CN116663889A - Novel power system risk assessment method based on improved Gaussian model - Google Patents
Novel power system risk assessment method based on improved Gaussian model Download PDFInfo
- Publication number
- CN116663889A CN116663889A CN202310501193.8A CN202310501193A CN116663889A CN 116663889 A CN116663889 A CN 116663889A CN 202310501193 A CN202310501193 A CN 202310501193A CN 116663889 A CN116663889 A CN 116663889A
- Authority
- CN
- China
- Prior art keywords
- risk
- model
- gaussian model
- power system
- risk assessment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000012502 risk assessment Methods 0.000 title claims abstract description 43
- 238000013210 evaluation model Methods 0.000 claims abstract description 7
- 239000002245 particle Substances 0.000 claims description 30
- 230000006698 induction Effects 0.000 claims description 11
- 238000012360 testing method Methods 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 8
- 230000005284 excitation Effects 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 230000003247 decreasing effect Effects 0.000 claims description 3
- 238000012821 model calculation Methods 0.000 claims description 2
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 241000379147 Strawberry green petal phytoplasma Species 0.000 description 17
- 238000004364 calculation method Methods 0.000 description 8
- 238000005070 sampling Methods 0.000 description 7
- 238000013459 approach Methods 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 4
- 239000002131 composite material Substances 0.000 description 4
- 230000033228 biological regulation Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 3
- 238000005315 distribution function Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 238000007476 Maximum Likelihood Methods 0.000 description 2
- 238000000342 Monte Carlo simulation Methods 0.000 description 2
- ZZUFCTLCJUWOSV-UHFFFAOYSA-N furosemide Chemical compound C1=C(Cl)C(S(=O)(=O)N)=CC(C(O)=O)=C1NCC1=CC=CO1 ZZUFCTLCJUWOSV-UHFFFAOYSA-N 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000004931 aggregating effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000029087 digestion Effects 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- Educational Administration (AREA)
- General Business, Economics & Management (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Operations Research (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Game Theory and Decision Science (AREA)
- General Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Public Health (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Computational Linguistics (AREA)
- Probability & Statistics with Applications (AREA)
- Algebra (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
Abstract
The invention discloses a novel power system risk assessment method based on an improved Gaussian model, which comprises the following steps of: s1: acquiring a power sample; s2: establishing an evaluation model and setting risk indexes; s3: constructing a comprehensive index based on the risk index, and calculating a comprehensive index score; s4: and evaluating the risk level based on the comprehensive index score, and executing the solution based on the risk level. The beneficial effects of the invention are as follows: the probability distribution of uncertain input is not needed, the time of risk assessment of the power system can be reduced, and the real-time performance of risk assessment is improved.
Description
Technical Field
The invention belongs to the field of power system risk assessment, and particularly relates to a novel power system risk assessment method based on an improved Gaussian model.
Background
With the access of high-proportion new energy, the inertia of the novel power system is continuously reduced, the dynamic characteristics are changed, and the safety and stability operation risks of the power grid are increasingly outstanding. The risk assessment can quantify the risk of the power system, and is a reliable guarantee for safety analysis of the power system. With the construction of new power systems, conventional risk assessment has some problems in facing new energy power systems. The large-scale access proportion of new energy is fast and fast improved, the on-site digestion degree of the power grid is insufficient, the peak regulation capacity of the system is relatively large, the frequency regulation capacity of the power system is reduced, and the power grid has the risk of frequency out-of-limit and even stability damage; the reactive voltage regulation capability of new energy sources such as wind power, photovoltaic and the like is far lower than that of a conventional thermal power generating unit, so that the system faces the risk of voltage instability; the high-proportion new energy is connected and overlapped with the cross-region AC/DC interconnection of the power grid and the distributed micro-power grid, the local transient energy impact characteristic under the system disturbance event is more complex, and global stability risks are easy to cause.
In the prior art, monte Carlo (MC) based sampling methods and variants thereof are basic probability analysis tools, but they require an accurate Probability Distribution Function (PDF) of uncertain input, and are slow and costly to calculate. Other analysis methods, such as the integration of PDF by simplification and linearization, produce larger errors in the presence of larger uncertainties; the PCE approximates the model with an orthogonal polynomial, but requires knowledge of the uncertain input distribution and correlation. The gaussian process is data driven, but for high dimensional inputs it is computationally slow and costly. There is a problem in that the time for risk assessment of the power system cannot be reduced.
