A power distribution network harmonic evaluation method and system based on a hidden Markov model

Abstract

The application provides a power distribution network harmonic evaluation method and system based on a hidden Markov model, which describes the harmonic emission characteristics of selected typical loads in a power distribution network at each time sequence range through a preset hidden Markov model, and estimates the harmonic currents emitted by all loads in the power distribution network at any time in the time sequence range, and then calculates the harmonic voltages of all loads in the power distribution network at any time in the time sequence range, so as to realize the harmonic evaluation of the power distribution network. Compared with the existing harmonic state estimation method, the method does not need to install a large number of synchronous harmonic monitoring devices, and has low application cost. Meanwhile, the method effectively overcomes the problem of low evaluation accuracy caused by the time-varying and fluctuating nature of the harmonic currents emitted by the loads in the existing harmonic flow estimation method, can accurately evaluate the voltage level of the harmonic of the power distribution network, and has strong applicability.

Description

Technical Field

[0001] This invention relates to the field of power quality assessment technology, and in particular to a method and system for assessing harmonics in distribution networks based on a hidden Markov model. Background Technology

[0002] With the continuous development of power electronics technology, the use of nonlinear loads and power electronic devices is increasing, leading to the widespread propagation of harmonics in distribution networks. This has become one of the most common power quality disturbances in power systems. Harmonic distortion problems in distribution networks can cause reduced power efficiency, power equipment failures, shortened equipment lifespan, increased temperature and power loss in cables and overhead lines, resulting in significant economic losses. Therefore, assessing the harmonic distortion level of distribution networks is crucial for power quality management and improvement.

[0003] The main purpose of distribution network harmonic assessment is to understand the harmonic voltage level of the distribution network, thereby providing a valid basis for assessing the harmonic distortion level, allocating harmonic responsibility, and mitigating harmonics. Current methods for distribution network harmonic assessment mainly include harmonic state estimation and harmonic power flow estimation. The former primarily constructs the harmonic transfer equation of the distribution network in the frequency domain, using harmonic data measured by installed harmonic monitoring devices as measurements and unmonitored harmonic states as state variables, and then solving them using methods such as least squares. This method requires a large number of harmonic monitoring devices and demands that the measurement data collected from the distribution network be synchronized, which significantly limits its application scope. Compared to the former, the latter, being unrestricted by measurement devices, has gained wider application. It mainly estimates the harmonic state of the distribution network by studying the harmonic currents injected by nonlinear loads at each node and calculating the harmonic power flow in a forward direction. However, the harmonic currents emitted by loads have a certain time-varying and fluctuating nature, making it difficult to guarantee the accuracy of this method, which is prone to significant assessment errors due to incorrect estimation of harmonic currents. Summary of the Invention

[0004] The present invention aims to provide a method and system for evaluating harmonics in distribution networks based on hidden Markov models, in order to solve the above-mentioned technical problems. It can accurately evaluate the voltage level of harmonics in distribution networks without the need for a large number of synchronous harmonic monitoring devices, and has strong applicability.

[0005] To address the aforementioned technical problems, this invention provides a method for evaluating harmonics in distribution networks based on a hidden Markov model, comprising the following steps:

[0006] Select typical loads in the distribution network and determine the harmonic state of these typical loads;

[0007] Determine the time series range for evaluation, and based on the preset hidden Markov model, obtain the harmonics emitted by the typical load harmonic state at any time within the time series range, and determine the emission characteristics of each harmonic of the typical load within the time series range.

[0008] Within the time series, the harmonic emission characteristics of each typical load are used to replace the harmonic emission characteristics of all loads in the distribution network, so as to obtain the harmonic currents emitted by all loads in the distribution network at any time within the time series.

[0009] Based on the harmonic currents emitted by all loads in the distribution network at any time within the time series, the harmonic voltages of all loads in the distribution network at any time within the time series are calculated, thereby realizing the harmonic assessment of the distribution network.

[0010] In the above scheme, a hidden Markov model is used to describe the harmonic emission characteristics of a typical load selected in the distribution network within a certain time series. Based on this, the harmonic currents emitted by all loads in the distribution network at any time within the time series are estimated, and then the harmonic voltages of all loads in the distribution network at any time within the time series are calculated, thus realizing the harmonic assessment of the distribution network. Compared with existing harmonic state estimation methods, this method does not require the installation of a large number of synchronous harmonic monitoring devices, resulting in low application costs. At the same time, it effectively overcomes the problem of low assessment accuracy caused by the time-varying and fluctuating nature of the harmonic currents emitted by loads in existing harmonic power flow estimation methods. It can accurately assess the voltage level of harmonics in the distribution network and has strong applicability.

[0011] In the above scheme, the harmonic emission characteristics of each harmonic of a typical load are flexibly used to replace the harmonic emission characteristics of all loads in the distribution network. This enables the description of the time-varying characteristics of all loads connected to all nodes in the distribution network. It overcomes the problem of difficulty in description, calculation and evaluation caused by the different harmonic current emission levels of the loads connected to each node in the distribution network under different operating conditions or modes, i.e., the different harmonic states of the loads.

