.9 means p t+1 Within 0.9, Sev(p t+1 ) is equal to 0; when if(pt+1 )>0.9 indicates p t+1 Greater than 0.9, Sev(p t+1 ) is equal to p t+1 The difference between 0.9 and 0.9.

[0101] Step 105: Perform an alarm operation on the power system according to the risk value.

[0102] In this embodiment, the risk value reflects the risk level of possible failure or instability of the power system in the future, identifies and responds to potential failures or unstable conditions in the power system, and triggers an alarm operation based on the calculated risk value, thereby ensuring the stability and reliability of the power system.

[0103] Exemplarily, the embodiment of the present invention performs alarm operation on the power system in two ways as follows:

[0104] Method 1: If the risk value is greater than or equal to the preset risk threshold, an alarm operation is performed on the power system.

[0105] Method 2: Based on the risk values ​​at multiple moments in the future, the system state of the power system is constructed, and multiple consecutive system states are screened out according to the preset time step to obtain the state trajectory. By observing the system state evolution path, the dynamic changes that may occur in the power system over a long period of time can be captured, thereby identifying possible emergencies or long-term accumulated problems;

[0106] For multiple state trajectories, if a power outage occurs in the state trajectory, the fault information is marked on the state trajectory, and the topological state where the power outage occurs is recorded as the target topological structure. The possible transmission line faults in the state trajectory are marked, and the topological state at the time of the fault is recorded, which provides accurate background information for future alarm operations, and also provides important basis for fault cause analysis, prediction and solution formulation;

[0107] The ratio between the number of fault information and the number of state trajectories is calculated as the fault probability. The frequency or possibility of faults in the power system can be quantified through the ratio of the number of fault information and state trajectories, thereby providing data support for alarm decision-making. If the fault probability is high, it means that the power system faces greater risks in the future, and the triggering conditions of the alarm operation will be more stringent.

[0108] Perform alarm operations on the power system based on the target topology and fault probability.

[0109] In order to better understand this embodiment, the following Figure 4 The figure is a schematic diagram of the structure of the power system risk alarm, showing the execution process of the embodiment of the present invention:

[0110] In the data preprocessing module, a multi-source database is formed by collecting and generating historical data and simulation data (i.e., constructing data samples at each moment). In the topology prediction model, a training set of the multi-source database is received, including topology data, PF data (i.e., tidal data), and meteorological data. These training sets are the generator active power at the current moment, the generator reactive power at the current moment, the load active power at the current moment, the load reactive power at the current moment, the topological state of the power system at the current moment, and the meteorological data at the next moment. These training sets are input into the topology prediction model for training, and the topological state of the power system at the next moment (such as Topo1...Topo N ) and the probability corresponding to the topological state at the next moment (such as Prob1....Prob N );

[0111] In the operation state prediction model, data from a multi-source database is received to form a training set of the operation state prediction model, including topological data, PF data, meteorological data and prediction data. Feature selection and data integration are performed on the data of the training set to form an operation state prediction model knowledge base. The operation state prediction model knowledge base data and the topological state of the power system output by the topological prediction model at the next moment are input into the operation state prediction model together to obtain the bus voltage of the power system at the next moment and the load rate at the next moment.

[0112] In the risk assessment and early warning model, the probabilities corresponding to the bus voltage at the next moment, the load rate at the next moment, and the topological state at the next moment are input into the severity calculation model, and the risk value (i.e., Sev*Prob) of each state (such as state 1...state N) (i.e., the topological state at the next moment) is output, and risk assessment and early warning are performed based on the risk value.