For example, a "method for analyzing the adequacy of a power system based on the monte carlo simulation method" disclosed in the chinese patent literature, its bulletin number: CN104156770a, filing date: the invention comprises the following steps of: collecting component data to form system information; sampling the element state duration to form a system state sequence; evaluating the states of all the systems in the power system; and obtaining the severity of the power system fault. The method can calculate the component failure rate of the system under the condition of disaster weather, and calculate the evaluation result of the power grid by adopting a Monte Carlo simulation method based on component state duration sampling, but has the problem that the time of risk evaluation of the power system cannot be reduced because an accurate probability distribution function of uncertain input is required.
Disclosure of Invention
Aiming at the defects that the time of risk assessment of a power system cannot be reduced and the real-time performance of the risk assessment cannot be improved due to the fact that an uncertain input accurate probability distribution function is needed in the prior art, the invention provides a novel power system risk assessment method based on an improved Gaussian model, probability distribution of uncertain input is not needed, the time of risk assessment of the power system can be reduced, and the real-time performance of the risk assessment is improved.
The technical scheme of the invention is as follows, a novel power system risk assessment method based on an improved Gaussian model comprises the following steps:
s1: acquiring a power sample;
s2: establishing an evaluation model and setting risk indexes;
s3: constructing a comprehensive index based on the risk index, and calculating a comprehensive index score;
s4: and evaluating the risk level based on the comprehensive index score, and executing the solution based on the risk level.
In the scheme, a power sample is acquired and used as original data of risk assessment of a power system, an assessment model is established, a sparse Gaussian model is established by using random variation reasoning and a probability framework is embedded, uncertain input probability distribution is not needed, an SGP optimal parameter is obtained based on an improved particle swarm algorithm, accordingly, the time of risk assessment of the power system is reduced, risk indexes are set and used for assessing risks, comprehensive indexes are established based on the risk indexes, comprehensive index scores are calculated, risk grades are evaluated based on the comprehensive index scores, solutions are executed based on the risk grades, and accordingly corresponding solutions are selected according to the numerical risks.
Preferably, in S1, the power sample includes power data and device data, and the power data includes voltage and power.
Preferably, in the step S2 of establishing the evaluation model, a sparse Gaussian model is established by using random variation reasoning and a probability frame is embedded, and the method comprises the following steps of:
s211: establishing a Gaussian model;
s212: establishing a sparse Gaussian model based on the Gaussian model;
s213: processing the mean and variance of the sparse Gaussian model;
s214: and calculating the KL divergence between the variational posterior and the true posterior.
In this scheme, gaussian model parameters are learned by maximum likelihood estimation, with posterior being derived from a priori and bayesian rule-based likelihood methods. As a non-parametric approach, the Gaussian model implements regression data driven. Performance is generally better than sample-based methods, with fewer samples being required for an uncertain input. The sparse Gaussian model improves expandability and reduces the risk assessment time of the power system.
Preferably, in the step S2 of establishing the evaluation model, the optimal parameters of the sparse Gaussian model are obtained based on an improved particle swarm algorithm, and the method comprises the following steps:
s221: initializing particle positions and speeds based on the to-be-determined particle swarm;
s222: updating the position and the speed of the particles during iterative circulation;
s223: obtaining a dynamic inertia factor based on the linearly decreasing weight;
s224: updating and recording the current optimal value and the global optimal value of the particle;
s225: if the difference between the upper limit of the cycle times or algebra meets the error constraint condition, outputting a result; otherwise, step S222 is performed.
In this scheme, to enhance the robustness of the limited samples, an improved particle swarm algorithm is used for parameter estimation. Because of the significantly reduced computational complexity, the sparse gaussian model enables voltage risk assessment for large power systems with high dimensional uncertainty inputs.
Preferably, in S2, the risk index includes a voltage out-of-limit risk index, a voltage deviation index, and a power flow section out-of-limit risk index.
In the scheme, a voltage out-of-limit risk value is obtained through a voltage out-of-limit risk index; obtaining a voltage deviation value through a voltage deviation index; obtaining a power flow section out-of-limit risk value through the power flow section out-of-limit risk index; thereby performing risk assessment on the power system.
