Distribution network situation prediction method and system based on state assessment

Abstract

The present disclosure belongs to the technical field of distribution networks, and specifically relates to a distribution network situation prediction method and system based on state assessment, comprising: acquiring real-time operation data of the distribution network; calculating a distribution network security situation index based on the acquired real-time operation data of the distribution network; using an autoregressive sliding average model to predict node voltages at future times based on the historical operation data of the distribution network and the obtained distribution network security situation index, calculating the predicted value of the distribution network security situation index, and realizing short-term prediction of the distribution network security situation awareness index.

Description

Technical Field

[0001] The present disclosure belongs to the technical field of distribution network, and particularly relates to a distribution network situation prediction method and system based on state assessment. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] The distribution network is responsible for distributing electricity to end users within the power system, directly impacting the quality of the system's power supply. Its importance is self-evident. With rapid economic development, energy demand and electricity consumption are rapidly increasing, leading to increasing attention for renewable energy and distributed generation. Distributed energy resources are typically directly connected to the distribution network. With the integration of diverse loads, the operation and control of the distribution network have become increasingly complex and diverse, increasing the risks faced by distribution systems during operation.

[0004] The inventors understand that the operational status of a distribution network refers to the current state and changing trends of electrical equipment, network power flow, and user-side electricity consumption. Distribution network status awareness is an important technical means of understanding the operational status of a distribution network. This allows for the recognition and understanding of the various factors affecting distribution network operation within a specific time and space, effectively grasping the operational status of the distribution network and improving the reliability of grid operation. Summary of the Invention

[0005] In order to solve the above problems, the present disclosure proposes a distribution network status prediction method and system based on state assessment, which predicts the changing trend of the voltage of each node, facilitates a comprehensive grasp of the changes in various state quantities, timely grasps the operating status of the distribution network, obtains the state quantity of the future state, and then evaluates the operating status at future moments, timely discovers abnormal power supply and potential faults, thereby strengthening the management of the distribution network and improving the level of economic operation.

[0006] According to some embodiments, a first solution of the present disclosure provides a distribution network situation prediction method based on state assessment, which adopts the following technical solutions:

[0007] A distribution network situation prediction method based on state assessment, comprising:

[0008] Obtain real-time operation data of the distribution network;

[0009] Calculate the distribution network security status indicators based on the acquired real-time operation data of the distribution network;

[0010] Based on the historical operation data of the distribution network and the obtained distribution network security situation indicators, the autoregressive sliding average model is used to predict the node voltage at future times, calculate the predicted value of the distribution network security situation indicators, and realize the short-term prediction of the distribution network security situation awareness indicators.

[0011] As a further technical limitation, in the process of obtaining the real-time operation data of the distribution network, the state detection information of the distribution network is obtained through the synchronous vector measurement unit, the remote terminal unit and the feeder terminal unit respectively to obtain the real-time operation data of the distribution network.

[0012] As a further technical limitation, the specific process of calculating the distribution network security situation index is as follows:

[0013] Perform multi-time-scale state estimation on the acquired real-time operation data of the distribution network to obtain the node voltage phasor information within the update period;

[0014] Calculating the branch current and node injection current of the distribution network node according to the obtained node voltage phasor information;

[0015] Calculate the complex power of the node injection power and the complex power of the branch power according to the obtained branch current and node injection current;

[0016] The injected power at the node where the distribution transformer is located is regarded as the distribution transformer load, and the injected power at the node where the main transformer is located is regarded as the main transformer load. The main transformer overload rate, distribution transformer overload rate, line overload rate and line severe overload ratio are calculated in real time to complete the risk assessment of the distribution network safety situation.

[0017] Furthermore, whether the node voltage is out of bounds is determined based on the obtained node voltage phasor information, and it is determined that the node voltage exceeds the limit value.

[0018] As a further technical limitation, the autoregressive moving average model adopts time series analysis, and the process of establishing the autoregressive moving average model is:

[0019] The input data is judged to determine whether it is a stationary non-pure random sequence. If it is stationary, it will directly proceed to the next step; otherwise, data processing is required before proceeding to the next step.

[0020] The stationary non-pure random sequence is identified and ordered by using autocorrelation and partial autocorrelation functions in combination with AIC criterion or BIC criterion, and parameter estimation is performed;

[0021] After completing the parameter estimation, the adaptability test of the fitted model is carried out. If the fitted model passes the test, the prediction phase begins; if the model fails the test, the identification and test are repeated.

[0022] Use highly adaptable fitting models to predict future trends in the sequence.