[0113] In the embodiment of the present invention, data samples are constructed at each moment, each data sample includes bus voltage, generator active power, generator reactive power, load active power, load reactive power, topological state of the power system, load rate and meteorological data of transmission lines and transformers, the generator active power at the current moment, the generator reactive power at the current moment, the load active power at the current moment, the load reactive power at the current moment, the topological state at the current moment and the meteorological data at the next moment are input into a preset topological prediction model, the topological state at the next moment is calculated, the data samples at multiple moments in history, the topological state at the next moment and the meteorological data at the next moment are input into a preset operation state prediction model for operation, the bus voltage at the next moment and the load rate at the next moment are obtained, and for the next moment, the risk value of the power system is calculated according to the topological state, the bus voltage and the load rate, and the power system is alarmed according to the risk value. Through the prediction of the topological state, the operation state and the risk value, potential faults can be discovered in advance and early warnings can be given, thereby reducing the risk of power outages and accidents. At the same time, the combination of meteorological data enhances the ability of the power system to respond to emergencies. It comprehensively considers data from multiple aspects and uses topology prediction models and operating status prediction models to predict the risks of the power system in turn, thereby strengthening the correlation between multiple data and improving the accuracy of power system risk prediction.

[0114] Embodiment 2

[0115] Figure 5 A schematic diagram of the structure of a power system risk warning device provided in the second embodiment of the present invention is shown in FIG. Figure 5 As shown, the device comprises:

[0116] The data sample construction module 501 is used to construct data samples at various times; each of the data samples includes bus voltage, generator active power, generator reactive power, load active power, load reactive power, topological state of the power system, load rate of transmission lines and transformers, and meteorological data;

[0117] A topology prediction module 502 is used to input the generator active power at the current moment, the generator reactive power at the current moment, the load active power at the current moment, the load reactive power at the current moment, the topology state at the current moment, and the meteorological data at the next moment into a preset topology prediction model to calculate the topology state at the next moment;

[0118] The motion state prediction module 503 is used to input the data samples at multiple moments in history, the topological state at the next moment and the meteorological data at the next moment into a preset operation state prediction model for calculation, so as to obtain the bus voltage at the next moment and the load rate at the next moment;

[0119] A risk value calculation module 504 is used to calculate the risk value of the power system according to the topological state, the bus voltage and the load rate at the next moment;

[0120] The power system alarm module 505 is used to perform an alarm operation on the power system according to the risk value.

[0121] In one embodiment of the present invention, the topology prediction module 502 includes:

[0122] A posterior distribution setting module, used to set a posterior distribution for a preset topology prediction model; the topology model includes an input layer, a hidden layer and an output layer;

[0123] An input data processing module is used to receive the generator active power at the current moment, the generator reactive power at the current moment, the load active power at the current moment, the load reactive power at the current moment, the topological state at the current moment, and the meteorological data at the next moment at the input layer, and to perform linear transformation on the generator active power at the current moment, the generator reactive power at the current moment, the load active power at the current moment, the load reactive power at the current moment, the topological state at the current moment, and the meteorological data at the next moment at the hidden layer to obtain sample features, and to convert the sample features into the topological state of the power system at the next moment according to the posterior distribution at the output layer.

[0124] In one embodiment of the present invention, the posterior distribution setting module includes:

[0125] A posterior distribution function construction module is used to construct a posterior distribution function for a preset topology prediction model according to the Bayesian theorem;

[0126] An approximate posterior acquisition module, used for performing distribution reasoning according to the posterior distribution function to obtain an approximate posterior as the posterior distribution of the topological model;

[0127] Wherein, the posterior distribution function is expressed as:

[0128]

[0129] Wherein, p(θ|D) represents the posterior distribution function, θ represents the network parameters of the topology prediction model, D represents the data sample, p(θ) represents the prior distribution, p(D|θ) represents the likelihood function, and p(D) represents the marginal likelihood. represents the i-th data sample, represents the i-th topological state, Indicated in and θ, and N represents the number of data samples.