Preferably, S3 comprises the steps of:
s31: constructing a judgment matrix based on the risk index;
s32: calculating the weight of the reference layer to the target layer and the weight of the index layer to the reference layer;
s33: calculating a combination weight and a comprehensive weight;
s34: and calculating the comprehensive index score by a analytic hierarchy process.
In the scheme, the weight of a reference layer to a target layer and the weight of an index layer to the reference layer are calculated through a judgment matrix, and consistency test is carried out; combining the weight coefficients to obtain the combined weight of the influence of the index layer element on the target layer, wherein the calculation method is that each element weight of the index layer is multiplied by the corresponding reference layer weight respectively; finally, based on the combined weight data, obtaining comprehensive weights; calculating a combination weight and a comprehensive weight; and calculating the comprehensive index score by a analytic hierarchy process. And the risk assessment of the power system is quantified, so that the risk situation can be intuitively known.
Preferably, S4 comprises the steps of:
s41: setting a risk comparison table containing a comprehensive index score range, a risk grade and a solution;
s42: obtaining a comprehensive index score;
s43: matching a comprehensive index score range to which the comprehensive index score belongs;
s44: the risk level is matched based on the comprehensive index score range, and the solution is matched based on the risk level.
In the scheme, in the risk comparison table, different comprehensive index score ranges correspond to different risk levels, and the risk levels correspond to the same or different solutions; and judging the comprehensive index score range corresponding to the comprehensive index score, thereby obtaining the risk grade corresponding to the comprehensive index score and the solution corresponding to the risk grade.
Preferably, a test system for model calculation is provided, comprising: determining a general dynamic model; distributing inertia of the test system and matching planning cases; and adjusting the model parameters of the speed regulator.
Preferably, the sparse Gaussian model is built by introducing induction points in the random variation reasoning process, so that the complexity of the model is reduced.
In the scheme, a set of induction points z= { Z1,..once, zl }, and a function value u=f (Z) are introduced in the process of establishing a used sparse Gaussian model by random variation reasoning. By using an induction point, the model complexity is reduced to O (nl 2), where the dimension of the induction point is l, which is much smaller than n.
Preferably, the general dynamic model includes a generator model, a turbine governor model and an excitation system model, and the general dynamic model is a low-order dynamic model.
In the scheme, low-order dynamic models such as a generator model, a turbine speed regulator model and an excitation system model are adopted, and the complexity of the model is reduced on the premise of meeting risk assessment, so that the time of risk assessment of a power system is reduced.
The beneficial effects of the invention are as follows: the probability distribution of uncertain input is not needed, the time of risk assessment of the power system can be reduced, and the real-time performance of risk assessment is improved.
Drawings
Fig. 1 is a flow chart of the present invention.
Fig. 2 is a comparison of the different methods in estimating the voltage amplitude.
Fig. 3 is a comparison of SGP and a sample-based method in estimating voltage amplitude.
Detailed Description
The technical scheme of the invention is further specifically described below through examples and with reference to the accompanying drawings.
Embodiment one:
as shown in fig. 1, a novel power system risk assessment method based on an improved gaussian model includes the following steps:
s1: acquiring a power sample;
s2: establishing an evaluation model and setting risk indexes;
s3: constructing a comprehensive index based on the risk index, and calculating a comprehensive index score;
s4: and evaluating the risk level based on the comprehensive index score, and executing the solution based on the risk level.
Acquiring a power sample as original data of power system risk assessment, establishing an assessment model, establishing a sparse Gaussian model by using random variation reasoning, embedding a probability framework, acquiring optimal parameters of SGP (generalized mean value) without uncertain input probability distribution, obtaining optimal parameters based on an improved particle swarm algorithm, thereby reducing time of power system risk assessment, setting a risk index for assessing risk, constructing a comprehensive index based on the risk index, calculating a comprehensive index score, digitizing the assessed risk, assessing a risk level based on the comprehensive index score, and executing a solution based on the risk level, thereby realizing selection of a corresponding solution according to the digitized risk.
In step S1, a power sample is obtained, where the power sample includes power data such as voltage and power, and may further include device data such as a type and a model of the power device. The power samples are acquired by the sensor acquisition power equipment or directly reference the acquired and edited power data, such as importing the power data through EXCEL and receiving the power data provided by a third party power data platform.