[0012] Furthermore, the process of selecting typical loads and determining the harmonic state of typical loads in the distribution network specifically involves:

[0013] Typical loads are selected from all loads in the distribution network according to preset conditions. The harmonic status of the typical loads is determined by cluster analysis of historical monitoring data of each harmonic current of the typical loads.

[0014] In the above scheme, by selecting the most representative typical charge in the distribution network with the most harmonic emission characteristics, and collecting historical monitoring data of each harmonic current of the typical load in the distribution network using power quality monitoring devices, clustering algorithms such as k-means algorithm are used to perform cluster analysis on the harmonic current sample data. By determining the harmonic state of the typical load, the harmonic current emission level of the typical load under different harmonic states can be reflected.

[0015] Furthermore, the pre-setting process of the Hidden Markov Model is specifically as follows:

[0016] The harmonic currents of the constructed load at different times are selected as the state observation vectors of the hidden Markov model, and the initial state probability vector and state transition matrix of the harmonic state of the constructed load at the initial time are calculated.

[0017] The joint probability density of the state observation vector is described by a Gaussian mixture distribution, and the joint probability density parameters of the state observation vector are obtained, resulting in the mean matrix and covariance matrix.

[0018] An initial hidden Markov model is constructed based on the initial state probability vector, state transition matrix, mean matrix, and covariance matrix.

[0019] The initial hidden Markov model is trained by selecting a load harmonic data sample set as training data. After determining the state of each harmonic, the parameters of the initial hidden Markov model are solved to complete the preset of the hidden Markov model.

[0020] In the above scheme, a hidden Markov model of the load harmonic current emission level is established, and the model parameters are identified to analyze the transfer characteristics between load harmonic states.

[0021] Furthermore, the process of determining the time series range for evaluation, and obtaining the harmonics emitted by the typical load harmonic state at any time within the time series range based on a preset hidden Markov model, and determining the emission characteristics of each harmonic of the typical load within the time series range, specifically involves:

[0022] Determine the time series range to be estimated; based on the initial state probability vector in the Hidden Markov Model, randomly generate the harmonic state of the typical load at the initial time, obtain the joint probability density parameter of the state observation vector corresponding to the harmonic state at the initial time, and randomly sample and generate each harmonic emitted by the typical load at the initial time.

[0023] The Monte Carlo method is used to calculate the cumulative state transition probability matrix of a typical load to determine the harmonic state of the typical load at the next moment; the joint probability density parameters of the state observation vector corresponding to the harmonic state are obtained, and the harmonics emitted by the typical load at that moment are randomly generated.

[0024] Repeat the above operation until the harmonics emitted by the typical load harmonic state at any time within the time series are obtained, thereby determining the emission characteristics of each harmonic of the typical load within the time series.

[0025] Furthermore, the process of obtaining the harmonic currents emitted by all loads in the distribution network at any time within the time series by using the harmonic emission characteristics of typical loads to replace the harmonic emission characteristics of all loads in the distribution network, within the time series, specifically involves:

[0026] The harmonic emission characteristics of typical loads are used to replace the harmonic emission characteristics of all loads in the distribution network. The Monte Carlo method is used to calculate and obtain the harmonic currents emitted by all loads in the distribution network at any time within the time series.

[0027] In the above scheme, based on the established hidden Markov model, the Monte Carlo method is used to generate the time series of load-injected harmonic currents into the distribution network, and then harmonic power flow calculation is performed to analyze the harmonic voltage level of the distribution network. Without the need to install a large number of synchronous harmonic monitoring devices, the voltage level of the distribution network harmonics can be accurately assessed, and the scheme has strong applicability.

[0028] The above scheme enables accurate estimation of harmonic voltage levels at various nodes of the distribution network at different times, improving the accuracy and applicability of harmonic assessment of the distribution network.

[0029] This invention also provides a distribution network harmonic assessment system based on a hidden Markov model, comprising:

[0030] The typical load determination module is used to select typical loads in the distribution network and determine the harmonic state of the typical loads;

[0031] The time series range determination module is used to determine the time series range for evaluation;

[0032] Hidden Markov Model Preset Module: Used to preset Hidden Markov Models;

[0033] The typical harmonic emission characteristics determination module is used to obtain the harmonics emitted by the typical load harmonic state at any time within the time series based on the hidden Markov model, and to determine the emission characteristics of each harmonic of the typical load within the time series.

[0034] The load harmonic current acquisition module is used to replace the harmonic emission characteristics of all loads in the distribution network with the harmonic emission characteristics of typical loads within a time series range, and to acquire the harmonic currents emitted by all loads in the distribution network at any time within the time series range.

[0035] The load harmonic voltage calculation module is used to calculate the harmonic voltage of all loads in the distribution network at any time within the time series range, based on the harmonic current emitted by all loads in the distribution network at any time within the time series range.