[0130] In one embodiment of the present invention, the approximate posterior acquisition module includes:

[0131] A variational distribution setting module, used for setting a variational distribution in the process of performing distribution inference on the posterior distribution function;

[0132] An evidence lower bound acquisition module, used to calculate the divergence between the variational distribution and the posterior distribution function to obtain an evidence lower bound;

[0133] An evidence lower bound optimization module, used for optimizing the evidence lower bound according to a loss function and a Bayesian point estimate, to obtain an approximate posterior as a posterior distribution of the topological model;

[0134] Among them, the lower bound of the evidence is expressed as:

[0135]

[0136] In the formula, represents the divergence between the variational distribution and the posterior distribution function, θ represents the network parameters of the topology prediction model, represents the variational distribution, p(θ|D) represents the posterior distribution function, p(D,θ) represents the joint distribution, p(D) represents the marginal likelihood, express Expected value, IE q [logp(D,θ)] represents the expected value of p(D,θ), represents the lower bound of evidence, and const represents a constant;

[0137] The loss function is expressed as:

[0138]

[0139] In the formula, represents the loss function, y n represents the nth topological state, β -n represents the local random parameter, α represents the global deterministic parameter, η(y n ,β -n ,α) represents y n ,β -n , the natural parameter of α, denotes the parameters of the variational distribution, represents η(y n ,β -n ,α)’s expected value.

[0140] In one embodiment of the present invention, the running state prediction model includes a two-layer long short-term memory network and a two-layer fully connected network; the motion state prediction module 503 includes:

[0141] A first time series feature acquisition module, used to input the data samples at multiple moments in history into the first layer of the double-layer long short-term memory network in time sequence to extract a first time series feature;

[0142] A second time series feature acquisition module, used for inputting the first time series feature into the second layer of the double-layer long short-term memory network to extract a second time series feature;

[0143] A first abstract feature acquisition module, used for inputting the topological state at the next moment and the meteorological data at the next moment into the first layer of the two-layer fully connected network to extract the first abstract feature;

[0144] A second abstract feature acquisition module, used for inputting the first abstract feature into the second layer of the two-layer fully connected network to extract a second abstract feature;

[0145] A feature concatenation module, used for concatenating the second temporal feature with the second abstract feature to obtain a concatenated feature;

[0146] The splicing feature calculation module is used to calculate the splicing feature in a preset operation state prediction model to obtain the bus voltage at the next moment and the load rate at the next moment.

[0147] In one embodiment of the present invention, the risk value is expressed as:

[0148] risk(t+1,OS i )=p t+1 (Topo i )*Sev(V t+1 ,p t+1 );

[0149]

[0150] In the formula, risk(t+1,OS i ) represents the risk index, t represents the time, OS i Represents the combination of the bus voltage and the load factor, Topo i represents the topological state, p t+1 (Topo i ) represents the probability of the topological state at the next moment, V t+1 represents the bus voltage at the next moment, p t+1 represents the load factor of the transmission line and transformer at the next moment, Sev(V t+1 ,pt+1 ) represents the severity between the bus voltage at the next moment and the load rate of the transmission line and transformer at the next moment, α i represents the i-th global deterministic parameter, β i represents the jth local random parameter, N represents the number of global deterministic parameters, L represents the number of local random parameters, Sev(V t+1 ) represents the severity of the bus voltage at the next moment, Sev(p t+1 ) represents the severity of the load rate of the transmission line and transformer at the next moment, Risk represents the risk value, Ω t Represents the set of topological states, and if represents a conditional judgment statement.

[0151] In one embodiment of the present invention, the power system alarm module 505 includes:

[0152] A risk value judgment module, configured to perform an alarm operation on the power system if the risk value is greater than or equal to a preset risk threshold;

[0153] and / or,

[0154] A system state construction module, used for constructing a system state of the power system according to the risk values ​​at multiple moments in the future;

[0155] A state trajectory acquisition module is used to filter out a plurality of continuous system states according to a preset time step to obtain a state trajectory;

[0156] A target topology structure recording module is used for marking fault information on a plurality of the state trajectories if a power outage of a transmission line occurs in the state trajectories, and recording the topology state where the power outage of the transmission line occurs as a target topology structure;

[0157] A fault probability calculation module, used to calculate the ratio between the number of the fault information and the number of the state trajectories as the fault probability;

[0158] An alarm operation module is used to perform an alarm operation on the power system according to the target topology and the fault probability.