In step S2, an evaluation model is built. Establishing a sparse Gaussian model by using random variation reasoning, embedding a probability frame, and obtaining optimal parameters of SGP based on an improved particle swarm algorithm; setting a risk index.
A sparse Gaussian model is established by using random variation reasoning and a probability framework is embedded, and the method comprises the following steps:
s211: and establishing a Gaussian model.
The gaussian model (GP) describes the output as y=f (X) +e, where the inputAt the same time system noise->The mapping function f is defined by a mean function and a covariance function:
in the above formula, f (·) is a nonlinear mapping function; m (x) is a mean (average) function, typically expressed in the form of a polynomial function; k (·) is a covariance function, setting the covariance between the points between x and x', which allows us to capture the correlation, linear or nonlinear dependence between inputs; x, x 'is a power sample, x' represents a power sample excluding x, and θ represents GP model parameters.
GP model parameters are learned by maximum likelihood estimation, with posterior being derived from a priori and bayesian rule-based likelihood methods. As a non-parametric approach, GP implements regression data driven. Performance is generally better than sample-based methods, with fewer samples being required for an uncertain input.
S212: the extended gaussian model is a sparse gaussian model.
And establishing a sparse Gaussian model by expanding the Gaussian model. A set of induction points z=z is introduced 1 ,…,z l The function value u=f (Z). Assuming that the real objective function f and the pseudo objective function f are conditionally independent given the induction point z=u, the joint distribution p (·) can be approximated by an inferred probability distribution q (·):
in the above formula, Z is an induction point, u is a function value, f is a real objective function, f is a pseudo objective function, q (·) is a probability distribution, and p (·) is a joint distribution.
By using induction points, model complexity is reduced to O (nl 2 ) Where the number of induction points is l, which is much smaller than n. This explains that in theory, the sparse gaussian model (Sparse Gaussian Process, SGP) is much more scalable for larger scale power systems than GP.
S213: and processing the mean and variance of the SGP model.
And processing the mean and variance of the SGP model. Formally, the probability distribution of the variational posterior approximates q (f, u) =p (f|u) q (u), which becomes the following equation after marginalizing u:
q(f|0,S)＝∫p(f|u)q(u)du
in the above formula, q (·) is probability distribution, p (·) is joint distribution, u is a function value, and S is a variation parameter.
Mean mu * Sum of variancesObtained by the formula:
in the above, mu * Is the mean value of the two values,for variance->K XZ And K ZZ Representing a cross covariance matrix between the induced points and the training points and an auto covariance matrix of the induced points, m Z For the mean value of the induction point Z, u is the function value, S, m q To change the parameters, P T For K is the auto-covariance matrix of the training points.
S214: and calculating the KL divergence between the variational posterior and the true posterior.
Minimizing the Kullback-Leibler (KL) divergence between the variational posterior q and the true posterior p is equivalent to maximizing the lower limit of the true log marginal likelihood as shown below:
in the above formula, E (·) is the desired operator, f i ,m q S is a variation parameter which is maximized to obtain an estimate, x i For the samples, KL is the KL divergence calculation, N is the normalization function, q (·) is the probability distribution, and p (·) is the joint distribution.
To enhance the robustness of the limited samples, improved particle swarm algorithms are used for parameter estimation. Due to the significantly reduced computational complexity, the proposed SGP enables us to perform voltage risk assessment for large power systems with high-dimensional uncertainty inputs.
Obtaining optimal parameters of SGP based on an improved particle swarm algorithm, comprising the following steps:
s221: and forming a group of particles to be determined, and finishing initializing the position and the speed of the particles.
Initializing the particle position and velocity is accomplished by forming a population of particles x to be defined.
x＝[ω m ,μ m ,Σ m ]
In the above formula, x is a group of particles to be determined, (ω) m ,μ m ) For the position of the particles Σ m Is the velocity of the particles.
S222: the position and velocity of the particles are updated again in each iteration loop.
The position and speed of the particles are updated again by performing an iterative loop of the following formula:
v i ＝λv i +c 1 random(0,1)(p i -x i )+c 2 random(0,1)(g i -x i )
x i ＝x i +v i
in the above, c 1 And c 2 For acceleration constant, 0.ltoreq.c may be taken 1 ＝c 2 ≤4，v i And x i Representing the velocity and current position of the particle swarm, p i And g i For the best and global best positions searched by the current particle, range (0, 1) represents the interval [0,1 ]]The random number, lambda is an inertia factor, and the dynamic lambda can obtain a better optimizing effect.