[0023] Furthermore, the specific process of predicting the node voltage at a future moment is as follows:

[0024] Reading historical operation data of the distribution network, wherein the historical operation data of the distribution network includes node voltage amplitudes and phase angles;

[0025] Select a certain time series to calculate the correlation between the voltage state and historical state information in the time series, and select historical state information with strong correlation as the independent variable of the autoregressive moving average model;

[0026] Construct a prediction matrix based on covariance and least squares estimation;

[0027] The voltage of each node at the future time t+1 is predicted based on the matrix at the latest time t.

[0028] As a further technical limitation, the specific process of calculating the predicted values ​​of the distribution network safety situation indicators is as follows: based on the obtained node voltage short-term situation prediction information, the voltage over-limit index is calculated, the distribution network node injection current and branch current are calculated through the admittance matrix, and the node injection power and node flow power are calculated based on the current and voltage to obtain the predicted values ​​of various distribution network safety situation indicators.

[0029] According to some embodiments, a second solution of the present disclosure provides a distribution network situation prediction system based on state assessment, which adopts the following technical solutions:

[0030] A distribution network situation prediction system based on state assessment, comprising:

[0031] An acquisition module configured to acquire real-time operation data of the distribution network;

[0032] A calculation module is configured to calculate a distribution network security situation indicator based on the acquired real-time distribution network operation data;

[0033] The prediction module is configured to use the autoregressive sliding average model to predict the node voltage at future times based on the historical operation data of the distribution network and the obtained distribution network security situation indicators, calculate the predicted value of the distribution network security situation indicators, and realize the short-term prediction of the distribution network security situation awareness indicators.

[0034] According to some embodiments, a third solution of the present disclosure provides a computer-readable storage medium, which adopts the following technical solution:

[0035] A computer-readable storage medium stores a program thereon, which, when executed by a processor, implements the steps in the distribution network situation prediction method based on state assessment as described in the first aspect of the present disclosure.

[0036] According to some embodiments, a fourth solution of the present disclosure provides an electronic device, which adopts the following technical solution:

[0037] An electronic device comprises a memory, a processor and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the distribution network situation prediction method based on state assessment as described in the first aspect of the present disclosure are implemented.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] The present invention discloses a security situation prediction based on state estimation, predicts the changing trend of the voltage of each node, timely grasps the operating situation of the distribution network, obtains the state quantity of the future state, and then evaluates the operating status at future moments, timely discovers abnormal power supply and potential faults, thereby strengthening the management of the distribution network and improving the level of economic operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings, which constitute a part of the present disclosure, are used to provide a further understanding of the present disclosure. The exemplary embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation to the present disclosure.

[0041] Figure 1 is a flow chart of a distribution network situation prediction method based on state assessment in the first embodiment of the present disclosure;

[0042] Figure 2 This is a block diagram of the distribution network situation awareness security assessment indicator in the first embodiment of the present disclosure;

[0043] Figure 3 This is a block diagram of security situation assessment based on PMU state estimation in the first embodiment of the present disclosure;

[0044] Figure 4 This is a flowchart of security situation prediction based on state estimation in the first embodiment of the present disclosure;

[0045] Figure 5 is a comparison diagram of the voltage amplitude change of node 3 in the first embodiment of the present disclosure;

[0046] Figure 6 is a comparison diagram of the voltage amplitude change of node 5 in the first embodiment of the present disclosure;

[0047] Figure 7 is a comparison diagram of the voltage amplitude change of node 13 in the first embodiment of the present disclosure;

[0048] Figure 8 is a comparison diagram of the voltage phase angle change of node 3 in the first embodiment of the present disclosure;

[0049] Figure 9 is a comparison diagram of the voltage phase angle change of node 3 in the first embodiment of the present disclosure;

[0050] Figure 10This is a structural block diagram of the distribution network situation prediction system based on state assessment in the second embodiment of the present disclosure. DETAILED DESCRIPTION

[0051] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0052] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present disclosure belongs.

[0053] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0054] In the absence of conflict, the embodiments of the present disclosure and the features thereof may be combined with each other.

[0055] Example 1

[0056] Embodiment 1 of the present disclosure introduces a distribution network situation prediction method based on state assessment.

[0057] like Figure 1 A distribution network situation prediction method based on state assessment is shown, comprising:

[0058] Obtain real-time operation data of the distribution network;

[0059] Calculate the distribution network security situation indicators based on the acquired real-time operation data of the distribution network;

[0060] Based on the historical operation data of the distribution network and the obtained distribution network security situation indicators, the autoregressive sliding average model is used to predict the node voltage at future times, calculate the predicted value of the distribution network security situation indicators, and realize the short-term prediction of the distribution network security situation awareness indicators.