[0159] The power system risk warning device provided in the embodiment of the present invention can execute the power system risk warning method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the power system risk warning method.

[0160] Embodiment 3

[0161] See also Figure 6, showing a schematic diagram of the structure of a computer device provided by an embodiment of the present invention. The computer device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, blade servers, mainframe computers, and other suitable computers. The computer device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0162] like Figure 6 As shown, the computer device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the computer device 10 can also be stored. The processor 11, ROM 12 and RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0163] A number of components in the computer device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the computer device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0164] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as the power system risk warning method.

[0165] In some embodiments, the power system risk warning method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the computer device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the power system risk warning method described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to execute the power system risk warning method in any other appropriate manner (e.g., by means of firmware).

[0166] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0167] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0168] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0169] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer device. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0170] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0171] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0172] Embodiment 4

[0173] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the power system risk warning method provided by any embodiment of the present invention is implemented.

[0174] In the process of implementation, the computer program product can be written in one or more programming languages ​​or a combination thereof to perform the computer program code of the present invention, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0175] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0176] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A power system risk warning method, characterized in that: include: Constructing data samples at various times; each of the data samples includes bus voltage, generator active power, generator reactive power, load active power, load reactive power, topological state of the power system, load rate of transmission lines and transformers, and meteorological data; Input the generator active power at the current moment, the generator reactive power at the current moment, the load active power at the current moment, the load reactive power at the current moment, the topological state at the current moment, and the meteorological data at the next moment into a preset topological prediction model to calculate the topological state at the next moment; Input the data samples at multiple historical moments, the topological state at the next moment, and the meteorological data at the next moment into a preset operation state prediction model for calculation, so as to obtain the bus voltage at the next moment and the load rate at the next moment; For the next moment, calculating the risk value of the power system according to the topological state, the bus voltage and the load rate; An alarm operation is performed on the power system according to the risk value.

2. The method according to claim 1, characterized in that The step of inputting the generator active power at the current moment, the generator reactive power at the current moment, the load active power at the current moment, the load reactive power at the current moment, the topological state at the current moment, and the meteorological data at the next moment into a preset topological prediction model to calculate the topological state at the next moment includes: Setting a posterior distribution for a preset topology prediction model; the topology model includes an input layer, a hidden layer and an output layer; The input layer receives the generator active power at the current moment, the generator reactive power at the current moment, the load active power at the current moment, the load reactive power at the current moment, the topological state at the current moment, and the meteorological data at the next moment, and the hidden layer performs linear transformation on the generator active power at the current moment, the generator reactive power at the current moment, the load active power at the current moment, the load reactive power at the current moment, the topological state at the current moment, and the meteorological data at the next moment to obtain sample features, and the output layer converts the sample features into the topological state of the power system at the next moment according to the posterior distribution.

3. The method according to claim 2, characterized in that The step of setting a posterior distribution for a preset topology prediction model includes: Construct a posterior distribution function for the preset topology prediction model based on Bayes' theorem; Performing distribution inference according to the posterior distribution function to obtain an approximate posterior as the posterior distribution of the topological model; Wherein, the posterior distribution function is expressed as: Wherein, p(θ|D) represents the posterior distribution function, θ represents the network parameters of the topology prediction model, D represents the data sample, p(θ) represents the prior distribution, p(D|θ) represents the likelihood function, and p(D) represents the marginal likelihood. represents the i-th data sample, represents the i-th topological state, Indicated in and θ, and N represents the number of data samples.