S223: the dynamic inertia factor is obtained based on the linearly decreasing weight.
The dynamic inertia factor is obtained by linearly decrementing the weight as follows:
in the above, lambda start And lambda (lambda) end Respectively representing an initial weight and a final weight, k represents the cycle number, T max Representing an upper limit on the number of iterations.
S224: the current optimum value and the global optimum value of the particle are updated and recorded.
The updated particle is calculated according to the set objective function formula (formula in step S214), and the current optimum value and the global optimum value of the particle are updated and recorded.
S225: it is checked whether the upper limit of the number of loops is reached or whether the difference between algebra satisfies an error constraint.
Checking whether the cycle number upper limit is reached or whether the difference between algebra meets the error constraint condition, if so, ending calculation and outputting a result; otherwise, the process goes to step S222 to continue the calculation.
Setting a risk index, comprising the following steps:
s231: and setting a voltage out-of-limit risk index. Obtaining a voltage out-of-limit risk value through a voltage out-of-limit risk index:
Risk(V)＝F(V)Sev hv (V max )
in the above, V is voltage, V max At maximum voltage, F (V) is a multinode voltage joint probability density function, sev hv (V max ) To describe the function of node voltage out-of-limit severity, it is defined as:
in the above, V is voltage, V max Is the maximum voltage.
V max ＝max{V 1 ,V 2 ,...,V k When studying the severity function of k nodes out of limit simultaneously, the node voltage with the largest deviation from the rated voltage is selected to be brought into the severity function.
S232: setting a voltage deviation index. The voltage deviation index BVDI is used for obtaining a voltage deviation value:
in the above, U i Represents the specific value of voltage observation, the subscript i represents the observation times,representing the average value of the sampled voltage, m represents the power sample capacity of the sampling survey. The BVDI value represents concentration or dispersion of bus voltage distribution indexes in a power grid, and the smaller the voltage distribution index is, the more concentrated the bus voltage distribution of the power grid is, and the better the electric energy quality is.
S233: and setting a power flow section out-of-limit risk index. Obtaining the out-of-limit risk value of the power flow section through the out-of-limit risk index of the power flow section:
Risk(P)＝F(P)Sev op (ΔP max )
in the above formula, P is the power of the section tide, F (P) is the joint probability density function of the multi-section tide, and delta P max Sev as maximum power difference op (ΔP max ) To describe the function of the cross-section tidal current out-of-limit severity, the severity is defined as follows:
in the above, P is the section tidal power, P n Max { P-P }, for rated power n And the maximum value of deviation of the investigated tide section from the rated transmission power is shown. When all section power flows are not greater than rated transmission power, namely P<P n When the severity function value is 0, otherwise, taking the maximum value of the section tide deviationValues.
In step S3, a comprehensive index is constructed based on the risk index, and a comprehensive index score is calculated, including the following steps:
s31: and constructing a judgment matrix in the risk index system.
By determining the matrix m= (M ij ) a×a Calculating the weight alpha of a reference layer to a target layer n×1 (n is the number of reference layer elements) and index layer to reference layer weight beta n×m (m is the number of index layer elements corresponding to each reference layer) and consistency check is performed.
S32: and calculating the weight of the reference layer to the target layer and the weight of the index layer to the reference layer.
Combining the weight coefficients to obtain the combination weight gamma of the target layer influence by the index layer element, wherein the calculation method is that each element weight of the index layer is multiplied by the corresponding reference layer weight respectively; and finally, based on the combined weight data, obtaining the comprehensive weight W.
S33: and calculating a combination weight gamma and a comprehensive weight W.
Calculating the score of the comprehensive index as
In the above, S j And the comprehensive weight of the j-th lower index in the index layer is represented, and k is the number of elements contained in the index layer in the evaluation index system.
S34: according to the expression, the comprehensive index score is calculated by a hierarchical analysis method.