[0061] As one or more implementation methods, this embodiment targets the important components of the 10kV regional distribution network: main transformers, distribution transformers and 10kV feeder lines. The indicator types are divided into two categories: overload indicators and risk indicators. The three evaluation tasks are divided into normal operation evaluation, K(N-1+1) verification and risk evaluation. Figure 2 As shown:

[0062] Among them, the overload indicators are divided into the main transformer load rate, the 10kV line heavy overload ratio, the distribution transformer load rate and the voltage over-limit value. The short-term state estimation based on the synchronous phase measurement synchronous phasor measurement unit (PMU) provides real-time updated node voltage monitoring information. The operation center can directly judge whether the voltage is over-limit online based on the node voltage information. Other overload indicators cannot be obtained directly, but can be calculated online based on the network relationship: the branch current and node injection current are calculated in real time according to the grid structure and parameters. Since the current and voltage states are accompanied by phase angle information, the branch power and node injection power can be obtained online through complex power calculation. The load of the main transformer and distribution transformer can be regarded as the injection power of the node where the transformer is located. The load rate of the main transformer and matching transformer is calculated from the state estimation result. Similarly, the heavy overload ratio of the 10kV line can be obtained by calculating the power flowing through each line to calculate the safety situation indicator.

[0063] While remote terminal units (RTUs) and feeder terminal units (FTUs) can also measure branch and node power, their accuracy is low and the sampling intervals are long, which doesn't meet the requirements for situational awareness, which requires accurate and real-time perception of system status. While PMUs offer powerful performance, their configuration is limited. Therefore, using multi-timescale state estimation based on hybrid PMU measurements for online security situation assessment is more appropriate than directly applying state monitoring information. Based on the online calculation results of overload indicators, statistical analysis methods can be used to obtain distribution network risk indicators online. Based on this indicator evaluation process, the operations center can obtain complete, real-time security situation assessment indicators.

[0064] like Figure 3 As shown in Figure 2, the specific process of calculating the distribution network security status indicators in real time based on the state estimation results is as follows:

[0065] Acquire status monitoring information of measurement equipment such as PMU, RTU, and FTU, and perform multi-timescale state estimation. Multi-timescale state estimation can provide node voltage phasor information U with an update period of 1s;

[0066] Directly judge whether the voltage of each node exceeds the limit based on the node voltage phasor information U;

[0067] The node injection current can be calculated based on the voltage at both ends of the node and the impedance of the intermediate branch by calculating the current IL of each branch of the regional distribution network or directly using the regional distribution network admittance matrix Y. According to the node Kirchhoff's current law (i.e., KCL principle), the node injection current can be calculated:

[0068] I ij =(U i-U j ) / Z ij (1)

[0069] I=YU (2)

[0070]

[0071] Where j is the node adjacent to node i, and N is the number of nodes adjacent to node i. Since the voltage in the equation is phasor information, the calculated current also includes phase angle information.

[0072] According to the calculated branch current, node current and node voltage, the complex power of branch power and node injection power can be obtained respectively by quantizing them into rectangular coordinate system and multiplying them;

[0073]

[0074]

[0075] The injected power at the node where the distribution transformer is located is regarded as the distribution transformer load, and the injected power at the node where the main transformer is located is regarded as the main transformer load. The main transformer overload rate, distribution transformer overload rate, main line overload rate and main line heavy overload ratio are calculated in real time.

[0076] Based on real-time safety indicator assessment scores and historical safety situation data, analysis is conducted, and statistical methods are used to analyze the overall risk of each device and regional distribution network system.

[0077] The security situation prediction process based on PMU hybrid measurement state estimation results is as follows: Figure 4 As shown in the figure, since state estimation provides real-time information updated every 1 second, the results based on state estimation should predict the state within a shorter timeframe. Long-term state law analysis should consider the daily and weekly periodic fluctuations of load and the system control method. It is not appropriate to rely solely on the situation prediction results based on state estimation. The prediction target for the safety situation prediction based on state estimation is set to the safety situation for a maximum of 5 minutes, based on the 5-minute standard period of ultra-short-term load fluctuations. Based on another research topic of this project, load forecasting, the load changes within the next hour (one prediction time point every 15 minutes) are obtained, and offline power flow calculations are performed to obtain the safety situation index evaluation score every 15 minutes. The two scores are weighted according to the operating status of the distribution network, and the final safety situation prediction result for the regional distribution network is obtained. This considers both the short-term volatility and randomness of the regional distribution network and the periodic laws and long-term development trends of the distribution network load fluctuations.

[0078] The model used to analyze online state estimation results and predict security situation indicators is the ARMA prediction model, also known as the Autoregressive Moving Average model. This model uses a time series analysis method. This method makes predictions by analyzing historical information. It can calculate a series of future information arranged in chronological order and can be dynamically updated according to the progress of time. Therefore, it meets the requirements of online security situation prediction.