[0036] Furthermore, in the typical load determination module, typical loads are selected from all loads in the distribution network according to preset conditions, and the harmonic state of the typical loads is determined by cluster analysis of historical monitoring data of each harmonic current of the typical loads.

[0037] Furthermore, in the Hidden Markov Model Preset Module, the preset process of the Hidden Markov Model is specifically as follows:

[0038] The harmonic currents of the constructed load at different times are selected as the state observation vectors of the hidden Markov model, and the initial state probability vector and state transition matrix of the harmonic state of the constructed load at the initial time are calculated.

[0039] The joint probability density of the state observation vector is described by a Gaussian mixture distribution, and the joint probability density parameters of the state observation vector are obtained, resulting in the mean matrix and covariance matrix.

[0040] An initial hidden Markov model is constructed based on the initial state probability vector, state transition matrix, mean matrix, and covariance matrix.

[0041] The initial hidden Markov model is trained by selecting a load harmonic data sample set as training data. After determining the state of each harmonic, the parameters of the initial hidden Markov model are solved to complete the preset of the hidden Markov model.

[0042] Furthermore, in the typical harmonic emission characteristic determination module, the time series range to be estimated is first determined; based on the initial state probability vector in the hidden Markov model, the harmonic state of the typical load at the initial time is randomly generated, the joint probability density parameter of the state observation vector corresponding to the harmonic state at the initial time is obtained, and each harmonic emitted by the typical load at the initial time is randomly sampled and generated.

[0043] The Monte Carlo method is used to calculate the cumulative state transition probability matrix of a typical load to determine the harmonic state of the typical load at the next moment; the joint probability density parameters of the state observation vector corresponding to the harmonic state are obtained, and the harmonics emitted by the typical load at that moment are randomly generated.

[0044] Repeat the above process until the harmonics emitted by the typical load harmonic state at any time within the time series are obtained, thereby determining the emission characteristics of each harmonic of the typical load within the time series.

[0045] Furthermore, in the load harmonic current acquisition module, the harmonic emission characteristics of each harmonic of a typical load are used to replace the harmonic emission characteristics of all loads in the distribution network. The Monte Carlo method is used to calculate and obtain the harmonic currents emitted by all loads in the distribution network at any time within the time series.

[0046] The present invention also provides a distribution network harmonic assessment device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the aforementioned distribution network harmonic assessment method based on a hidden Markov model.

[0047] The present invention also provides a distribution network harmonic assessment storage medium, the storage medium including a stored computer program, wherein, when the computer program is running, it controls the device where the storage medium is located to execute the aforementioned distribution network harmonic assessment method based on a hidden Markov model. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of a distribution network harmonic assessment method based on a hidden Markov model provided by an embodiment of the present invention.

[0049] Figure 2 This is a schematic diagram of a hidden Markov model structure provided in an embodiment of the present invention;

[0050] Figure 3 This is a comparison chart of the evaluation results of an embodiment of the present invention and the evaluation results of existing technical solutions;

[0051] Figure 4 This is a schematic diagram of the system structure connection of a distribution network harmonic assessment method based on a hidden Markov model provided in an embodiment of the present invention. Detailed Implementation

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] Please see Figure 1 This invention provides a method for evaluating harmonics in a distribution network based on a hidden Markov model, comprising the following steps:

[0054] S1: Select typical loads in the distribution network and determine the harmonic state of the typical loads;

[0055] S2: Determine the time series range for evaluation. Based on the preset hidden Markov model, obtain the harmonics emitted by the typical load harmonic state at any time within the time series range, and determine the emission characteristics of each harmonic of the typical load within the time series range.

[0056] S3: Within the time series range, the harmonic emission characteristics of each typical load are used to replace the harmonic emission characteristics of all loads in the distribution network, so as to obtain the harmonic currents emitted by all loads in the distribution network at any time within the time series range.

[0057] S4: Calculate the harmonic voltage of all loads in the distribution network at any time within the time series based on the harmonic current emitted by all loads in the distribution network at any time within the time series, and realize the harmonic assessment of the distribution network.

[0058] The typical load is a manually selected load determined based on experience.

[0059] In this embodiment of the invention, a pre-set Hidden Markov Model is used to describe the harmonic emission characteristics of a typical load selected in the distribution network within a certain time series. Based on this, the harmonic currents emitted by all loads in the distribution network at any time within the time series are estimated, and then the harmonic voltages of all loads in the distribution network at any time within the time series are calculated, thus realizing the harmonic assessment of the distribution network. Compared with existing harmonic state estimation methods, this method does not require the installation of a large number of synchronous harmonic monitoring devices, resulting in low application costs. At the same time, it effectively overcomes the problem of low assessment accuracy caused by the time-varying and fluctuating nature of the harmonic currents emitted by loads in existing harmonic power flow estimation methods. It can accurately assess the voltage level of harmonics in the distribution network and has strong applicability.