4. The method according to claim 3, characterized in that The performing distribution inference according to the posterior distribution function to obtain an approximate posterior as the posterior distribution of the topological model includes: In the process of performing distributional inference on the posterior distribution function, setting a variational distribution; Calculating the divergence between the variational distribution and the posterior distribution function to obtain an evidence lower bound; Optimizing the lower bound of the evidence according to a loss function and a Bayesian point estimate to obtain an approximate posterior as the posterior distribution of the topological model; Among them, the lower bound of the evidence is expressed as: In the formula, represents the divergence between the variational distribution and the posterior distribution function, θ represents the network parameters of the topology prediction model, represents the variational distribution, p(θ|D) represents the posterior distribution function, p(D,θ) represents the joint distribution, p(D) represents the marginal likelihood, express Expected value, IE q [logp(D,θ)] represents the expected value of p(D,θ), represents the lower bound of evidence, and const represents a constant; The loss function is expressed as: In the formula, represents the loss function, y n represents the nth topological state, β -n represents the local random parameter, α represents the global deterministic parameter, η(y n ,β -n ,α) represents y n ,β -n , the natural parameter of α, denotes the parameters of the variational distribution, represents η(y n ,β -n ,α)’s expected value.

5. The method according to claim 1, characterized in that The operation state prediction model includes a two-layer long short-term memory network and a two-layer fully connected network; the data samples at multiple historical moments, the topological state at the next moment, and the meteorological data at the next moment are input into the preset operation state prediction model for calculation to obtain the bus voltage at the next moment and the load rate at the next moment, including: Inputting the data samples at multiple moments in history into the first layer of the double-layer long short-term memory network in time sequence to extract first time sequence features; Inputting the first time series feature into the second layer of the two-layer long short-term memory network to extract the second time series feature; Inputting the topological state at the next moment and the meteorological data at the next moment into the first layer of the two-layer fully connected network to extract the first abstract feature; Inputting the first abstract feature into the second layer of the two-layer fully connected network to extract the second abstract feature; Concatenate the second temporal feature with the second abstract feature to obtain a concatenated feature; The splicing feature is operated in a preset operation state prediction model to obtain the bus voltage at the next moment and the load rate at the next moment.

6. The method according to any one of claims 1 to 5, characterized in that: The risk value is expressed as: risk(t+1,OS i )=p t+1 (Topo i )*North(V t+1 ,p t+1 ); In the formula, risk(t+1,OS i ) represents the risk index, t represents the time, OS i Represents the combination of the bus voltage and the load factor, Topo i represents the topological state, p t+1 (Topo i ) represents the probability of the topological state at the next moment, V t+1 represents the bus voltage at the next moment, p t+1 represents the load factor of the transmission line and transformer at the next moment, Sev(V t+1 ,p t+1 ) represents the severity between the bus voltage at the next moment and the load rate of the transmission line and transformer at the next moment, α i represents the i-th global deterministic parameter, β i represents the jth local random parameter, N represents the number of global deterministic parameters, L represents the number of local random parameters, Sev(V t+1 ) represents the severity of the bus voltage at the next moment, Sev(p t+1 ) represents the severity of the load rate of the transmission line and transformer at the next moment, Risk represents the risk value, Ω t Represents the set of topological states, and if represents a conditional judgment statement.

7. The method according to claim 6, characterized in that The step of issuing an alarm to the power system according to the risk value includes: If the risk value is greater than or equal to a preset risk threshold, an alarm operation is performed on the power system; and / or, constructing a system state of the power system according to the risk value at multiple future moments; Filter out a plurality of continuous system states according to a preset time step to obtain a state trajectory; For the plurality of state trajectories, if a power outage of a transmission line occurs in the state trajectories, fault information is marked on the state trajectories, and the topological state in which the power outage of the transmission line occurs is recorded as a target topological structure; Calculating a ratio between the number of the fault information and the number of the state trajectories as a fault probability; An alarm operation is performed on the power system according to the target topology and the fault probability.

8. A computer device, characterized in that: The computer device comprises: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the power system risk warning method as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the power system risk warning method as described in any one of claims 1-7 is implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the power system risk warning method as described in any one of claims 1 to 7 is implemented.

Images