[0079] A time series is a sequence of data points arranged in chronological order. Typically, the time interval of a time series is a constant (e.g., 1 second, 5 minutes, 12 hours, 7 days, 1 year), so time series can be analyzed and processed as discrete-time data. Time series are widely used in mathematical statistics, signal processing, pattern recognition, econometrics, mathematical finance, weather forecasting, earthquake prediction, electroencephalography, control engineering, aeronautics, communications engineering, and most other applied sciences and engineering involving temporal data measurement.

[0080] The basic idea of ​​time series analysis is to establish a mathematical model that can more accurately reflect the dynamic dependencies contained in the sequence based on the system's operating status information (status monitoring data) within a limited time, and to use it to predict the future of the system.

[52] This dynamic dependency is designed to satisfy the stability condition. Assuming that the time series variables are randomly extracted from a random process and arranged in time, there must be a stable trend, that is, the time series eventually tends to a constant or a linear function.

[0081] However, in reality, the relationship between the historical information provided and the future information required to be predicted is often unstable, and the variance between the information is not fixed. Therefore, it is necessary to adopt a corresponding model to solve the problem of time series variance fluctuation and time series trend instability. Time series generally exhibit the following characteristics:

[0082] 1. Long-term trend changes: The time series shows a certain tendency, which grows or decreases steadily according to certain rules.

[0083] 2. Seasonal cyclical changes: Affected by factors such as seasonal changes, the sequence changes regularly according to a certain period.

[0084] 3. Cyclic change: Periodic change with irregular period.

[0085] 4. Random changes: sequence changes caused by many uncertain factors.

[0086] Taking the voltage variations of regional distribution networks as an example, regional distribution networks exhibit long-term trend changes and seasonal cyclical variations on long time scales of years, months, and days. However, when the time scale is shortened to within an hour, both deterministic and random variations coexist. If the time scale is shortened to five minutes, random variations become dominant. Based on the different variation characteristics of time series, time series analysis methods are divided into deterministic variation analysis and random variation analysis. Deterministic variation analysis focuses on the characteristics of long-term trend changes and cyclical variations, and includes trend variation analysis, cyclic variation analysis, and cyclic variation analysis. Although random variation analysis only focuses on random variation characteristics, because there are multiple methods to address time-varying variance and trend instability, each with its own advantages and disadvantages, there are also various models. Currently, the most widely used are the AR (autoregressive model), the MA (moving average model), and the ARMA model, which combines the characteristics of both models.

[0087] Although the time series method is simple and easy to use, and can predict future states using linear error statistical analysis methods, it is not suitable for long-term predictions because it does not focus on the physical nature behind the data.

[0088] The autoregressive model (AR) primarily addresses the problem of non-constant error variance in time series. It uses data from different historical states of the same variable x, from x1 to xt-1, to analyze the future time xt, assuming a linear relationship between these time series data. This is essentially a linear regression analysis method, but rather than using binary variables x and y, it uses historical data from x to predict x itself; hence the name autoregression.

[0089] The autoregressive model is:

[0090]

[0091] Where c is the term in the linear fit; ε t is a time-varying error. The autoregressive model assumes that the error has a mean of 0 and a standard deviation of σ; and σ is set to a time-invariant constant. Through this model assumption, X t It is equal to a linear combination of multiple old X variables, and the variance is constant.

[0092] The advantages of the autoregressive model are obvious. It is just a linear combination of its own past states, does not require other information, and the resources used for prediction are simple. However, the state X analyzed by the autoregressive model must satisfy the autocorrelation coefficient, which is understood in the autoregressive model as X t-i The coefficient of The autocorrelation coefficient is close to 0. It measures the correlation between the same event in different periods. It is the correlation coefficient between a signal and itself at different time points. It can be used to find repeated patterns in the data. The calculation formula is:

[0093]

[0094] Among them, E is the expected function, X i is the historical state information at time i, μ i is the expected state information at time i, X i+k is the historical state information at time i+k, μ i+k is the expected state information at time i+k; the value range of R(k) is [-1,1]. If R(k) is 1, then state X is the maximum positive correlation; if R(k) is -1, then state X is the maximum negative correlation; if R(k) is 0, then state X is irrelevant.

[0095] In the autoregressive model, the autocorrelation coefficient of X is the key to its prediction accuracy. If the autocorrelation coefficient is less than 0.5, it should not be used, otherwise the prediction result will be very inaccurate. If the autocorrelation coefficient of X is less than -0.5, it means that X's future state will decrease due to the increase of its historical state. In the power system, it can be understood that some control measures have been taken or it is at the inflection point of cyclical change.

[0096] While autoregressive models can only predict state information for adjacent moments in the future, it's possible to predict state information for a series of future moments by treating the predicted state information as historical state information and then deleting the historical state information furthest from the current moment. However, this recursive approach assumes that the predicted historical state information is sufficiently accurate, and in real-world applications, the accuracy of this series of future moments will decrease.

[0097] The moving average model (MA) is a simple and effective smoothing prediction model. Its basic idea is to translate the state information arranged in time series item by item to the future prediction moment and perform weighted average summation. When the time series formed by the state information does not increase or decrease rapidly and there is no seasonal change, the moving average method can effectively eliminate the random fluctuations in the prediction caused by model errors. When the values ​​of the time series show an unstable development trend due to the interference of short-term changes and random fluctuations, the moving average method can often effectively eliminate the interference and make the change trend of the time series become a constant or linear function, thus solving the problem of unstable change trend of the time series.

[0098] Moving average methods are divided into simple moving average, weighted moving average, and exponential moving average. The simple moving average method is to give equal weight to each piece of historical state information. Although the calculation and programming are simple, the accuracy is low and the calculation effect is poor. Therefore, the weighted moving average method is generally used in applications. The weighted moving average method (WMA) is to give each different piece of historical state information involved in the average summation. In time series analysis, the model of the weighted moving average method is:

[0099]

[0100] Among them, WMA M That is, the future predicted state X in the autoregressive model t , p M That is the historical status information X i , it can be seen that the weighted moving average method believes that the closer the historical state information is to the prediction moment, the greater the impact on the prediction moment. However, when the overall trend of the data is not always linearly increasing or decreasing, the error of the model given by the moving average method is large. In addition, since the weighted average method can only estimate the future state by calculating the average value of the past state information, there is always an error that cannot be eliminated between it and the actual future state, because it does not reflect the changing trend of the state in the model parameters.

[0101] When predicting future states based on multi-time-scale state estimation results based on PMU hybrid measurements, it is also necessary to consider that the errors of state estimation results at different times are inconsistent, so this simple weighted average method cannot be fully adopted.

[0102] The exponential moving average (EMA) method can better reflect the time-varying characteristics of the situation in the sliding average model. The exponential moving average method is that the weight of each historical state information decreases exponentially over time. The decreasing speed is determined by a constant α. It is generally adopted N is the number of historical sampling information taken, and the EMA model can be expressed as follows:

[0103]

[0104] If the two-parameter weighted average method or the two-parameter exponential sliding average method is used, the changing trend of the time series can be reflected. However, due to the large amount of calculation, it is only suitable for analyzing data from two historical moments. This method is used in the prediction module of dynamic state estimation at multiple time scales.

[0105] (4) Autoregressive Moving Average Model

[0106] The goal of situational awareness prediction based on state estimation is to achieve online prediction of the safety situation within the next five minutes. The system state quantities required for safety index calculation include node voltage, branch current, and branch power. The time series changes of these quantities within five minutes are highly random and volatile. Therefore, there are problems such as the error variance is not a fixed value and the time series change trend is unstable. The AR model and the MA model can only solve one of these problems. Therefore, a time series analysis model that combines the AR model and the MA model is adopted, namely the Auto-Regression and Moving Average Model (ARMA).

[0107] The mathematical model of time series analysis autoregressive moving average (ARMA) is as follows:

[0108]

[0109] The autoregressive process AR is:

[0110]

[0111] Where ε is random white noise. If the lag operator L is used to represent the historical state, then Equation (11) can be described by the p-order polynomial of the lag operator:

[0112]

[0113] Where φ(L) is called the autoregressive operator. Since there are p historical state information, that is, p lags are performed, the process is called the p-order autoregressive process AR(p).

[0114] The sliding average process is:

[0115] x t =ε t -θ1ε t-1 -θ2ε t-2 -…-θ q ε t-q (12)

[0116] Similarly, this process is called the q-order moving average process, abbreviated as MA(q). q are the parameters of the moving average model. This autoregressive moving average process is ARMA(p,q).

[0117] Before establishing the ARMA model, the correlation of historical state quantities and the stability of time series changes must be calculated to determine whether they meet the stationary and correlation conditions. When the autocorrelation is determined, the orders p and q of the ARMA model are also determined.

[0118] The ARMA modeling steps are as follows:

[0119] (1) Determine whether the input data is a stationary non-pure random sequence. If it is stationary, proceed directly to (2); if it is not stationary, process the data and proceed to (2) only after processing.

[0120] (2) The established model is identified and ordered using autocorrelation and partial autocorrelation functions in combination with the AIC or BIC criteria.

[0121] (3) After completing model identification and order determination, enter the model parameter estimation stage.

[0122] (4) After completing parameter estimation, the fitted model is tested for adaptability. If the fitted model passes the test, the prediction phase begins. If the model fails the test, the model identification and test are repeated, i.e., step (2) is repeated to reselect the model.

[0123] (5) Use a highly adaptable fitting model to predict the future trend of the sequence.

[0124] Among them, the AIC criterion is a weighted function of fitting accuracy and number of parameters. The model that minimizes the AIC is considered the optimal model. The AIC criterion function is defined as follows:

[0125]

[0126]

[0127] In the formula, M(N) can generally be taken as or N is the total number of available historical state samples, and p0 is the optimal autoregressive model order.

[0128] The BIC criterion is similar to the AIC criterion and is defined as follows:

[0129]

[0130] Among them, n is the number of state variables to be taken, M(N) can generally be taken as or n′0 is the optimal order.

[0131] Generally speaking, the order of the AR process in the ARMA model is higher, and the order of the MA model is smaller. When analyzing ARMA, sometimes n'0 is close to 0, then the ARMA model is converted to an AR model because the algorithm of the ARMA model is much more complicated than that of the AR model.

[0132] In reality, the time series of voltage amplitudes and phase angles between nodes in a regional distribution network exhibit strong correlation. Therefore, using electrical state variables as time series yields multiple, highly correlated sequences. While a single sequence often fails to meet the correlation and self-stability requirements of the ARMA model, an analysis of the entire set of sequences reveals a high degree of correlation and cointegration—that is, a very close long-term equilibrium relationship exists between the sequences. In this context, the traditional ARMA time series analysis model needs to be improved and transformed into a multivariate time series analysis model, the statistical essence of which is a combination of multivariate regression analysis and time series analysis.

[0133] Since multivariate time series analysis mainly focuses on the correlation between time series of different variables, the multivariate ARMA model takes the multivariate AR process as the main process. The multivariate AR model is also called a vector autoregression model, that is, different variables form a vector, and the time series group of different variables becomes a time series of a vector.

[0134] Taking voltage exceeding the limit as the target safety situation for prediction as an example, the model of the multivariate AR model is as follows:

[0135] Assume that the dependent variable node voltage u and the independent variable x have the following relationship:

[0136] u=α0+α1x1+α2x2+…+α m x m +ε (17)

[0137] Where u refers to the voltage amplitude or phase angle at a node, and the independent variable x is the output and extrapolated result of the historical state estimation, including the historical information of u itself and the voltage and current information of adjacent nodes and branches. α is the independent variable coefficient of the model, which can be calculated using maximum correlation analysis.

[0138] Expanding the formula to all nodes, we obtain the multivariate AR prediction model for the future security situation of the regional distribution network:

[0139]

[0140] Let U=[u1,u2,…,u n ],X=[x1,x2,…,x m ], it can be expressed as

[0141] Y=AX+ε (19)

[0142] Because X contains historical information about each state estimation result, its dimension m should be a multiple of the dimension n of the voltage state output, resulting in large X and A. However, in practice, the voltage state, in addition to its high autocorrelation, only has a strong correlation with the voltage and current states of adjacent nodes and branches. Generally, only the voltage and current of adjacent nodes and branches at adjacent moments, along with a series of time series of the state itself, are required. This results in a more sparse matrix A and higher computational efficiency.

[0143] In addition, X can also be divided into X(1), X(2), ..., X(t-1) at different times, where X(k) refers to the state estimation at time k, and the matrix corresponding to X(k) is A(k). Although this division increases the number of vector groups, the calculation is simple and the computational efficiency is further improved.

[0144] The steps for calculating future voltage states using the multivariate ARMA / VAR time series analysis model are as follows:

[0145] (1) Read historical status information, including node voltage amplitude and phase angle;

[0146] (2) Select a certain time series to calculate the correlation between the voltage state and the historical state information in the time series, and select the historical state information with strong correlation as the independent variable X of the multivariate ARMA model;

[0147] (3) Construct the prediction matrix A (or matrix group A(k)) based on the covariance and least squares estimation;

[0148] (4) Based on the matrix A at the latest time t, predict the voltage of each node at the future time t+1.

[0149] Although the multivariate time series analysis model can predict the short-term trend of distribution network voltage and related situation awareness indicators, it does not pay attention to the physical characteristics of the distribution network state. When the time scale increases, the accuracy of the multivariate time series analysis model will be lower than the load forecasting model that reflects the physical laws of changes in the electrical quantity of the regional distribution network. This research project studied the ultra-short-term load forecasting technology of the regional distribution network with a five-minute cycle, and successfully realized the short-term load forecast of the regional distribution network with a time scale of 5 minutes. In conjunction with this situation forecasting technology, the multivariate time series analysis model established by using the real-time estimation results of state estimation and historical information is used to predict the regional distribution network node voltage in the next 10s, 20s, 30s, and 40s with a cycle of 10s. The predicted values ​​of the relevant situation awareness indicators are calculated with the help of the node voltage prediction values, so as to achieve the purpose of short-term prediction of the regional distribution network security situation awareness indicators.

[0150] This example uses a 33-node distribution network as the state estimation object to verify the effectiveness of the multivariate time series analysis prediction model. The load fluctuation is the same as the previous state estimation simulation verification. Each node has a load fluctuation of 30% to 50% with a constant power factor within 5 minutes. The first section of nodes is regarded as a motor with infinite capacity. The dynamic adjustment of the grid frequency is not considered. The distribution network is regarded as a steady-state environment. Only the multivariate time series prediction model is considered to be able to follow the changes in the voltage phasor.

[0151] The historical information obtained from state estimation is filtered with a 10-second cycle. This historical information includes voltage amplitude and voltage phase angle information, totaling a massive amount of data from 33 nodes at 29 moments. A VAR multivariate time series model is used for fitting. To reduce the model's overfitting, only the voltage phasors of adjacent nodes and the current amplitudes of connected branches are included in the multivariate time series modeling of a node's state. The first 25 historical moments are used as the sample set for training, and the last four historical moments are used as the control set. This assumes that the "present" moment is the 250th second, and the "future" voltage states of 260, 270, 280, and 290 seconds are predicted. To facilitate the identification of predicted values ​​and state estimation sample values, the window is scaled to 16 moments, from 140 to 290 seconds. The multivariate time series model for the first 12 moments is identical to the state estimation sample. Only the model for the last four moments, i.e., the "future period," exhibits model errors.

[0152] like Figure 5 、 Figure 6 and Figure 7 As shown in the figure, although there is a certain error in the voltage amplitude, it can accurately reflect the changing trend of the node voltage. The error in the prediction model is due to the following reasons:

[0153] (1) The error of the state estimation measurement is set as Gaussian white noise error, and the error is normally randomly distributed, which means that the multivariate time series analysis model cannot effectively predict the error. It can be seen from the figure that the maximum error of the prediction model is about 0.002, which is the same as the PMU measurement noise;

[0154] (2) Since the voltage amplitude prediction of a single node refers to the time series variation of the voltage phasor of the adjacent nodes,

[0155] Therefore, the prediction model shows certain overfitting characteristics.

[0156] Since the PMU-based state estimation provides a large amount of accurate phase angle historical information, the multivariate time series analysis model also achieves effective prediction of the phase angle.

[0157] like Figure 8 and Figure 9As shown in the figure, the multivariate time series prediction model can effectively predict the short-term changes in the amplitude and phase angle information of the voltage at each node in the regional distribution network within the range of allowable error. However, it has certain overfitting characteristics and is not suitable for long-term situation prediction. With the help of the obtained short-term situation prediction information of the node voltage, the distribution network operation center can first calculate the voltage over-limit index, and then calculate the distribution network node injection current and branch current through the admittance matrix, and calculate the node injection power and node flow power based on the current and voltage, and obtain the predicted values ​​of various distribution network security situation indicators.

[0158] This embodiment predicts the changing trend of the voltage at each node based on the safety situation prediction of state estimation, timely grasps the operating situation of the distribution network, obtains the state quantity of the future state, and then evaluates the operating status at future moments, timely discovers abnormal power supply and potential faults, thereby strengthening the management of the distribution network and improving the level of economic operation.

[0159] Example 2

[0160] The second embodiment of the present disclosure introduces a distribution network situation prediction system based on state assessment.

[0161] like Figure 10 A distribution network situation prediction method based on state assessment is shown, comprising:

[0162] An acquisition module configured to acquire real-time operation data of the distribution network;

[0163] A calculation module is configured to calculate a distribution network security situation indicator based on the acquired real-time distribution network operation data;

[0164] The prediction module is configured to use the autoregressive sliding average model to predict the node voltage at future times based on the historical operation data of the distribution network and the obtained distribution network security situation indicators, calculate the predicted value of the distribution network security situation indicators, and realize the short-term prediction of the distribution network security situation awareness indicators.

[0165] The detailed steps are the same as those of the distribution network status prediction method based on state assessment provided in Example 1, and will not be repeated here.

[0166] Example 3

[0167] A third embodiment of the present disclosure provides a computer-readable storage medium.

[0168] A computer-readable storage medium stores a program thereon, which, when executed by a processor, implements the steps in the distribution network situation prediction method based on state assessment as described in the first embodiment of the present disclosure.

[0169] The detailed steps are the same as those of the distribution network status prediction method based on state assessment provided in Example 1, and will not be repeated here.

[0170] Example 4

[0171] A fourth embodiment of the present disclosure provides an electronic device.

[0172] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of the distribution network status prediction method based on state assessment as described in the first embodiment of the present disclosure are implemented.

[0173] The detailed steps are the same as those of the distribution network status prediction method based on state assessment provided in Example 1, and will not be repeated here.

[0174] The foregoing description is merely a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. Those skilled in the art will readily appreciate that various modifications and variations are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present disclosure shall be included within the scope of protection of the present disclosure.

Claims

1. A distribution network situation prediction method based on state assessment, characterized in that: include: Obtain real-time operation data of the distribution network; Calculate the distribution network security situation indicators based on the acquired real-time operation data of the distribution network; Based on the historical operation data of the distribution network and the obtained distribution network security situation indicators, a multivariate autoregressive sliding average model is used to predict the node voltage at future moments, calculate the predicted value of the distribution network security situation indicators, and realize the short-term prediction of the distribution network security situation awareness indicators; The autoregressive moving average model adopts time series analysis, and the process of establishing the autoregressive moving average model is as follows: The input data is judged to determine whether it is a stationary non-pure random sequence. If it is stationary, it will directly proceed to the next step; otherwise, data processing is required before proceeding to the next step. Identify and determine the order of the stationary non-pure random sequence through autocorrelation and partial autocorrelation functions in combination with AIC criterion or BIC criterion, and estimate parameters; After completing the parameter estimation, the fit model is tested for adaptability. If the fit model passes the test, the prediction phase begins; If the model fails the test, the identification and test are repeated; Use highly adaptable fitting models to predict future trends of sequences.

2. A distribution network situation prediction method based on state assessment as claimed in claim 1, characterized in that: In the process of obtaining the real-time operation data of the distribution network, the state detection information of the distribution network is obtained through the synchronous vector measurement unit, the remote terminal unit and the feeder terminal unit respectively, and the real-time operation data of the distribution network is obtained.

3. A distribution network situation prediction method based on state assessment as claimed in claim 1, characterized in that: The specific process of calculating the distribution network security situation index is as follows: Perform multi-time-scale state estimation on the acquired real-time operation data of the distribution network to obtain the node voltage phasor information within the update period; Calculating the branch current and node injection current of the distribution network node according to the obtained node voltage phasor information; Calculate the complex power of the node injection power and the complex power of the branch power according to the obtained branch current and node injection current; The injected power at the node where the distribution transformer is located is regarded as the distribution transformer load, and the injected power at the node where the main transformer is located is regarded as the main transformer load. The main transformer overload rate, distribution transformer overload rate, line overload rate and line severe overload ratio are calculated in real time to complete the risk assessment of the distribution network safety situation.

4. A distribution network situation prediction method based on state assessment as described in claim 3, characterized in that: It is determined whether the node voltage is out of bounds based on the obtained node voltage phasor information, and the node voltage is judged to be out of limit.

5. A distribution network situation prediction method based on state assessment as claimed in claim 1, characterized in that: The specific process of predicting node voltage at future time is: Reading historical operation data of the distribution network, wherein the historical operation data of the distribution network includes node voltage amplitudes and phase angles; Select a certain time series to calculate the correlation between the voltage state and historical state information in the time series, and select historical state information with strong correlation as the independent variable of the autoregressive moving average model; Construct a prediction matrix based on covariance and least squares estimation; The voltage of each node at the future time t+1 is predicted based on the matrix at the latest time t.

6. A distribution network situation prediction method based on state assessment as claimed in claim 1, characterized in that: The specific process of calculating the predicted values ​​of distribution network security status indicators is as follows: based on the obtained node voltage short-term status prediction information, the voltage limit index is calculated, the distribution network node injection current and branch current are calculated through the admittance matrix, and the node injection power and node flow power are calculated based on the current and voltage, so as to obtain the predicted values ​​of various distribution network security status indicators.

7. A distribution network situation prediction system based on state assessment, characterized in that: include: An acquisition module configured to acquire real-time operation data of the distribution network; A calculation module is configured to calculate a distribution network security situation indicator based on the acquired real-time distribution network operation data; A prediction module is configured to use a multivariate autoregressive sliding average model to predict node voltages at future times based on historical distribution network operation data and the obtained distribution network security situation indicators, calculate the predicted values ​​of the distribution network security situation indicators, and achieve short-term prediction of the distribution network security situation awareness indicators; The autoregressive moving average model adopts time series analysis, and the process of establishing the autoregressive moving average model is as follows: The input data is judged to determine whether it is a stationary non-pure random sequence. If it is stationary, it will directly proceed to the next step; otherwise, data processing is required before proceeding to the next step. Identify and determine the order of the stationary non-pure random sequence through autocorrelation and partial autocorrelation functions in combination with AIC criterion or BIC criterion, and estimate parameters; After completing the parameter estimation, the fit model is tested for adaptability. If the fit model passes the test, the prediction phase begins; If the model fails the test, the identification and test are repeated; Use highly adaptable fitting models to predict future trends of sequences.

8. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps of the distribution network situation prediction method based on state assessment as described in any one of claims 1 to 6 are implemented.

9. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps in the distribution network situation prediction method based on state assessment according to any one of claims 1 to 6 are implemented.

Images