In step S4, a composite index score is obtained, a risk level is evaluated based on the composite index score and the risk lookup table, and a solution is executed based on the risk level. In the risk comparison table, different comprehensive index score ranges correspond to different risk levels, and the risk levels correspond to the same or different solutions. For example, the composite index score range A 1 、A 2 、A 3 Respectively correspond to risk levels R 1 、R 2 、R 3 ，R 1 Corresponding solution S 1 ，R 2 And R is 3 Corresponding solution S 2 . Judging the comprehensive index to obtainAnd (5) dividing the corresponding comprehensive index score range, thereby obtaining the risk grade corresponding to the comprehensive index score and the solution corresponding to the risk grade.
Embodiment two:
in this embodiment, a test system is provided for better executing the model proposed by the present invention and performing the calculation of the model. Extracting features from the WECC system forms a minispec system that simplifies the model of the WECC by aggregating nodes below a certain voltage level. The miniWECC system comprises 243 buses, 146 generator sets (comprising 109 synchronous motors and 37 renewable generators), 329 transmission lines, 122 transformers, 7 switching shunts and 139 loads. The renewable energy laboratory proposed method maps the historical data of the original WECC system into this miniWECC system. The "real" data of the minispec system refers to the real historical data mapped from the original WECC system. The uncertain sources include all loads and renewable generators.
Step S5: a miniWECC test system is established, which comprises the following steps:
s51: a generic dynamic model is determined.
For each synchronous generator, a generator model, a turbine governor model, and an excitation system model will be simulated. Due to the aggregate nature of the generator, a low-order dynamic model is selected.
S52: and (5) distributing system inertia, and matching WECC planning cases.
In order for the restored WECC system to have a true inertial frequency response, the inertial constants of the generator model are carefully selected so that the total kinetic energy of the restored system matches the total kinetic energy of the detailed WECC planning case. For this, the 2015 summer WECC planning case was chosen because the online capacity was close to the miniWECC model. The net kinetic energy of the hydro and non-hydro generators for each region in the WECC planning case is mapped to the net kinetic energy of the 240 bus test case.
S53: fine tuning the governor model.
The governor model is fine tuned to maintain the system level frequency response, particularly the frequency nadir. And adjusting the droop setting and the parameters and time constants of the base load generator so as to enable the droop setting and the parameters and the time constants to meet the actual conditions.
As shown in table 1, the system event measurements recorded in the WECC are recorded system responses.
Table 1 index weights and composite scores.
| Index (I) | Weighting of | Score of |
| Reliability of power supply | 0.3798 | 18.7815 |
| Quality of voltage | 0.3757 | 15.1230 |
| Adaptability to | 0.1142 | 5.7975 |
| Economical efficiency | 0.1303 | 6.8376 |
| Comprehensive synthesis | 1 | 46.5296 |
Embodiment III:
this example provides one implementation: the voltage at the test display bus 2202 is low and there are cases where the voltage amplitude is below 0.9, which can cause serious problems for system operation. As can be seen from fig. 2 and table I, the original PCE performs the worst because it cannot handle complex correlations between uncertain inputs. Compared to the original PCE, SPCE has better performance and lower computational cost, but there is a larger error in peak and tail distribution. Note that PCE and SPCE require an accurate PDF of uncertain input, which is a difficult task in practice. In contrast, GP can approximate PDF distribution with better results, which demonstrates its advantage in terms of accuracy. SGPs have similar performance but are computationally more efficient with the aid of sparse methods. SGP can also capture tail distribution, which is important for successful prediction of extreme conditions, such as voltage violations, i.e., voltage amplitudes below 0.9 (about 0.87). Comparison of the CPU time of the two GP-based methods in table I shows that conventional GP cannot be extended to handle high-dimensional uncertainty inputs in the system, while SGP is sufficient to handle high-dimensional data by reducing CPU time by 90%.
For the sample-based approach, LHS achieves acceptable accuracy with fewer samples than the original MC, and QMC is based on low-variance sequences, i.e., sobol sequences herein. In fig. 3, it can be observed that QMC based on Sobol sequences is slightly superior to LHS. As shown in Table I, the efficiency of the enhanced sampling method is highly dependent on the number of samples used. However, both sample-based methods have difficulty capturing the extreme case of a small number of samples, i.e., a large number of samples are required, which can be very time consuming. Furthermore, sample-based methods also require an uncertainty about the exact PDF of the input, which can be very difficult if the amount of historical data is small. In contrast, SGP achieves the best performance in terms of percentage error and computational efficiency, as shown in table 2.