[0060] In this embodiment of the invention, the harmonic emission characteristics of each harmonic of a typical load are flexibly used to replace the harmonic emission characteristics of all loads in the distribution network, thereby enabling the description of the time-varying characteristics of all loads connected to all nodes in the distribution network. This overcomes the problem of difficulty in description, calculation and evaluation caused by the different harmonic current emission levels of the loads connected to each node in the distribution network under different operating conditions or modes, i.e., the different harmonic states of the loads.

[0061] Furthermore, the process of selecting typical loads and determining the harmonic state of typical loads in the distribution network specifically involves:

[0062] Typical loads are selected from all loads in the distribution network according to preset conditions. The harmonic status of the typical loads is determined by cluster analysis of historical monitoring data of each harmonic current of the typical loads.

[0063] In this embodiment of the invention, by selecting the most representative typical charge in the distribution network with the most harmonic emission characteristics, and collecting historical monitoring data of each harmonic current of the typical load in the distribution network using a power quality monitoring device, clustering algorithms such as k-means algorithm are used to perform cluster analysis on the harmonic current sample data. By determining the harmonic state of the typical load, the harmonic current emission level of the typical load under different harmonic states can be reflected.

[0064] Furthermore, the pre-setting process of the Hidden Markov Model is specifically as follows:

[0065] The harmonic currents of the constructed load at different times are selected as the state observation vectors of the hidden Markov model, and the initial state probability vector and state transition matrix of the harmonic state of the constructed load at the initial time are calculated.

[0066] The joint probability density of the state observation vector is described by a Gaussian mixture distribution, and the joint probability density parameters of the state observation vector are obtained, resulting in the mean matrix and covariance matrix.

[0067] An initial hidden Markov model is constructed based on the initial state probability vector, state transition matrix, mean matrix, and covariance matrix.

[0068] The initial hidden Markov model is trained by selecting a load harmonic data sample set as training data. After determining the state of each harmonic, the parameters of the initial hidden Markov model are solved to complete the preset of the hidden Markov model.

[0069] In this embodiment of the invention, a hidden Markov model of the load harmonic current emission level is established, and the model parameters are identified to analyze the transfer characteristics between load harmonic states.

[0070] Furthermore, the process of determining the time series range for evaluation, and obtaining the harmonics emitted by the typical load harmonic state at any time within the time series range based on a preset hidden Markov model, and determining the emission characteristics of each harmonic of the typical load within the time series range, specifically involves:

[0071] Determine the time series range to be estimated; based on the initial state probability vector in the Hidden Markov Model, randomly generate the harmonic state of the typical load at the initial time, obtain the joint probability density parameter of the state observation vector corresponding to the harmonic state at the initial time, and randomly sample and generate each harmonic emitted by the typical load at the initial time.

[0072] The Monte Carlo method is used to calculate the cumulative state transition probability matrix of a typical load to determine the harmonic state of the typical load at the next moment; the joint probability density parameters of the state observation vector corresponding to the harmonic state are obtained, and the harmonics emitted by the typical load at that moment are randomly generated.

[0073] Repeat the above operation until the harmonics emitted by the typical load harmonic state at any time within the time series are obtained, thereby determining the emission characteristics of each harmonic of the typical load within the time series.

[0074] Furthermore, the process of obtaining the harmonic currents emitted by all loads in the distribution network at any time within the time series by using the harmonic emission characteristics of typical loads to replace the harmonic emission characteristics of all loads in the distribution network, within the time series, specifically involves:

[0075] The harmonic emission characteristics of typical loads are used to replace the harmonic emission characteristics of all loads in the distribution network. The Monte Carlo method is used to calculate and obtain the harmonic currents emitted by all loads in the distribution network at any time within the time series.

[0076] In this embodiment of the invention, based on the established hidden Markov model, the Monte Carlo method is used to generate the time series of load-injected harmonic currents into the distribution network, and then harmonic power flow calculation is performed to analyze the harmonic voltage level of the distribution network. Without the need to install a large number of synchronous harmonic monitoring devices, the voltage level of the distribution network harmonics can be accurately assessed, and the applicability is strong.

[0077] The embodiments of the present invention enable accurate estimation of harmonic voltage levels at various nodes of the distribution network at different times, thereby improving the accuracy and applicability of harmonic assessment of the distribution network.

[0078] Furthermore, to further illustrate the implementation process of the distribution network harmonic assessment method based on the Hidden Markov Model, this embodiment provides a specific implementation process. The application of parameters and specific algorithms in this embodiment is only a supplementary explanation of this scheme and should not be interpreted as the only way to implement this scheme.

[0079] Furthermore, the process of selecting typical loads and determining their harmonic states in the distribution network can be achieved through the following steps:

[0080] A typical load with the most representative harmonic emission characteristics in the distribution network was selected, and historical monitoring data of each harmonic current of the typical load was collected using a power quality monitoring device, as shown in equations (1)-(3):

[0081]

[0082]

[0083] S = [S1, S2] (3)

[0084] In the formula: S is the historical monitoring dataset of each harmonic current, which is an N×2H matrix; N is the number of monitoring samples for each harmonic; H is the highest order of the monitored harmonic; I h,jThis is the monitoring sample of the j-th amplitude of the h-th harmonic current of a typical load; θ h,j This is the j-th phase angle monitoring sample of the h-th harmonic current of a typical load.

[0085] Next, the k-means algorithm is used to cluster the historical monitoring data of each harmonic current of a typical load. First, the number of load harmonic states K is set as needed, and K harmonic current samples are randomly selected as cluster centers. Then, the distance between the N sets of sample data in S and the cluster centers is calculated according to equation (4), and each set of sample data is assigned to the nearest cluster:

[0086]

[0087] In the formula: s j Let C be the j-th data sample in S; k s is the kth cluster center; j,,h C k,h s j and C k The h-th eigenvalue is the h-th harmonic eigenvalue.

[0088] After dividing the N sets of sample data into clusters, calculate the average value of each feature of all objects in the K clusters, use this average value as the new cluster center, and repeat the above steps until the cluster centers no longer change. Ultimately, K clusters can be determined, each cluster representing a harmonic state of the load; therefore, K harmonic states of a typical load can be determined.

[0089] Q = [Q1, Q2, ..., Q] k Q K (5)

[0090] In the formula: Q k Let Q represent the k-th harmonic state, and let K represent the set of harmonic states.

[0091] Furthermore, a hidden Markov model of the load harmonic current emission level in the distribution network is established. This model describes the probabilistic mapping relationship between the load harmonic current emission level and the load harmonic state. Specifically:

[0092] The harmonic currents of the load at different times are used as the state observation vectors of the model. The state observation vector at time t is shown in Equation (6).

[0093] O(t) = [p I (t), p θ (t)] T (6)

[0094] In the formula: p I (t)=[I2(t), I3(t),...,I h(t), ..., I H [(t)] represents the vector composed of the amplitudes of each harmonic current emitted by the load at time t, I h (t) represents the amplitude of the h-th harmonic emitted by the load at time t.

[0095] In the formula: p θ (t)=[θ2(t), θ3(t),...,θ h (t), ..., θ H [(t)] represents the vector consisting of the phase angles of the harmonic currents emitted by the load at time t, θ h (t) represents the phase angle of the h-th harmonic emitted by the load at time t.

[0096] Based on the selected model state observation vector, the initial state probability vector of the load harmonic state at the initial time t1 is shown in equations (7)-(8).

[0097] π = [π1, π2, ..., π] k , ..., π K (7)

[0098] π k =P(Q(t1)=Q k ), 1≤k≤K (8)

[0099] Where: π k This indicates that the load at the initial time t1 is in a harmonic state Q. k The probability; Q(t1) is the harmonic state of the load at time t1. The state transition matrix A is represented by equations (9)-(10):

[0100]

[0101] a rk =P(Q(t+1)=Q k |Q(t)=Q r ), 1≤r, k≤K (10)

[0102] In the formula: a rk Let Q be the element in row r and column k of A, representing the harmonic state of the load from state Q at time t. r The state transitions to Q at time t+1. k The probability of.

[0103] If a mixture Gaussian distribution is used to describe the joint probability density of the state observation vector, then the state Q... k The joint probability density of the corresponding state observation vector is given by equation (11).

[0104] b k [O(t)]=N(μ k, ∑ k ), 1≤k≤K (11)

[0105] In the formula: O(t) follows a mean of u k The covariance matrix is ​​∑ k The mixture of Gaussian distributions.

[0106] Therefore, the joint probability density parameters of each state observation vector can form the mean matrix and covariance matrix as shown in equation (12).

[0107]

[0108] Finally, a hidden Markov model can be constructed as shown in equation (13).

[0109] λ=(π,A,μ,∑) (13)

[0110] After constructing the Hidden Markov Model, this embodiment selects the load harmonic data sample set S obtained from monitoring as the training data, and determines the K harmonic states Q1-Q based on the clustering algorithm. K The Baum-Welch algorithm can be used to solve for the parameters of the Hidden Markov Model and complete the pre-set Hidden Markov Model.

[0111] Furthermore, the process of determining the time series range for evaluation, and obtaining the harmonics emitted by the typical load harmonic state at any time within the time series range based on a preset hidden Markov model, and determining the emission characteristics of each harmonic of the typical load within the time series range, specifically involves:

[0112] First, determine the time series range within which the harmonic state of the distribution network needs to be estimated:

[0113] T = {t1, t2, ..., t} j , ..., t end} (14)

[0114] In the formula: t j t represents the j-th time point in the time series. end This represents the last moment in the time series.

