A pcam-gmm-based source-grid-load-storage coordination optimization method for rural distribution network

Abstract

The present application belongs to the technical field of scene optimization technology combined with multi-objective optimization decision, and particularly relates to a rural power distribution network source-grid-load-storage coordinated optimization method based on PCAM-GMM. The problems of high system operation cost and large system pollutant emission in the prior art are solved. The present application optimizes the scheduling of system energy conversion equipment and shared energy storage systems, improves the electric load demand side response capability and demand side response level of the multi-regional integrated energy system under the premise of meeting the safe operation of the multi-regional integrated energy system, improves the energy utilization rate and system economy of the system, and reduces the emission of pollutants.

Description

Technical Field

[0001] This invention belongs to the technical field of combining scenario optimization technology with multi-objective optimization decision-making, specifically involving a coordinated optimization method for rural power distribution network source-grid-load-storage based on PCAM-GMM. Background Technology

[0002] With the integration of numerous distributed power sources into rural power grids, and given the random and unpredictable nature of renewable energy output, rural power distribution network dispatching is significantly impacted. Furthermore, the unique characteristics of agricultural production, including various seasonal loads characterized by short cycles and high loads, contribute to overload or power outages in rural power grids, severely affecting grid security and agricultural production. Therefore, researching coordinated optimization techniques for rural power distribution networks with multiple power supply modes under distributed power source integration is of great significance.

[0003] Currently, scholars have conducted relevant research on the coordinated operation of power sources, grids, loads, and storage in distribution networks. For the problem that neither wind power and photovoltaic power generation on the energy supply side nor load demand on the user side can be accurately predicted, existing research often uses the k-means algorithm for scenario clustering. However, the time complexity of the k-means algorithm increases with the number of samples, and the sample data must have a relatively uniform spherical or near-spherical cluster structure; otherwise, the dataset will not converge. This characteristic makes it difficult to apply to cluster analysis of massive, high-dimensional scenarios. Furthermore, existing research rarely considers the participation of rural loads in demand response, largely failing to account for the differences and uncertainties in the responses of different types of rural loads. There is also a lack of research on incentive compensation mechanisms for user-interrupted loads and transferred loads, necessitating the development of new, flexible, incentive-based demand response mechanisms. Summary of the Invention

[0004] This invention provides a coordinated optimization method for rural power distribution networks based on PCAM-GMM, addressing the problems of high system operating costs and large pollutant emissions in existing technologies. This invention optimizes the scheduling of system energy conversion equipment and shared energy storage systems, improving the demand-side response capability and level of the multi-regional integrated energy system while ensuring its safe operation. This enhances the system's energy utilization rate and economic efficiency, and reduces pollutant emissions.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A coordinated optimization method for rural distribution network sources-grid-load-storage based on PCAM-GMM includes:

[0007] S1. Construct the PCAM-GMM algorithm to generate typical load scenarios;

[0008] S2. Based on the reduction results of typical load scenarios, construct a demand response mechanism and response model that considers both fixed and flexible constraints of agricultural load;

[0009] S3. Construct a coordinated optimization model for rural power distribution network sources-grid-load-storage based on typical load scenarios and demand response mechanisms.

[0010] The step 1 of constructing the PCAM-GMM algorithm to generate typical load scenarios includes using the PCAM-GMM algorithm to perform dimensionality reduction and clustering of multiple scenarios, while retaining the original scenario data, and obtaining typical load scenarios of wind and solar loads through dimensionality reduction and clustering.

[0011] In step 1, the PCAM steps involved in constructing the PCAM-GMM algorithm to generate typical load scenarios include:

[0012] S101. Normalize the raw scene data of photovoltaic power output;

[0013] S102. Calculate the covariance coefficients after data normalization to form the covariance matrix and the orthogonal matrix of the covariance matrix. Perform a linear transformation on the orthogonal matrix to obtain the principal component matrix.

[0014] S103. Calculate the feature retention degree of a single principal component and the feature retention degree of a previous principal component, and extract the dimensionality-reduced information through the feature retention degree index.

[0015] In step 1, the GMM clustering step in constructing the PCAM-GMM algorithm to generate typical load scenarios includes:

[0016] S104. Calculate the posterior probability based on the log-likelihood function of the weights, mean, and covariance matrix.

[0017] S105. Based on the posterior probability, recalculate the weights, mean, and covariance matrix;

[0018] S106. Calculate the log-likelihood function of the Gaussian mixture model;

[0019] S107. Check if the weights, mean, and covariance matrix or the log-likelihood function have converged. If they have not converged, return to S104 until they converge.

[0020] In step 1, the PCAM-GMM algorithm is used to generate typical load scenarios. The PCAM-GMM algorithm is used to calculate the concentration and dispersion of the comprehensive clustering index, and typical load scenarios are generated according to the accuracy requirements.

[0021] The demand response mechanism and response model that considers both fixed and flexible constraints of agricultural load include: a price-based demand response model and an incentive-based demand response model; the total network loss, node voltage deviation, and operating cost are minimized by calculating the price-based demand response model and the incentive-based demand response model.

[0022] The construction of the rural power distribution network source-grid-load-storage coordinated optimization model includes:

[0023] S301. Construct constraints with the optimization objectives of minimizing total network loss, minimizing node voltage deviation, and minimizing operating cost;

[0024] S302. Select a compromise solution from the Pareto front that balances economy and system security, and use a fuzzy multi-attribute decision method to select the best compromise solution.

[0025] The constraints include safety constraints, MT unit operation constraints, new energy power generation operation constraints, and energy storage operation constraints.

[0026] The beneficial effects of this invention are as follows: It establishes a coordinated optimization model for rural power distribution network source-grid-load-storage based on PCAM-GMM scenario reduction, which reduces the operating cost of rural power distribution network and improves the clean energy absorption capacity, power supply reliability, economy and ability to cope with source-load uncertainty. Attached Figure Description

[0027] Figure 1 This is a flowchart of the typical scene generation process using the PCAM-GMM algorithm of this invention;

[0028] Figure 2 This is a diagram of the demand response mechanism of the present invention, taking into account both fixed and flexible constraints on rural loads;

[0029] Figure 3 This is a structural diagram of the PCAM-GMM rural power distribution network source-grid-load-storage coordination optimization model of the present invention; Detailed Implementation

[0030] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments:

[0031] Example 1

[0032] 1. Construct the PCAM-GMM algorithm to generate typical load scenarios:

[0033] The PCAM-GMM algorithm for generating typical load scenarios refers to the large number of typical days comprised of wind power, photovoltaic, and conventional load operation data in rural power distribution networks. The dimensionality of this data is related to equipment type, typical day time period, and the number of typical days, making it difficult for conventional clustering algorithms to cluster high-dimensional data. Therefore, this solution first uses principal component analysis to reduce the dimensionality of the massive operation scenarios of wind power, photovoltaic, and conventional loads, extracting key information, and then performs clustering to improve efficiency and the adaptability of the results.

[0034] First, Principal Component-Gaussian Mixture Clustering (PCAM-GMM) algorithm is used to achieve dimensionality reduction clustering of massive scenes. While fully preserving the original scene data, typical load scenes of wind, solar and load are obtained. The specific approach is as follows.

[0035] 1.1 Principal Component Analysis

[0036] The main idea of ​​principal component analysis (PCAM) is to replace the original n-dimensional data with a k-dimensional orthogonal basis, thereby reducing the data from n-dimensional to k-dimensional. The following section uses high-dimensional photovoltaic power output data as an example to introduce the specific steps of PCAM.

[0037] Assuming the original photovoltaic output data includes N l T l dimensional vector The specific steps of PCAM are as follows:

[0038] 1) Normalization of raw scene data for photovoltaic power output:

[0039]

[0040]

[0041]

[0042] In the formula:

[0043] —for vectors Normalized photovoltaic power output data;

[0044] —respectively vectors The first moment;

[0045] —respectively vectors The second moment.

[0046] 2) Calculate the normalized covariance coefficients of the data and form the covariance matrix:

[0047]

[0048]

[0049] In the formula:

[0050] r mn — Represents two Ts l dimensional vector The covariance between the m-th element and the n-th element in the matrix.

[0051] 3) Calculate the eigenvalues ​​and eigenvectors of R:

[0052]

[0053] In the formula:

[0054] — represents the eigenvalues ​​of the covariance matrix R, and has U is an orthogonal matrix. Based on the eigenvalues, the corresponding eigenvectors a1, a2, ..., a... can be easily calculated. L .

[0055] 4) By performing a linear transformation on matrix U, we can obtain the principal component matrix Z:

[0056] Z = U T S (7)

[0057] 5) Calculate the individual feature retention ω of the nth principal component. n Characteristic retention of the first n principal component components

[0058]

[0059]

[0060] ω n The larger the value, the more original photovoltaic power output information the reduced component contains. Similarly, The larger the value, the more likely it is that the first n principal component components z1, z2, ..., z n The more comprehensive the internal information of the photovoltaic output data, the better. Therefore, an appropriate dimensionality reduction dimension can be selected through a feature retention index to balance the trade-off between computational complexity and the degree of preservation of original data information.

[0061] 1.2 Gaussian Mixture Clustering Algorithm

[0062] Existing research often uses the k-means clustering algorithm to generate typical scenarios because the k-means algorithm has the advantages of being fast and simple. However, it also has disadvantages such as limited accuracy, inflexible class shapes, qualitative probability of a sample belonging to each cluster, and inability to output probability values.

[0063] Based on this, this scheme proposes a Principal Component-Gaussian Mixture Clustering (PCAM-GMM) algorithm, building upon the Gaussian Mixture Model. The biggest difference between this algorithm and the Gaussian Mixture Model is that Principal Component Analysis (PCAM) is used to reduce the dimensionality of massive datasets before clustering. While the conventional Gaussian Mixture Model has the advantages of soft clustering, it also suffers from center selection and sensitivity to noise. PCAM uses principal component analysis to replace high-dimensional data to achieve dimensionality reduction. While ensuring computational accuracy and speed, noise in the data is removed, leaving only linearly independent principal component components, thus reducing the impact of center selection and noise on Gaussian Mixture Clustering.

[0064] A Gaussian mixed model (GMM) is a linear combination of multiple Gaussian distributions, and theoretically can fit arbitrarily complex distributions. The parameters of a Gaussian mixed model include: number K, weights φ, etc. w Mean μ w The covariance matrix ∑w, for a random variable X, is expressed by the GMM as follows:

[0065]

[0066] In the formula:

[0067] φ w —The weight of the w-th component satisfies 0≤φ w ≤1; N w (x|μ w Let ∑w) be the w-th component in the mixture model, expressed as:

[0068]

[0069] GMM uses the Expectation-Maximization (EM) algorithm to fit the mixed Gaussian distribution. Each iteration of the EM algorithm consists of two steps:

[0070] 1) E-step: Calculate the expected value of the estimated parameter;

[0071] 2) M-step: Based on the obtained expected value, solve for the maximum value of the likelihood function.

[0072] Before applying the EM algorithm, it is necessary to estimate the maximum likelihood estimator of the parameters, with a mean μ. wFor example, we find its maximum likelihood estimator. Taking the logarithm of equation (10) yields the log-likelihood function, and then we calculate the maximum likelihood estimator for μ. w Taking the derivative and setting it to zero, we can obtain the maximum log-likelihood estimator, as shown in equation (12):

[0073]

[0074] The specific steps of GMM clustering are as follows:

[0075] Step 1: According to φ w μ w Calculate the posterior probability γ(z) using the log-likelihood function of ∑w. nw ).

[0076]

[0077] Step 2: Based on the posterior probability γ(z) calculated in Step 1 nw ), recalculate φ w μ w 、∑w.

[0078]

[0079]

[0080]

[0081]

[0082] Step 3: Calculate the log-likelihood function of the Gaussian mixture model.

[0083]

[0084] Step 4: Check if the parameters or log-likelihood function have converged. If they have not converged, return to step 1 until they converge.

[0085] 1.3 Comprehensive Clustering Indicators

[0086] The number of clusters significantly impacts computational speed and results. Too few clusters result in less representative clusters, while too many clusters diminish representativeness and increase computational difficulty. Therefore, a reasonable number of clusters is needed to balance representativeness and computational speed. To quantitatively assess the compactness of clustered scenes within the same cluster, an intra-cluster concentration B is defined. r :

[0087]

[0088] In the formula:

[0089] G rj —This refers to the j-th scene in class r;

[0090] n r — This represents the number of scenes in the r-th category.

[0091] To quantitatively evaluate the dispersion between clusters after clustering, dispersion B is defined as follows:

[0092]

[0093] In the formula:

[0094] R represents the total number of categories.

[0095] Based on intra-cluster concentration and dispersion, a comprehensive clustering index O(R) is proposed:

[0096]

[0097] The larger the overall clustering index, the more compact the scenarios within each category, the more obvious the boundaries between categories, and the better the clustering effect.

[0098] 2. Demand response mechanism and response model considering fixed and flexible agricultural load constraints:

[0099] The aforementioned demand response mechanism and model considering both fixed and flexible constraints of rural loads refers to a segmented compensation price method for interruptible loads, addressing the problem that current incentive-based pricing compensation mechanisms cannot meet the differentiated needs of various types of rural loads. A transferable load compensation model with quantified load transfer costs is proposed; and considering the uncertainty of interruptible loads, an optimized operation model for the participation of multiple types of rural loads in the rural distribution network demand response is established.

[0100] The main agricultural load is divided into four parts: crop farming, animal husbandry, agricultural product processing, and residential agricultural load, which participate in centralized agricultural services, as shown in Table 1. Rural load aggregators are established to conduct price-based and incentive-based demand-side responses. Based on various load types and main energy consumption characteristics, rural load-side resources are mainly divided into reducible loads and time-shiftable loads. These two types of demand response resources can respond to grid demand at different time scales.

[0101] Table 1 Main Electricity Loads in Rural Areas

[0102]

[0103]

[0104] 2.1 Price-based demand response model

[0105] Price-based demand response mainly guides users to adjust their electricity consumption periods by implementing time-of-use pricing. The relationship between load response and price change is described by the elasticity coefficient, as shown in equation (22).

[0106]

[0107] In the formula:

[0108] ε — the price elasticity of demand coefficient;

[0109] L 0 —This refers to the amount of electricity;

[0110] p 0 —For electricity prices;

[0111] ΔL and Δp represent the relative increments of electricity consumption and electricity price, respectively.

[0112] A user's electricity demand at a given moment is not only related to the current electricity price, but also affected by electricity prices at other times. Therefore, the relationship between a user's electricity demand and electricity prices across multiple time periods is as follows:

[0113]

[0114] In the formula:

[0115] ε th —This is the mutual elasticity coefficient;

[0116] —These represent the electricity load demand at time t and the electricity price at time h before the implementation of PBDR;

[0117] ΔL t Δp h — These represent the changes in electricity consumption at time t and the changes in prices at time h before and after the implementation of PBDR, respectively.

[0118] The load changes after users participate in PBDR are as follows:

[0119]

[0120] After implementing PBDR, the electricity load L of user f at time t f,t The formula is as follows:

[0121]

[0122] Since user load fluctuates with changes in electricity prices, this scheme introduces a floating factor τ. f,t Simulate price changes in response over different time periods.

[0123]

[0124] In the formula:

[0125] —Total electricity load of user f before PBDR implementation;

[0126] —This represents the average electrical load;

[0127] p f,t —The final response price for user f at time t.

[0128] Combining equations (25) and (26), the total electrical load of users participating in PBDR can be obtained as follows:

[0129]

[0130] In the formula:

[0131] N f —The total number of users participating in PBDR.

[0132] In summary, the PBDR compensation cost (user peak-valley arbitrage) is:

[0133]

[0134] 2.2 Incentive-Based Demand Response Model

[0135] (1) Modeling of interruptible load compensation mechanism

[0136] Because load reduction causes economic losses to users, the power grid implements incentive mechanisms to compensate them. For centralized agricultural service providers, users submit phased compensation price quotations to rural distribution network operators, including the amount of load reduction and related quotes, such as... Figure 2 As shown in the upper section, the first response phase is the basic response phase, and the other response phases are the flexible response phases. While accepting the quote, rural distribution network operators will require users of centralized agricultural service providers to reduce their load and compensate for cost C. IL for:

[0137]

[0138] In the formula:

[0139] N l —The total number of users participating in the demand response;

[0140] N m —The total number of stages for progressively pricing the user;

[0141] c t,l,m —The price quote for the m-th stage provided to user l;

[0142] — The planned response power consumption for user l during time period t.

[0143] (2) Modeling of transferable load compensation mechanism

[0144] This scheme uses a two-dimensional alternating function based on load transferable time (LTT) and load transferable power (LTP) to quantify the compensation cost of transferable load. The load transfer cost is a nonlinear function of LTT and LTP, and the cost increases more significantly with the increase of LTT and LTP. The formula for the TL compensation cost CTL is as follows:

[0145]

[0146]

[0147] In the formula:

[0148] N g —The total number of users participating in TL;

[0149] T r —This is the period for load transfer;

[0150] —For user g in time period t r The load transfer power;

[0151] c E,base c T,base —Base prices based on LTT and LTP, respectively;

[0152] —This represents the unit price increment;

[0153] —The final prices for LTT and LTP, respectively;

[0154] —For user g in time period t r The increase in the originally planned electricity transfer;

[0155] —For user g during the originally planned transfer period t r The time period increment;

[0156] E n T n —These represent the actual load transfer amount and the actual load transfer period, respectively;

[0157] N E N T—These represent the user's original planned load transfer amount and load transfer period, respectively.

[0158] 2.3 Modeling Uncertainty in User Response

[0159] Considering the uncertainty of user responses, the actual response capability of incentive-based demand responses will differ from expectations. This solution proposes an incentive mechanism based on fixed and flexible constraints to avoid the impact of uncertainty, such as... Figure 2 As shown in the lower part, ω l,m,t This refers to the fluctuation range of uncertain response quantities within the real-time range.

[0160] based on Figure 2 The planned response power of the user in phase m should meet the capacity limit of phase m in the contract.

[0161]

[0162] In the formula:

[0163] —The planned response power for user l in stage m;

[0164] — This represents the maximum response amount in the m-th stage of the contract.

[0165] The actual response power of the user is based on the planned response power, and fluctuations are allowed within the range specified in the contract. The uncertainty of the response load with normal distribution characteristics is described by a stochastic programming method.

[0166]

[0167] In the formula:

[0168] —This represents the deviation between the actual response power and the planned response power.

[0169] —This is to account for the actual electricity consumption of user l in the m-th stage t-period after taking into account the uncertainty of demand response;

[0170] δ — the standard deviation of the normal distribution.

[0171] Figure 2 The first stage is the fixed compensation stage, and the other stages are flexible compensation stages. The range of quantities in each stage is determined by the probability density function of the normal distribution mentioned above. The compensation cost, taking into account the uncertainty of demand response, is as follows:

[0172]

[0173] 3. Coordinated Optimization Model for Rural Power Distribution Network (Source-Grid-Load-Storage):

[0174] The aforementioned rural power distribution network source-grid-load-storage coordination optimization model refers to the optimization of source-grid-load-storage coordination by combining typical scenarios generated by the PCAM-GMM algorithm and various rural load demand response models, with the objective functions of minimizing total network loss, voltage deviation, and operating cost, to ensure that the optimal scheduling scheme can take into account both system reliability and economy under uncertain risks.

[0175] 3.1 Constructing the objective function

[0176] (1) Security

[0177] The optimization objective is to minimize system network loss and node voltage deviation, specifically expressed as:

[0178] minF1=min{f loss ,f u} (35)

[0179] In the formula:

[0180] f loss —This represents the total network loss of the system, including active power losses in AC and DC branches and within the converter station.

[0181] f u — This refers to the node voltage deviation of the system, including the voltage deviation on both the AC and DC sides.

[0182] 1) Objective function 1: Minimize total network loss.

[0183]

[0184] In the formula:

[0185] T represents the total number of scheduling periods;

[0186] p s —The probability of scenario s occurring;

[0187] —These are the sets of branches in AC and DC distribution networks, respectively;

[0188] N VSC —A collection of converter stations;

[0189] — This refers to the active power loss of converter station i during time period t under scenario s;

[0190] These are branch lines l in the AC distribution network and branch lines l in the DC distribution network during time period t under scenario s. The current flowing through it;

[0191] These are branch 1 in the AC distribution network and branch 1 in the DC distribution network. The resistance.

[0192] 2) Objective function 2: Minimize node voltage deviation

[0193]

[0194] In the formula:

[0195] —These are the sets of nodes for AC and DC distribution networks, respectively;

[0196] — This represents the per-unit voltage value of node j during time period t in scenario s.

[0197] (2) Economic efficiency

[0198] The objective function is to minimize the overall system operating cost. This overall operating cost includes not only the energy consumption cost of the gas turbine generator set, but also the cost of wind and solar curtailment penalties, demand response costs, and energy storage charging and discharging costs. The objective function expression is as follows:

[0199]

[0200] In the formula:

[0201] F2 represents the comprehensive operating cost of coordinated and optimized scheduling of source-grid-load-storage;

[0202] Ω G — Represents a collection of thermal power units;

[0203] —This represents the active power of unit i during time period t;

[0204] Ω W — Represents the set of new energy power station nodes;

[0205] C curt,k — This represents the cost coefficient for the curtailment of wind and solar power at the k-th renewable energy power station;

[0206] δ k,t — This represents the wind and solar curtailment rate of the k-th renewable energy power station during time period t, with a value ranging from 0 to 1;

[0207] — This represents the random active power of the k-th renewable energy power station during time period t;

[0208] E(·) — represents the expectation operation of a random variable;

[0209] Ω DR —This represents the set of DRs participating in the scheduling;

[0210] CDR,d —This represents the response cost coefficient of the d-th DR;

[0211] — This represents the active power response during the d-th DR time period t;

[0212] a, b, and c are parameters of the consumption characteristic curve, respectively.

[0213] ε c ε d —These represent the cost coefficients for charging and discharging the energy storage system, respectively;

[0214] —The charging power of the energy storage system during time period t;

[0215] — This represents the discharge power of the energy storage system during time period t.

[0216] 3.2 Constructing Constraints

[0217] (1) Safety constraints

[0218] The safety constraints of rural power distribution networks mainly include node voltage, branch current upper and lower limit constraints and branch transmission power limits, as shown in Equation (39).

[0219]

[0220] In the formula:

[0221] — This represents the active and reactive power of unit i during time period t;

[0222] —These represent the voltage amplitude of node i during time period t;

[0223] — This represents the active and reactive power transmission of transmission lines i and j during time period t. The subscripts min and max represent the lower and upper limits of this variable, respectively.

[0224] (2) Operating constraints of MT units

[0225]

[0226]

[0227] In the formula:

[0228] —These are the lower and upper limits of MT's output, respectively;

[0229] —These are the technical parameters for the MT unit's upward and downward ramping operations, respectively.

[0230] (3) Operational constraints of new energy power generation

[0231]

[0232]

[0233] In the formula:

[0234] —Reduced output active power for typical wind power scenarios;

[0235] —The active power of wind power under dispatch;

[0236] —Reduced output active power for typical photovoltaic power generation scenarios;

[0237] — This refers to the active power of photovoltaic power generation under dispatch.

[0238] (4) Energy storage operation constraints

[0239]

[0240]

[0241]

[0242] In the formula:

[0243] —The maximum allowable charging and discharging power of the energy storage device;

[0244] CAP BS —This refers to the maximum capacity of the energy storage device;

[0245] —The amount of electricity stored in the energy storage device;

[0246] β BS —This represents the loss rate of the energy storage device.

[0247] 3.3 Fuzzy Multi-Attribute Decision-Making Method

[0248] Solving multi-objective problems is not merely an optimization problem, but also a decision-making problem. To obtain the final scheduling scheme, a compromise solution that balances economy and system security needs to be selected from the Pareto front for the scheduler to choose from. This scheme employs a fuzzy multi-attribute decision-making method to select the optimal compromise solution, the process of which is as follows:

[0249] (1) Calculate fuzzy sets

[0250] The fuzzy set is defined by the membership function shown in equation (47). Sure

[0251]

[0252] In the formula:

[0253] —The maximum and minimum values ​​of the i-th objective in the Pareto front.

[0254] (2) Calculate the fuzzy membership degree

[0255] For the j-th non-dominated solution in the Pareto front, its standardized fuzzy membership degree μ k The highest fuzzy membership value can be calculated by equation (48), and the final output is shown in equation (49).

[0256]

[0257] opt={o|μ o =max(μ k )} (49)

[0258] In the formula:

[0259] n—the number of non-dominated solutions in the Pareto front;

[0260] m—the number of objectives; a compromise solution is the set {μ}. k The solution corresponding to the larger value in}.

[0261] Example 2

[0262] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the source-grid-load-storage coordinated optimization method for rural distribution networks based on PCAM-GMM provided in Embodiment 1 of this invention.

[0263] Example 3

[0264] This embodiment provides a computer-readable storage medium storing a computer program, characterized in that, when the program is executed by a processor, it implements a coordinated optimization method for rural distribution network source-grid-load-storage based on PCAM-GMM provided in Embodiment 1 of the present invention.

[0265] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0266] The solutions in this application embodiment can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0267] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0268] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0269] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0270] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0271] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A PCAM-GMM based source-grid-load-storage coordination optimization method for rural distribution network, characterized in that, include: S1. Construct the PCAM-GMM algorithm to generate typical load scenarios; S2. Based on the reduction results of typical load scenarios, construct a demand response mechanism and response model that considers both fixed and flexible constraints of agricultural load; S3. Construct a coordinated optimization model for rural power distribution network sources-grid-load-storage based on typical load scenarios and demand response mechanisms; Among them, the demand response mechanism and response model considering fixed and flexible constraints of agricultural load include: price-based demand response model and incentive-based demand response model; the price-based demand response model and incentive-based demand response model are used to calculate the minimum total network loss, the minimum node voltage deviation and the minimum operating cost; the price-based demand response model guides users to adjust their electricity consumption time by implementing time-of-use pricing, and the relationship between load response and price change is described by elasticity coefficient, as shown in equation (22); （22） In the formula: —This is the price elasticity of demand coefficient; —This refers to the amount of electricity; —For electricity prices; , —These represent the relative increases in electricity consumption and electricity price, respectively; The relationship between user electricity demand and electricity prices across different time periods is as follows: （23） In the formula: —This is the mutual elasticity coefficient; , —These represent the electricity load demand at time t and the electricity price at time h before the implementation of PBDR; , — These represent the changes in electricity consumption at time t and the changes in prices at time h before and after the implementation of PBDR, respectively; The load changes after users participate in PBDR are as follows: （24） The electricity load of user f at time t after PBDR implementation The formula is as follows: （25） Since user load fluctuates with changes in electricity prices, this solution introduces a floating factor. Simulate price changes in response over different time periods; （26） In the formula: —Total electricity load of user f before PBDR implementation; —This represents the average electrical load; —The final response price for user f at time t; Combining equations (25) and (26), the total power load of users participating in PBDR can be obtained as follows: （27） In the formula: —The total number of users participating in PBDR; In summary, the PBDR compensation cost (user peak-valley arbitrage) is: （28） In the incentive-based demand response model, the first response stage is the basic response stage, and the other response stages are the flexible response stages. Rural distribution network operators, while accepting the quoted price, will require users of centralized agricultural service providers to reduce their load to compensate for costs. for: （29） In the formula: —The total number of users participating in the demand response; —The total number of stages for progressively pricing the user; —For users The price quote provided for the m-th stage; —For users Planned response power during time period t; Load transfer cost is a nonlinear function of LTT and LTP. The cost increases more significantly with increasing LTT and LTP. The formula for TL compensation cost CTL is as follows: （30） （31） In the formula: —The total number of users participating in TL; —This is the period for load transfer; —For user g during the time period The load transfer power; , —Base prices based on LTT and LTP, respectively; —This represents the unit price increment; , —The final prices for LTT and LTP, respectively; —For user g during the time period The increase in the originally planned electricity transfer; —For user g during the originally planned transfer period The time period increment; , —These represent the actual load transfer amount and the actual load transfer period, respectively; , —These represent the user's originally planned load transfer amount and load transfer period, respectively; The construction of the rural power distribution network source-grid-load-storage coordinated optimization model includes: S301. By minimizing total network loss, minimizing node voltage deviation, and minimizing operating cost as optimization objectives, constraints are constructed; the constraints include safety constraints, MT unit operation constraints, new energy power generation operation constraints, and energy storage operation constraints. S302. Select a compromise solution from the Pareto front that balances economy and system security, and use a fuzzy multi-attribute decision-making method to select the best compromise solution; Minimizing system network loss and node voltage deviation is the optimization objective, specifically expressed as: （35） In the formula: —This represents the total network loss of the system, including active power losses in AC and DC branches and within the converter station. — This refers to the node voltage deviation of the system, including the voltage deviation on both the AC and DC sides; 1) Objective function 1: Minimize total network loss; （36） In the formula: T represents the total number of scheduling periods; —The probability of scenario s occurring; , —These are the sets of branches in AC and DC distribution networks, respectively; —A collection of converter stations; — This refers to the active power loss of converter station i during time period t under scenario s; , These are the branch circuits in the AC distribution network during time period t under scenario s. Branch circuits in DC distribution network The current flowing through it; , Branch circuits in AC distribution network Branch circuits in DC distribution network The resistance; 2) Objective function 2: Minimize node voltage deviation （37） In the formula: , —These are the sets of nodes for AC and DC distribution networks, respectively; —This represents the per-unit voltage value of node j during time period t in scenario s; The objective function, which aims to minimize the overall system operating cost, is expressed as follows: （38） In the formula: —This represents the comprehensive operating cost of coordinated and optimized scheduling of power generation, grid, load, and storage. — Represents a collection of thermal power units; —This represents the active power of unit i during time period t; — Represents the set of new energy power station nodes; — This represents the cost coefficient for the curtailment of wind and solar power at the k-th renewable energy power station; — This represents the wind and solar curtailment rate of the k-th renewable energy power station during time period t, with a value ranging from 0 to 1; — This represents the random active power of the k-th renewable energy power station during time period t; — Represents the expectation operation of random variables; —This represents the set of DRs participating in the scheduling; —This represents the response cost coefficient of the d-th DR; — This represents the active power response during the d-th DR time period t; —These are the parameters of the consumption characteristic curve; —These represent the cost coefficients for charging and discharging the energy storage system, respectively; —The charging power of the energy storage system during time period t; —This represents the discharge power of the energy storage system during time period t; In the constraints: The specific safety constraints of rural power distribution networks are shown in equation (39); （39） In the formula: , — This represents the active and reactive power of unit i during time period t; —These represent the voltage amplitude of node i during time period t; , — This represents the active and reactive power transmission of transmission lines i and j during time period t. The subscripts min and max represent the lower and upper limits of this variable, respectively. MT unit operating constraints （40） （41） In the formula: , —These are the lower and upper limits of MT's output, respectively; , —These are the technical parameters for the MT unit's upward and downward ramps, respectively; New energy power generation operation constraints （42） （43） In the formula: —Reduced output active power for typical wind power scenarios; —The active power of wind power under dispatch; —Reduced output active power for typical photovoltaic power generation scenarios; —The active power of photovoltaic power generation under dispatch; Energy storage operation constraints （44） （45） （46） In the formula: —The maximum allowable charging and discharging power of the energy storage device; —This refers to the maximum capacity of the energy storage device; —The amount of electricity stored in the energy storage device; —This refers to the loss rate of the energy storage device; The process of the fuzzy multi-attribute decision method is as follows: The fuzzy set is represented by the membership function shown in equation (47). Sure （47） In the formula: , —The maximum and minimum values ​​of the i-th objective in the Pareto front; For the j-th non-dominated solution in the Pareto front, its standardized fuzzy membership degree It can be calculated by equation (48), and the final output of the highest fuzzy membership value is shown in equation (49); （48） （49） In the formula: —The number of non-dominated solutions in the Pareto front; —The number of objectives, a compromise solution is a set. The solution corresponding to the larger value in the middle.

2. The method for coordinated optimization of rural distribution network sources-grid-load-storage based on PCAM-GMM according to claim 1, characterized in that: The step 1 of constructing the PCAM-GMM algorithm to generate typical load scenarios includes using the PCAM-GMM algorithm to perform dimensionality reduction and clustering of multiple scenarios, while retaining the original scenario data, and obtaining typical load scenarios of wind and solar loads through dimensionality reduction and clustering.

3. The method for coordinated optimization of rural distribution network sources-grid-load-storage based on PCAM-GMM according to claim 2, characterized in that: In step 1, the PCAM steps involved in constructing the PCAM-GMM algorithm to generate typical load scenarios include: S101. Normalize the raw scene data of photovoltaic power output; S102. Calculate the covariance coefficients after data normalization to form the covariance matrix and the orthogonal matrix of the covariance matrix. Perform a linear transformation on the orthogonal matrix to obtain the principal component matrix. S103. Calculate the feature retention degree of a single principal component and the feature retention degree of a previous principal component, and extract the dimensionality-reduced information through the feature retention degree index.

4. The method for coordinated optimization of rural distribution network sources-grid-load-storage based on PCAM-GMM according to claim 2, characterized in that: In step 1, the GMM clustering step in constructing the PCAM-GMM algorithm to generate typical load scenarios includes: S104. Calculate the posterior probability based on the log-likelihood function of the weights, mean, and covariance matrix. S105. Based on the posterior probability, recalculate the weights, mean, and covariance matrix; S106. Calculate the log-likelihood function of the Gaussian mixture model; S107. Check if the weights, mean, and covariance matrix or the log-likelihood function have converged. If they have not converged, return to S104 until they converge.

5. A method for coordinated optimization of rural distribution network sources-grid-load-storage based on PCAM-GMM according to claim 2, characterized in that: In step 1, the PCAM-GMM algorithm is used to generate typical load scenarios. The PCAM-GMM algorithm is used to calculate the concentration and dispersion of the comprehensive clustering index, and typical load scenarios are generated according to the accuracy requirements.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the coordinated optimization method for rural power distribution network source-grid-load-storage based on any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the coordinated optimization method for source-grid-load-storage of rural distribution network based on PCAM-GMM as described in any one of claims 1 to 5.

Images