Table 2 results of different methods are compared.
| Method | CPU time |
| PCE | 735.66 |
| SPCE | 39.73 |
| GP | 125.68 |
| SGP | 13.96 |
| LHS | 90.50 |
| QMC | 89.68 |
Notably, SGP is the only method that can accurately delineate the tail of PDF, contributing to future risk analysis and preventive control.
In summary, both GP and SGP are data driven methods for better accuracy, without the need to uncertainty the probability distribution of the input. Thus, sampling errors of the uncertain input probability distribution are mitigated compared to model-based and PCE-based approaches. GP has the problem of handling high-dimensional uncertainty inputs, while SGP solves this problem by sparse techniques, resulting in better accuracy. In terms of computational efficiency, the sampling-based approach requires a large number of samples from the uncertainty-input probability distribution, which results in a higher computational demand than the GP-based approach. Theoretically, the computational complexity of the original GP method has cubic computational complexity. The method has linear calculation complexity by adopting a variable-fraction reasoning and sparse technology. Therefore, the method can be extended to a larger-scale power system and has high calculation efficiency. The time of risk assessment of the power system is reduced, and the real-time performance of risk assessment is improved.
The above embodiments are merely preferred embodiments of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions easily contemplated by those skilled in the art within the technical scope of the present invention should be included in the scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.
Claims (10)
1. The novel power system risk assessment method based on the improved Gaussian model is characterized by comprising the following steps of:
s1: acquiring a power sample;
s2: establishing an evaluation model and setting risk indexes;
s3: constructing a comprehensive index based on the risk index, and calculating a comprehensive index score;
s4: and evaluating the risk level based on the comprehensive index score, and executing the solution based on the risk level.
2. The method for risk assessment of a novel power system based on an improved gaussian model according to claim 1, wherein in S1, the power samples include power data and device data, and the power data includes voltage and power.
3. The novel power system risk assessment method based on the improved gaussian model according to claim 1, wherein in the step S2 of establishing the assessment model, a sparse gaussian model is established by using random variation reasoning and a probability frame is embedded, and the method comprises the following steps:
s211: establishing a Gaussian model;
s212: establishing a sparse Gaussian model based on the Gaussian model;
s213: processing the mean and variance of the sparse Gaussian model;
s214: and calculating the KL divergence between the variational posterior and the true posterior.
4. The novel power system risk assessment method based on the improved gaussian model according to claim 1 or 3, wherein in the step S2 of establishing the assessment model, optimal parameters of the sparse gaussian model are obtained based on an improved particle swarm algorithm, and the method comprises the following steps:
s221: initializing particle positions and speeds based on the to-be-determined particle swarm;
s222: updating the position and the speed of the particles during iterative circulation;
s223: obtaining a dynamic inertia factor based on the linearly decreasing weight;
s224: updating and recording the current optimal value and the global optimal value of the particle;
s225: if the difference between the upper limit of the cycle times or algebra meets the error constraint condition, outputting a result; otherwise, step S222 is performed.
5. A novel power system risk assessment method based on an improved gaussian model according to claim 1 or 3, characterized in that in S2, the risk indicators comprise a voltage out-of-limit risk indicator, a voltage deviation indicator and a power flow section out-of-limit risk indicator.
6. The novel power system risk assessment method based on the improved gaussian model according to claim 1, wherein S3 comprises the steps of:
s31: constructing a judgment matrix based on the risk index;
s32: calculating the weight of the reference layer to the target layer and the weight of the index layer to the reference layer;
s33: calculating a combination weight and a comprehensive weight;
s34: and calculating the comprehensive index score by a analytic hierarchy process.
7. The novel power system risk assessment method based on the improved gaussian model according to claim 1, wherein S4 comprises the steps of:
s41: setting a risk comparison table containing a comprehensive index score range, a risk grade and a solution;
s42: obtaining a comprehensive index score;
s43: matching a comprehensive index score range to which the comprehensive index score belongs;
s44: the risk level is matched based on the comprehensive index score range, and the solution is matched based on the risk level.
8. The method for risk assessment of a novel power system based on an improved gaussian model according to claim 1, wherein a test system for model calculation is provided, comprising: determining a general dynamic model; distributing inertia of the test system and matching planning cases; and adjusting the model parameters of the speed regulator.