[0115] Based on the initial state probability vector π in the Hidden Markov Model, the harmonic state Q(t1) of the load at the initial time t1 is randomly generated. After determining Q(t1), the joint probability density parameters of the state observation vectors corresponding to the harmonic state Q(t1) are obtained, for example, when Q(t1) = Q k Then its corresponding parameter is u k , Σ k At this point, the state observation vector O(t1) of Q(t1) is [p I (t1),p θ (t1)]T N(u) follows a mixture Gaussian distribution k ,Σ k Based on the distribution of the observation vector, the observation vector at time t1 can be randomly sampled and generated, which is the amplitude and phase angle of each harmonic current of the load at time t1.

[0116] Based on the previous moment t j-1 The harmonic state Q(t) j-1 The Monte Carlo method can be used to generate the next time step t. j The harmonic state Q(t) j First, calculate the cumulative state transition probability matrix C. The elements in C are taken as follows:

[0117]

[0118] In the formula: c rk Let c be the element in the r-th row and k-th column of C. Then, generate a random number ε based on the uniform distribution U(0,1) and compare it with all elements in the j-th row of C. When the value of ε is in the j-th row and k-th element c of C... jk And the (k+1)th element c in row j of C jk+1 When in between, take Q(t) j )=Q k In determining Q(t) j After that, Q(t) can be obtained. j The joint probability density parameters of the state observation vector corresponding to ) are obtained, and t is generated by random sampling. j The observation vector O(t) at time t j )=[p I (t j ),p θ (t j )] T .

[0119] Repeating the above steps yields the harmonics emitted by the load at any time t within the time series range T, resulting in:

[0120]

[0121] After obtaining the harmonics emitted by a typical load at any time within the time series, the emission characteristics of each harmonic of the typical load within the time series can be determined.

[0122] Furthermore, within the time series range, the harmonic emission characteristics of typical loads are used to replace the harmonic emission characteristics of all loads in the distribution network, to obtain the harmonic currents emitted by all loads in the distribution network at any time t within the time series range, specifically expressed as follows:

[0123]

[0124] In the formula: I inject (t) represents the phasor matrix of each harmonic current injected into the distribution network at time t by the loads connected to all nodes of the distribution network. The h-th harmonic phasor injected at time t for the load connected to the j-th node is shown in Equation (18); M is the number of nodes in the distribution network.

[0125]

[0126] In the formula: The amplitude of the h-th harmonic injected into the load connected to the j-th node at time t; The phase angle of the h-th harmonic injected into the load connected to the j-th node at time t.

[0127] Furthermore, based on the harmonic currents emitted by all loads in the distribution network at any given time within the time series, the harmonic voltages of all loads in the distribution network at any given time within the time series can be calculated using the following formula:

[0128]

[0129] In the formula: Let h be the h-th harmonic voltage of the j-th node at time t; The h-th harmonic impedance coefficient between node i and node j in the distribution network.

[0130] This embodiment comprises four parts: harmonic data clustering, constructing a Hidden Markov Model (HMM) of load harmonic current emission levels, generating a time series of harmonic currents injected into the distribution network, and harmonic power flow calculation. After collecting historical monitoring data of each harmonic current from typical loads in the distribution network, cluster analysis is performed to determine the load harmonic state. Then, an HMM of the load harmonic current emission levels is established, and the HMM parameters are solved. These parameters include the initial state occurrence probability, the state transition probability matrix, and the joint probability distribution of the observation vectors. Based on the HMM, Monte Carlo simulation is used to generate a time series of harmonic currents injected into the distribution network. Finally, harmonic power flow calculation is performed based on the generated harmonic current series to estimate the harmonic voltage levels at each node of the distribution network. This method can accurately assess the voltage levels of harmonics in the distribution network without requiring the installation of numerous synchronous harmonic monitoring devices, demonstrating strong applicability.

[0131] Please see Figure 2 Furthermore, in the hidden Markov model of the load harmonic current emission level, the generation of the initial state variable Q(t1) satisfies the initial state occurrence probability π, and the state variable Q(t1) - Q(t) jThe transition probabilities between +1) satisfy the state transition probability matrix A, and the state variables Q(t1)-Q(t) are given by Q(t1)-Q(t). j+11 ) to the observed variable O(t1)-O(t) j+1 The mapping of ) satisfies the emission probability matrix B.

[0132] Furthermore, to fully illustrate the technical effects of this embodiment, the evaluated value of the proposed solution is compared with the actual value and the evaluation results of existing technical solutions. This is achieved by comparing the third harmonic voltage values; for details, please refer to [link to relevant documentation]. Figure 3 .

[0133] Clearly, the proposed solution in this embodiment yields a more accurate evaluation result than the calculation results of existing technical solutions, thus making the proposed power distribution network harmonic evaluation method more accurate and more valuable for engineering applications.

[0134] Please see Figure 4 This embodiment provides a distribution network harmonic assessment system based on a Hidden Markov Model (HMM), used to implement a distribution network harmonic assessment method based on an HMM, specifically including:

[0135] The typical load determination module is used to select typical loads in the distribution network and determine the harmonic state of the typical loads;

[0136] The time series range determination module is used to determine the time series range for evaluation;

[0137] Hidden Markov Model Preset Module: Used to preset Hidden Markov Models;

[0138] The typical harmonic emission characteristics determination module is used to obtain the harmonics emitted by the typical load harmonic state at any time within the time series based on the hidden Markov model, and to determine the emission characteristics of each harmonic of the typical load within the time series.

[0139] The load harmonic current acquisition module is used to replace the harmonic emission characteristics of all loads in the distribution network with the harmonic emission characteristics of typical loads within a time series range, and to acquire the harmonic currents emitted by all loads in the distribution network at any time within the time series range.

[0140] The load harmonic voltage calculation module is used to calculate the harmonic voltage of all loads in the distribution network at any time within the time series range, based on the harmonic current emitted by all loads in the distribution network at any time within the time series range.

[0141] Furthermore, in the typical load determination module, typical loads are selected from all loads in the distribution network according to preset conditions, and the harmonic state of the typical loads is determined by cluster analysis of historical monitoring data of each harmonic current of the typical loads.

[0142] Furthermore, in the Hidden Markov Model Preset Module, the preset process of the Hidden Markov Model is specifically as follows:

[0143] The harmonic currents of the constructed load at different times are selected as the state observation vectors of the hidden Markov model, and the initial state probability vector and state transition matrix of the harmonic state of the constructed load at the initial time are calculated.

[0144] The joint probability density of the state observation vector is described by a Gaussian mixture distribution, and the joint probability density parameters of the state observation vector are obtained, resulting in the mean matrix and covariance matrix.

[0145] An initial hidden Markov model is constructed based on the initial state probability vector, state transition matrix, mean matrix, and covariance matrix.

[0146] The initial hidden Markov model is trained by selecting a load harmonic data sample set as training data. After determining the state of each harmonic, the parameters of the initial hidden Markov model are solved to complete the preset of the hidden Markov model.

[0147] Furthermore, in the typical harmonic emission characteristic determination module, the time series range to be estimated is first determined; based on the initial state probability vector in the hidden Markov model, the harmonic state of the typical load at the initial time is randomly generated, the joint probability density parameter of the state observation vector corresponding to the harmonic state at the initial time is obtained, and each harmonic emitted by the typical load at the initial time is randomly sampled and generated.

[0148] The Monte Carlo method is used to calculate the cumulative state transition probability matrix of a typical load to determine the harmonic state of the typical load at the next moment; the joint probability density parameters of the state observation vector corresponding to the harmonic state are obtained, and the harmonics emitted by the typical load at that moment are randomly generated.

[0149] Repeat the above process until the harmonics emitted by the typical load harmonic state at any time within the time series are obtained, thereby determining the emission characteristics of each harmonic of the typical load within the time series.

[0150] Furthermore, in the load harmonic current acquisition module, the harmonic emission characteristics of each harmonic of a typical load are used to replace the harmonic emission characteristics of all loads in the distribution network. The Monte Carlo method is used to calculate and obtain the harmonic currents emitted by all loads in the distribution network at any time within the time series.

[0151] This embodiment also provides a distribution network harmonic assessment device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the aforementioned distribution network harmonic assessment method based on a hidden Markov model.

[0152] This embodiment also provides a distribution network harmonic assessment storage medium, the storage medium including a stored computer program, wherein, when the computer program is running, it controls the device where the storage medium is located to execute the aforementioned distribution network harmonic assessment method based on a hidden Markov model.

[0153] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for assessing harmonics in a distribution network based on a hidden Markov model, characterized in that, Includes the following steps: Select typical loads in the distribution network and determine the harmonic state of these typical loads; The time series range for evaluation is determined. Based on a pre-defined Hidden Markov Model (HMM), the harmonics emitted by a typical load at any time within the time series range are obtained, and the emission characteristics of each harmonic of the typical load within the time series range are determined. Specifically, the pre-defined process of the HMM involves: selecting the harmonic currents of the constructed load at different times as the state observation vectors of the HMM; calculating the initial state probability vector and state transition matrix of the constructed load harmonic state at the initial time; describing the joint probability density of the state observation vectors using a Gaussian mixture distribution; obtaining the joint probability density parameters of the state observation vectors; and obtaining the mean matrix and covariance matrix; constructing an initial HMM based on the initial state probability vector, state transition matrix, mean matrix, and covariance matrix; training the initial HMM using a load harmonic data sample set as training data; determining the parameters of the initial HMM after determining each harmonic state; and completing the pre-defined HMM. The process of determining the emission characteristics of each harmonic of a typical load within a time series, based on a pre-defined Hidden Markov Model (HMM), involves: determining the time series range to be estimated; randomly generating the harmonic state of the typical load at the initial moment based on the initial state probability vector in the HMM; obtaining the joint probability density parameter of the state observation vector corresponding to the initial harmonic state; randomly sampling and generating each harmonic emitted by the typical load at the initial moment; calculating the cumulative state transition probability matrix of the typical load using the Monte Carlo method to determine the harmonic state of the typical load at the next moment; obtaining the joint probability density parameter of the state observation vector corresponding to the harmonic state; randomly generating each harmonic emitted by the typical load at that moment; repeating the above operations until the harmonics emitted by the typical load at any moment within the time series are obtained, thereby determining the emission characteristics of each harmonic of the typical load within the time series. Within the time series, the harmonic emission characteristics of each typical load are used to replace the harmonic emission characteristics of all loads in the distribution network, so as to obtain the harmonic currents emitted by all loads in the distribution network at any time within the time series. Calculate the harmonic voltage of all loads in the distribution network at any time within the time series based on the harmonic current emitted by all loads in the distribution network at any time within the time series. The typical load is a manually selected load determined based on experience.

2. The method for assessing harmonics in a distribution network based on a hidden Markov model according to claim 1, characterized in that, The process of selecting typical loads and determining their harmonic states in the distribution network specifically involves: Typical loads are selected from all loads in the distribution network according to preset conditions. The harmonic status of the typical loads is determined by cluster analysis of historical monitoring data of each harmonic current of the typical loads.

3. A method for assessing harmonics in a distribution network based on a hidden Markov model according to any one of claims 1 to 2, characterized in that, The process of obtaining the harmonic currents emitted by all loads in the distribution network at any time within the time series by using the harmonic emission characteristics of typical loads to replace the harmonic emission characteristics of all loads in the distribution network, within the time series, is as follows: The harmonic emission characteristics of typical loads are used to replace the harmonic emission characteristics of all loads in the distribution network. The Monte Carlo method is used to calculate and obtain the harmonic currents emitted by all loads in the distribution network at any time within the time series.

4. A distribution network harmonic assessment system based on a hidden Markov model, characterized in that, A method for assessing harmonics in a distribution network based on a hidden Markov model, applicable to any one of claims 1 to 3, includes: The typical load determination module is used to select typical loads in the distribution network and determine the harmonic state of the typical loads; The time series range determination module is used to determine the time series range for evaluation; A Hidden Markov Model (HMM) preset module is used to preset the Hidden Markov Model. The preset process of the HMM in this module is as follows: The harmonic currents of the constructed load at different times are selected as the state observation vectors of the HMM; the initial state probability vector and state transition matrix of the constructed load harmonic state at the initial time are calculated; a Gaussian mixture distribution is used to describe the joint probability density of the state observation vectors, and the joint probability density parameters of the state observation vectors are obtained, resulting in the mean matrix and covariance matrix; an initial HMM is constructed based on the initial state probability vector, state transition matrix, mean matrix, and covariance matrix; a load harmonic data sample set is selected as training data to train the initial HMM; after determining each harmonic state, the parameters of the initial hidden Markov model are solved, thus completing the preset of the HMM. A typical harmonic emission characteristic determination module is used to obtain the harmonics emitted by a typical load at any time within the time series based on a hidden Markov model, and to determine the emission characteristics of each harmonic of the typical load within the time series. In this module, the time series range to be estimated is first determined; based on the initial state probability vector in the hidden Markov model, the harmonic state of the typical load at the initial time is randomly generated; the joint probability density parameter of the state observation vector corresponding to the harmonic state at the initial time is obtained; and each harmonic emitted by the typical load at the initial time is randomly sampled and generated; the cumulative state transition probability matrix of the typical load is calculated using the Monte Carlo method to determine the harmonic state of the typical load at the next time step; the joint probability density parameter of the state observation vector corresponding to this harmonic state is obtained; and each harmonic emitted by the typical load at this time is randomly generated; the above process is repeated until each harmonic emitted by the typical load at any time within the time series is obtained, thereby determining the emission characteristics of each harmonic of the typical load within the time series. The load harmonic current acquisition module is used to replace the harmonic emission characteristics of all loads in the distribution network with the harmonic emission characteristics of typical loads within a time series range, and to acquire the harmonic currents emitted by all loads in the distribution network at any time within the time series range. The load harmonic voltage calculation module is used to calculate the harmonic voltage of all loads in the distribution network at any time within the time series range, based on the harmonic current emitted by all loads in the distribution network at any time within the time series range.

5. A distribution network harmonic assessment system based on a hidden Markov model according to claim 4, characterized in that, In the typical load determination module, typical loads are selected from all loads in the distribution network according to preset conditions, and the harmonic state of the typical loads is determined by cluster analysis of historical monitoring data of each harmonic current of the typical loads.

6. A distribution network harmonic assessment system based on a hidden Markov model according to any one of claims 4 to 5, characterized in that, In the load harmonic current acquisition module, the harmonic emission characteristics of each typical load are used to replace the harmonic emission characteristics of all loads in the distribution network. The Monte Carlo method is used to calculate and obtain the harmonic currents emitted by all loads in the distribution network at any time within the time series.

Images