9. The novel power system risk assessment method based on the improved Gaussian model according to claim 3, wherein the sparse Gaussian model is built by introducing induction points in the random variation reasoning process, and the complexity of the model is reduced.
10. The method for risk assessment of a novel power system based on an improved gaussian model according to claim 8, wherein the general dynamic model comprises a generator model, a turbine governor model and an excitation system model, and the general dynamic model is a low-order dynamic model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310501193.8A CN116663889A (en) | 2023-05-06 | 2023-05-06 | Novel power system risk assessment method based on improved Gaussian model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310501193.8A CN116663889A (en) | 2023-05-06 | 2023-05-06 | Novel power system risk assessment method based on improved Gaussian model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116663889A true CN116663889A (en) | 2023-08-29 |
Family
ID=87714374
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310501193.8A Pending CN116663889A (en) | 2023-05-06 | 2023-05-06 | Novel power system risk assessment method based on improved Gaussian model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116663889A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117406161A (en) * | 2023-12-11 | 2024-01-16 | 普华讯光（北京）科技有限公司 | Metering device magnitude deviation early warning method, system, equipment and medium |
-
2023
- 2023-05-06 CN CN202310501193.8A patent/CN116663889A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117406161A (en) * | 2023-12-11 | 2024-01-16 | 普华讯光（北京）科技有限公司 | Metering device magnitude deviation early warning method, system, equipment and medium |
| CN117406161B (en) * | 2023-12-11 | 2024-04-02 | 普华讯光（北京）科技有限公司 | Metering device magnitude deviation early warning method, system, equipment and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Decision tree based online stability assessment scheme for power systems with renewable generations | |
| Ma et al. | Ultra-short-term wind generation forecast based on multivariate empirical dynamic modeling | |
| Mestav et al. | State estimation for unobservable distribution systems via deep neural networks | |
| Cheng et al. | Enhanced state estimation and bad data identification in active power distribution networks using photovoltaic power forecasting | |
| CN110766200A (en) | Method for predicting generating power of wind turbine generator based on K-means mean clustering | |
| CN112069727B (en) | Intelligent transient stability evaluation system and method with high reliability for power system | |
| Zhang et al. | A new solar power output prediction based on hybrid forecast engine and decomposition model | |
| CN108847679B (en) | Wind generating set and subsynchronous oscillation identification method, device and system for wind generating set | |
| CN116663889A (en) | Novel power system risk assessment method based on improved Gaussian model | |
| Li et al. | GMM-HMM-based medium-and long-term multi-wind farm correlated power output time series generation method | |
| Jafarzadeh et al. | Probabilistic dynamic security assessment of large power systems using machine learning algorithms | |
| Liao et al. | Scenario generation for cooling, heating, and power loads using generative moment matching networks | |
| Sheng et al. | A hybrid model based on complete ensemble empirical mode decomposition with adaptive noise, GRU network and whale optimization algorithm for wind power prediction | |
| Liao et al. | Scenario generations for renewable energy sources and loads based on implicit maximum likelihood estimations | |
| Zhou et al. | A transfer learning method for forecasting masked-load with behind-the-meter distributed energy resources | |
| An et al. | Combined grey model based on entropy weight method for long-term load forecasting | |
| CN114221334A (en) | Rapid state estimation method based on graph neural network | |
| CN115828441A (en) | Fan output scene generation and reduction method based on condition generation countermeasure network | |
| CN113609456A (en) | Reliability analysis method, system, medium and terminal for photovoltaic charging station of electric vehicle | |
| Zhang et al. | A good point set-based knowledgebase generation scheme for power system intelligent dynamic security assessment | |
| CN113256018B (en) | Wind power ultra-short term probability prediction method based on conditional quantile regression model | |
| Wang et al. | Exploring the application of machine‐learning techniques in the next generation of long‐term hydropower‐thermal scheduling | |
| Wang et al. | Scenario Clustering for Long-term Hydropower-Thermal Scheduling using Shape Feature Extraction | |
| Zheng et al. | A Method for Generating Subsynchronous Oscillation Data of Power System Based on Wasserstein Generative Adversarial Network | |
| CN114256865A (en) | Wind power installed capacity calculation method considering load increase direction randomness |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |