Real-time stochastic dispatching method and system for active power distribution network based on fuzzy neural network pre-decision

Abstract

The application discloses a kind of active power distribution network real-time random scheduling method and system of fuzzy neural network predecision fusion.It is described that the uncertainty of photovoltaic output and the randomness of load are corresponding prediction error uncertainty based on the rederivation of power flow model based on Euler equation, and the probability distribution of uncertain variable is obtained by data-driven method;Random optimization power scheduling model of active power distribution network is established;Fuzzy neural network is used to describe the uncertainty of random optimization scheduling model, and historical data is used for training;The scheduling strategy is pre-decided by using trained neural network to input current time power distribution network information;The initial value of the solver optimization is obtained by using the pre-decision value, and then accurate solution is carried out, so as to ensure the accuracy of the solution result while accelerating the model solving process.The application overcomes the defect that the power flow model based on second-order cone relaxation may violate the relaxation condition and cause error, and ensures the real-time of scheduling strategy.

Description

Technical Field

[0001] This invention relates to the field of distribution network optimization technology, specifically to a real-time stochastic scheduling method and system for active distribution networks that integrates fuzzy neural network pre-decision. Background Technology

[0002] To alleviate environmental pollution, address the traditional energy crisis, and reduce carbon emissions from the power industry, distributed photovoltaic (PV) power is being extensively integrated into distribution networks. Future distribution networks will contain a high proportion of new energy sources and distributed power sources. To address the randomness of new energy sources, and with the increasing integration of distributed energy storage into distribution networks, traditional distribution networks are transforming into active distribution networks.

[0003] Optimal dispatching of distribution networks is essential for achieving economical and safe operation. Solving optimal dispatching models will become more frequent. However, as traditional distribution networks become increasingly complex, with the large-scale integration of distributed resources, uncertainty in the distribution network increases, making model solving more difficult and slower. Establishing effective models and finding efficient solution methods are challenges that must be addressed. On the one hand, establishing accurate and realistic active distribution network dispatching models hinges on using appropriate and reliable power flow models and accurately describing their uncertainties. Currently, power flow equations are widely based on second-order cone relaxation, but this may lead to errors due to the failure of relaxation conditions to accurately reflect the power flow equations before relaxation. On the other hand, it is necessary to develop fast and real-time solution methods. Traditional solvers may suffer from slow solution speeds in solving optimization problems. Due to the development of deep neural networks, some scholars have used them for solving optimization problems. However, current neural network-based methods may suffer from high training costs or deviations between predicted and actual values, leading to non-convergence of optimization constraints. Summary of the Invention

[0004] Purpose of the invention: To address the challenges of increasing uncertainty in distribution networks and the difficulty in directly solving optimization models when the number of variables increases, while also considering the potential inaccuracy of the solution results, this invention proposes a real-time stochastic scheduling method and system for active distribution networks that integrates fuzzy neural network pre-decision making, ensuring the real-time performance of the model without sacrificing solution accuracy.

[0005] Technical solution: A real-time stochastic dispatching method for active distribution networks that integrates fuzzy neural network pre-decision making, comprising the following steps:

[0006] Step 1: Based on the Euler equations, re-derive the power flow model of the active distribution network, including the network loss equation, active power balance equation, reactive power balance equation, voltage drop equation, and line current constraints. The derived model is equivalent to the original power flow model.

[0007] Step 2: Treat the predicted power values ​​of photovoltaic output and load as deterministic variables, and the prediction error as an uncertain variable. Decompose the actual power of photovoltaic output and load into two parts: predicted power value and prediction error. The uncertainty of photovoltaic output and the randomness of load are described as the uncertainty of the corresponding prediction error. The probability distribution of the uncertain variable is obtained through a data-driven method.

[0008] Step 3: Taking into account the power supply cost of the upstream substation, the distribution network loss, the dispatch cost of the distributed energy storage system (ESS), and the dispatch cost of the static var compensator (SVC), construct the objective function of the active distribution network stochastic optimization power dispatch model with the goal of minimizing cost. Its constraints include power flow constraints, safe operation constraints, substation injected power constraints, photovoltaic unit operation constraints, ESS operation constraints, SVC constraints, and flexible load constraints.

[0009] Step 4: The uncertainty of the active distribution network stochastic optimization power dispatch model is described by integrating a fuzzy neural network, and historical data is used to train the fuzzy neural network. The forward propagation process of the fuzzy neural network includes a data input layer, a membership function calculation layer, a rule generation layer, a normalization layer, and an output layer. The back propagation process uses the negative gradient descent algorithm to optimize the center point, width vector of the membership function, and connection weights of the output layer to obtain the parameters of the fuzzy neural network.

[0010] Step 5: Input the real-time status data of the active distribution network, including real-time power prediction values ​​and data-driven prediction errors, into the trained fuzzy neural network, and use the fuzzy neural network to quickly output the pre-decision of the scheduling strategy.

[0011] Step 6: Use the pre-decision value as the initial value for the solver to find the optimal value, and then use the solver to accurately solve the active distribution network stochastic power dispatch model in step (3) to obtain the real-time stochastic dispatch strategy of the active distribution network.

[0012] A real-time stochastic dispatching system for active power distribution networks that integrates fuzzy neural network pre-decision making includes:

[0013] The power flow model re-derivation module is used to re-derive the power flow model of the active distribution network based on the Euler equations, including network loss equations, active power balance equations, reactive power balance equations, voltage drop equations, and line current constraints. The derived model is equivalent to the original power flow model.

[0014] The uncertainty description module treats the power prediction values ​​of photovoltaic output and load as deterministic variables and the prediction error as an uncertain variable. It decomposes the actual power of photovoltaic output and load into two parts: power prediction value and prediction error. The uncertainty of photovoltaic output and the randomness of load are described as the uncertainty of the corresponding prediction error, and the probability distribution of the uncertain variable is obtained through a data-driven method.

[0015] The scheduling model construction module is used to comprehensively consider the power supply cost of the upstream substation, the distribution network loss, the scheduling cost of the distributed energy storage system (ESS), and the scheduling cost of the static var compensator (SVC). It constructs the objective function of the active distribution network stochastic optimization power scheduling model with the goal of minimizing costs. Its constraints include power flow constraints, safe operation constraints, substation injected power constraints, photovoltaic unit operation constraints, ESS operation constraints, SVC constraints, and flexible load constraints.

[0016] The fuzzy neural network description and training module is used to describe the uncertainty of the stochastic optimal power dispatch model of the active distribution network by integrating fuzzy neural networks, and to train the fuzzy neural network using historical data. The forward propagation process of the fuzzy neural network includes a data input layer, a membership function calculation layer, a rule generation layer, a normalization layer, and an output layer. The back propagation process uses the negative gradient descent algorithm to optimize the center point, width vector of the membership function, and connection weights of the output layer to obtain the parameters of the fuzzy neural network.

[0017] The pre-decision acquisition module is used to input real-time status data of the active distribution network, including real-time power prediction values ​​and data-driven prediction errors, into a trained fuzzy neural network, and use the fuzzy neural network to quickly output pre-decision of scheduling strategy.

[0018] The real-time random scheduling strategy acquisition module is used to take the pre-decision value as the initial value for the solver to optimize, and then use the solver to accurately solve the active distribution network random optimization power scheduling model in the scheduling model construction module to obtain the real-time random scheduling strategy of the active distribution network.

[0019] The present invention also provides a computer device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, wherein when the programs are executed by the processors, they implement the steps of the active distribution network real-time random scheduling method for pre-decision fusion of fuzzy neural networks as described above.

[0020] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the active distribution network real-time random scheduling method for pre-decision fusion of fuzzy neural networks as described above.

[0021] Beneficial effects:

[0022] (1) This invention re-derives the power flow model and accurately describes the uncertain variables of the active distribution network, and establishes an economical and safe stochastic optimal power dispatch model for the active distribution network, which helps dispatchers to optimize the dispatch of the active distribution network with photovoltaic power output fluctuations and random load changes.

[0023] (2) This invention proposes a method for efficient solution of stochastic optimization power scheduling model by integrating fuzzy neural network pre-decision method. The membership function of fuzzy neural network can be used to characterize the probability distribution of uncertain variables in stochastic optimization. The output of the neural network is used to make pre-decision for decision variables. The pre-decision value is used as the initial point for the solver to find the optimal solution. In the end, the solution of the model is accelerated without sacrificing the accuracy of the model solution. Attached Figure Description

[0024] Figure 1 This is a flowchart of the method of the present invention;

[0025] Figure 2 A network topology diagram for an improved IEEE 33-node system;

[0026] Figure 3 This is a comparison chart of the predicted and actual values ​​of the energy storage charging and discharging power by a fuzzy neural network. Detailed Implementation

[0027] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0028] Figure 1 The flowchart of the real-time stochastic dispatching method for active distribution networks based on fuzzy neural network pre-decision proposed in this invention is shown, including the following steps:

[0029] Step 1: Based on the Euler equations, the power flow model is re-derived. The proposed model is equivalent to the original power flow model and can be solved efficiently.

[0030] Step 2: The uncertainty of photovoltaic output and the randomness of load are essentially described as the uncertainty of the corresponding prediction error, and the probability distribution of the uncertain variables is obtained through a data-driven approach;

[0031] Step 3: Based on Step 1 and Step 2, establish an economical and safe stochastic optimization power dispatch model for the active distribution network. The objective function comprehensively considers multiple factors, and the constraints include power flow constraints, safe operation constraints, etc.

[0032] Step 4: Use a fuzzy neural network to describe the uncertainty of the stochastic optimization scheduling model in Step 3, and train it using historical data to obtain the neural network parameters;

[0033] Step 5: Input the current distribution network information and use the trained neural network to make pre-decision decisions on the scheduling strategy;

[0034] Step 6: Use the pre-decision value as the initial value for the solver to find the optimal value, and then use the solver to accurately solve the stochastic optimization scheduling model in step (3), thereby accelerating the model solving process while ensuring the accuracy of the solution results, and obtaining the real-time stochastic scheduling strategy of the active distribution network.

[0035] This invention uses an improved IEEE 33-bus system as an experimental object to detail the specific implementation process of active distribution network stochastic optimization power dispatch using the method of this invention. The system network topology diagram is as follows: Figure 2 As shown in the diagram. In this system, 10 nodes are connected to a distributed photovoltaic (PV) system to simulate the uncertainties of new energy sources. Six energy storage devices and ten SVCs are also configured to ensure stable system operation. The PV output is 10%-90% of its rated output, and the load is 80%-100% of the rated load of the IEEE 33-node system. To generate power values ​​for both PV and load at multiple time points, the fluctuation range is uniformly sampled as the power prediction value. Typical scenarios for power prediction errors are obtained by scenario reduction from historical data of typical similar days. The historical data is generated through Monte Carlo simulation. Based on the constructed model data, the specific implementation steps of the method of this invention are as follows:

[0036] (1) The power flow model is re-derived based on the Euler equation, including the network loss equation, active power balance equation, reactive power balance equation, voltage drop equation, and line current constraint:

[0037] a) For the network loss equation, the original network loss function formula is expressed as:

[0038]

[0039] In the formula: P mn and Q mn These represent the active and reactive power flowing through line m and n, respectively. and These represent the active and reactive power losses on line mn, respectively. mn R is the current flowing through line mn. mn and X mn These represent the resistance and reactance of line mn, respectively; U m Let be the voltage amplitude at node m.

[0040] b) For the active and reactive power balance equations, the active and reactive power balance at node m is expressed as:

[0041]

[0042] In the formula: These represent the active power injected into node m from three types of distributed resources: substations, distributed photovoltaics, and distributed energy storage. These represent the reactive power injected into node m from the substation, distributed photovoltaic system, and static var compensator, respectively. The active and reactive power of the load at node m are P, respectively. km Q km These represent the active and reactive power injected into node m at line km, respectively, in P. mn Q mn These represent the active and reactive power flowing out of node m on line mn, respectively. This is a set of routes.

[0043] c) For the voltage drop equation, the initial definition of voltage drop is:

[0044]

[0045] In the formula: Z mn The impedance of the line is mn, i.e., Z mn =R mn +jX mn j is the imaginary unit, V m V n Let V be the complex voltage at nodes m and n. m =|V m |∠θ m V n =|V n |∠θ n ,|V m |、|V n | represents the voltage magnitudes at nodes m and n, respectively, and ∠θ m ,∠θ n The voltage phase angles at nodes m and n are respectively; S mn S represents the line apparent power. mn =P mn +jQ mn ; and S mn and V m The conjugate of complex numbers.

[0046] Since the initial definition formula for voltage drop contains complex terms, direct optimization is difficult. Therefore, we relax the voltage phase angle as follows:

[0047] First, introduce the concept of multiplying both sides.

[0048]

[0049] V m Vn Z mn S mn Expanding on this, we have:

[0050]

[0051] ∠θ nm writing Then, using Euler's formula: e jθ =cosθ+jsinθ, where ∠θ is the voltage phase angle difference. nm After rewriting, we get:

[0052] |V m | 2 -(|V m ||V n |cosθ nm +j|V m ||V n |sinθ nm )=(R mn +jX mn (P) mn -jQ mn )

[0053] Separate the real and the virtual:

[0054] |V m ||V n |cosθ mn =|V m | 2 -(R mn P mn +X mn Q mn )

[0055] |V m ||V n |sinθ mn =R mn Q mn -X mn P mn

[0056] Add the squares of the two equations together:

[0057] (|V m | 2 -(R mn P mn +X mn Q mn )) 2 +(R mn Q mn -X mn P mn ) 2 =(|Vm ||V n |) 2

[0058] Expand the squared term and divide by |V m | 2 :

[0059]

[0060] The above equation is the voltage drop formula after voltage phase angle relaxation. Compared with the commonly used second-order cone relaxation, this method does not require the introduction of assumptions and is equivalent to the original power flow equation. It can be solved efficiently using existing nonlinear solvers.

[0061] d) For line current constraints, we have:

[0062]

[0063]

[0064]

[0065] In the formula: This represents the upper limit of the current amplitude flowing through line mn.

[0066] (2) Description of uncertainties in active distribution networks:

[0067] First, due to the randomness of photovoltaic (PV) power and load, power distribution network scheduling is usually based on the power prediction data of both. Therefore, the power prediction data is a deterministic variable. The uncertainty in an active power distribution network is essentially the uncertainty of the prediction error. The actual power of PV power and load can be decomposed into two parts: the predicted power value and the prediction error, as shown below:

[0068] P PV =P PV,pre +ξ PV

[0069] P D0 =P D,pre +ξ DP

[0070] Q D0 =Q D,pre +ξ DQ

[0071] In the formula: P PV P D0 Q D0 These represent the actual values ​​of photovoltaic power output and active and reactive power of the load, respectively, P PV,pre P D,pre Q D,preξ represents the predicted values ​​of photovoltaic power output and active and reactive power of the load, respectively. PV ξ DP ξ DQ These represent the prediction errors for photovoltaic power output and active and reactive power of the load, respectively.

[0072] Secondly, the objective of stochastic optimization is to optimize the expected value of the objective function under uncertainty. In active distribution networks, the uncertainty variable is the uncertainty of power prediction error. To avoid excessive historical data scenarios causing the stochastic optimization scheduling model to have too high a dimension, resulting in slow model solution speed or inability to solve directly, k-means clustering algorithm and synchronous back-substitution elimination technique based on Kantorovich probability distance are used to reduce the prediction data of similar days of power prediction to obtain typical scenarios of prediction error and the corresponding probabilities of typical scenarios: First, the k-means clustering algorithm is used to divide the original scenarios into different clusters. Then, the synchronous back-substitution elimination algorithm based on Kantorovich distance is used to eliminate the scenario sets within each cluster into unique typical scenarios. The final scenario sets of each cluster are the reduced scenario sets that can represent the original scenario sets. Thus, a small number of typical scenarios can effectively represent all historical data scenarios, balancing the accuracy and real-time performance of the model solution. Since the scenario reduction technique used is an existing technology, the specific details are not elaborated here.

[0073] (3) Constructing a stochastic power dispatch model for active distribution networks:

[0074] First, taking into account the power supply cost of the upstream substation, the distribution network loss, the dispatch cost of the distributed energy storage system (ESS), and the dispatch cost of the static var compensator (SVC), the objective function of the active distribution network stochastic optimization power dispatch model is constructed as follows:

[0075]

[0076] Where: Ω L Ω ESS Ω SVC Ω S These are respectively the line set, ESS node set, SVC node set, and typical scenario set after data-driven processing; K T K L K E K S These are the substation power supply cost coefficient, network loss coefficient, ESS dispatch coefficient, and SVC dispatch coefficient, respectively; P T P l L P i,ESS Q i,SVCThe active power output of the substation, the active power loss of the line, the charging and discharging power of the ESS node, and the reactive power output of the SVC node are all included, where P i,ESS The positive and negative signs represent the charging and discharging state of the ESS, with positive indicating discharging and negative indicating charging; π(S) is the probability corresponding to a typical scenario obtained through data-driven analysis.

[0077] Secondly, the constraints on the safe operation of the power distribution network include:

[0078] a) Current constraint:

[0079]

[0080]

[0081]

[0082]

[0083]

[0084] In the formula: P mn,S and Q mn,S These represent the active and reactive power flowing through line mn in scenario S, respectively. and These represent the active and reactive power losses on line mn under scenario S, respectively, and R. mn and X mn These represent the resistance and reactance of line mn, respectively, U m,S U n,S Let m and n be the voltage amplitudes at nodes m and n respectively, in scenario S. Let m be the active power injected into the distributed photovoltaic system in scenario S. These represent the active power injected into node m of the distributed energy storage system in the substation. These represent the reactive power injected into node m from the substation, distributed photovoltaic system, and static var compensator, respectively. The active and reactive power of the load at node m are P, respectively. km,S Q km,S These represent the active and reactive power (P) injected into node m on line km under scenario S. mn,S Q mn,S These represent the active and reactive power flowing out of node m from line mn under scenario S. This is a set of routes.

[0085] b) Safety operation constraints, including line transmission current constraints, line transmission power constraints, and node voltage constraints:

[0086]

[0087] V imin ≤V i,S ≤V i max

[0088] In the formula: P mn,S Q mn,S These represent the active and reactive power transmitted by line mn under scenario S, respectively. mn,max V represents the maximum current carrying capacity of line mn. i,S Let V be the voltage amplitude at node i in scenario S. i max V i min These are the specified upper and lower limits of the node voltage. This represents the upper limit of the current amplitude flowing through line mn.

[0089] c) Substation injected power constraints:

[0090]

[0091] In the formula: P T Q T To contribute both active and reactive power to the substation These are the upper and lower limits of the active and reactive power output of the substation.

[0092] d) Operating constraints of photovoltaic units:

[0093]

[0094] In the formula: P PV,S For the active power output of distributed photovoltaic power in scenario S, Q PV S PV These refer to the reactive power output and capacity of distributed photovoltaic systems.

[0095] e) ESS operational constraints:

[0096]

[0097]

[0098]

[0099] In the formula: P i,ESS , The charging and discharging power of the ESS, and its upper and lower limits. The ESS charge level at adjacent times. η represents the upper and lower limits of the battery capacity. loss η ESS These are the power loss coefficient and the charging / discharging efficiency, respectively.

[0100] f) SVC constraint:

[0101]

[0102] In the formula: Q i,SVC , The reactive power compensated by SVC, and the upper and lower limits of reactive power compensation.

[0103] g) Flexible load constraints:

[0104] (1-K P )*P D0 ≤P D ≤(1+K P )*P D0

[0105] (1-K Q )*Q D0 ≤Q D ≤(1+K Q )*Q D0

[0106] In the formula: P D0 Q D0 To calculate the active and reactive power of the load according to step (2), K P K Q P represents the proportion of active and reactive flexible loads. D Q D The active and reactive power of the load after flexible load control.

[0107] (4) The uncertainty of the stochastic optimization scheduling model in step (3) is described by fusing a fuzzy neural network, and the neural network parameters are obtained by training it with historical data:

[0108] First, the forward propagation process of the fuzzy neural network learning historical data feature information of the stochastic optimization scheduling model includes a data input layer, a membership function calculation layer, a rule generation layer, a normalization layer, and an output layer.

[0109] a) Data Input Layer: The number of neurons in this layer is equal to the feature dimension of the input data; that is, when the feature dimension of the data is n, the number of neurons in the input layer is n. The input data includes the predicted photovoltaic active power P. PV,pre The predicted active and reactive power P of the load D,pre Q D,pre and its prediction error ξ in various typical scenarios PV ξ DP ξ DQ The input variables contain uncertainty.

[0110] b) Membership Function Calculation Layer: Each neuron represents a fuzzy variable. The probability that each neuron in the input layer belongs to the fuzzy variable is calculated using the membership function. The number of neurons is the number of possible fuzzy conditions formed by the input variable. All neurons form a fuzzy set. This layer is used to describe the probability of uncertain input variables. Since the range of uncertain variable changes in the stochastic optimization scheduling model proposed in step (3) is the set of all typical scenarios, the number of fuzzy classifications of the input variables using this layer is the number of typical scenarios, and the number of nodes in this layer is the number of typical scenarios. The membership function is a Gaussian function, as shown below:

[0111]

[0112] In the formula: u ij x is the output of the neuron. i For input data, c ij b is the center point of the membership function. ij is the width vector of the membership function, and m is the number of fuzzy classifications of the input, which in this paper is the number of typical scenarios.

[0113] c) Rule generation layer: Each neuron represents a fuzzy rule. The output of this layer is the applicability of each rule, calculated using the following formula:

[0114] α j =Πu ij j = 1, 2, ..., m

[0115] d) Normalization layer: The fitness of each rule in the rule generation layer is normalized using the following formula:

[0116]

[0117] e) The output layer performs weighted processing to obtain the final output. The calculation formula is as follows:

[0118]

[0119] In the formula: y i w is the output variable. ij is the corresponding weight, and k is the number of output variables.

[0120] Then, the backpropagation process of the fuzzy neural network learning the historical data feature information of the stochastic optimization scheduling model uses the negative gradient descent algorithm to optimize the center point, width vector of the membership function, and connection weights of the output layer.

[0121] a) Calculate the negative gradient of the error E with respect to the output layer, as follows:

[0122]

[0123] b) Calculate the negative gradient of error E with respect to the normalization layer, rule generation layer, and membership function layer in sequence. The negative gradient of error E with respect to the membership function layer can be obtained as follows:

[0124]

[0125] c) Error E versus weight w ij Center point c ij and width vector b ij The negative gradient is:

[0126]

[0127] d) The final parameter adjustment algorithm is as follows:

[0128]

[0129] In the formula: η, ξ, and ψ are the learning rates of the corresponding parameters, and r is the number of learning rounds of the neural network.

[0130] (5) Input information reflecting the real-time status of the distribution network, namely, real-time power prediction data P PV,pre P D,pre Q D,pre and the typical scenario error ξ obtained through data-driven methods PV ξ DP ξ DQ The neural network trained in step (4) is used to quickly output the estimated values ​​of the scheduling variables. Make preliminary decisions on scheduling strategies.

[0131] (6) The mature nonlinear solver IPOPT is adopted, and the pre-decision value is used as the initial point for the solver to optimize. The scheduling variables are accelerated to ensure the accuracy of the results, and the real-time solution of the stochastic optimization scheduling model proposed in step (3) is realized.

[0132] To demonstrate the effectiveness of the proposed scenario-based stochastic power dispatching method for active distribution networks, typical scenarios with 5, 10, 20, and 30 scenarios were extracted from historical data. The objective function values ​​at these 20 time points are shown in Table 1. Table 1 shows that when the number of typical scenarios is 10, 20, and 30, the objective function values ​​at these 20 time points are generally stable, with differences within ±100. It can be predicted that further increasing the number of typical scenarios will also stabilize the objective function values ​​within a stable range, indicating that the typical scenarios constructed from historical data by the proposed method can effectively characterize uncertain variables.

[0133] Table 1. Impact of the Number of Typical Scenarios on the Objective Function

[0134]

[0135]

[0136] To verify the performance of the proposed fuzzy neural network pre-decision solving method, the original problem was solved directly using the solver as a comparison. Furthermore, to reduce the random impact of computer performance on computation time, 20 time segments were solved, and the total solution time was obtained. Simulation results under different typical scenarios are shown in Table 2. In Table 2, T0 is the time required to solve directly using the solver, T1 is the time required for the second-stage correction of the solver, T2 is the fuzzy neural network pre-decision time, and T... 总 The total time required to integrate the fuzzy neural network solution method is given by α, which represents the speedup ratio of the proposed method compared to the direct solution.

[0137] Table 2 Comparison of computation time under different typical scenarios

[0138]

[0139] As shown in Table 2, directly solving the proposed model using the solver takes 249.023 seconds per solution for 30 scenarios, with a real-time scheduling timescale of 5 minutes, which barely meets the requirements. However, the time margin is small, and considering the time required for power prediction and communication latency, its real-time performance needs further improvement. The solution method incorporating fuzzy neural network pre-decision takes 159.827 seconds per solution for 30 scenarios, which better meets the requirements for real-time scheduling. Furthermore, the proposed method achieves an acceleration of over 30% for all typical scenarios. Especially for the typical scenario count of 30, the proposed method significantly reduces the solution time, shortening the total time for solving 20 power scheduling operations by 1760 seconds, effectively improving the model's solution speed and better meeting the real-time requirements of the current environment. Furthermore, as the number of typical scenarios increases, the complexity of the original optimization problem increases, which leads to a significant increase in the time required to solve the problem directly using the solver. However, the time for the first-stage pre-decision using the fuzzy neural network remains basically unchanged. Although the time for the second-stage correction using the solver also increases, the overall speedup ratio is slightly improved. This indicates that the proposed method has a more significant effect on improving the efficiency of solving complex systems.

[0140] Figure 3 Taking the energy storage system installed at node 12 as an example, a scatter plot is shown showing the relationship between the pre-decision value and the true value of the fuzzy neural network in the test set with 20 typical scenarios and 100 samples. Figure 3 The horizontal axis represents the true value, and the vertical axis represents the pre-decision value. It can be seen that the data points are roughly located near a straight line with a slope of 1, indicating that the pre-decision values ​​are close to the true values, thus demonstrating that the fuzzy neural network has good pre-decision performance.

[0141] In summary, this invention proposes a real-time stochastic scheduling method for active distribution networks that integrates fuzzy neural network pre-decision making, enabling real-time optimized scheduling of active distribution networks with uncertainties. The effectiveness of the proposed method is verified through case analysis.

[0142] Based on the same technical concept as the method embodiments, the present invention also provides an active distribution network real-time stochastic dispatching system that integrates fuzzy neural network pre-decision making, comprising:

[0143] The power flow model re-derivation module is used to re-derive the power flow model of the active distribution network based on the Euler equations, including network loss equations, active power balance equations, reactive power balance equations, voltage drop equations, and line current constraints. The derived model is equivalent to the original power flow model.

[0144] The uncertainty description module treats the power prediction values ​​of photovoltaic output and load as deterministic variables and the prediction error as an uncertain variable. It decomposes the actual power of photovoltaic output and load into two parts: power prediction value and prediction error. The uncertainty of photovoltaic output and the randomness of load are described as the uncertainty of the corresponding prediction error, and the probability distribution of the uncertain variable is obtained through a data-driven method.

[0145] The scheduling model construction module is used to comprehensively consider the power supply cost of the upstream substation, the distribution network loss, the scheduling cost of the distributed energy storage system (ESS), and the scheduling cost of the static var compensator (SVC). It constructs the objective function of the active distribution network stochastic optimization power scheduling model with the goal of minimizing costs. Its constraints include power flow constraints, safe operation constraints, substation injected power constraints, photovoltaic unit operation constraints, ESS operation constraints, SVC constraints, and flexible load constraints.

[0146] The fuzzy neural network description and training module is used to describe the uncertainty of the stochastic optimal power dispatch model of the active distribution network by integrating fuzzy neural networks, and to train the fuzzy neural network using historical data. The forward propagation process of the fuzzy neural network includes a data input layer, a membership function calculation layer, a rule generation layer, a normalization layer, and an output layer. The back propagation process uses the negative gradient descent algorithm to optimize the center point, width vector of the membership function, and connection weights of the output layer to obtain the parameters of the fuzzy neural network.

[0147] The pre-decision acquisition module is used to input real-time status data of the active distribution network, including real-time power prediction values ​​and data-driven prediction errors, into a trained fuzzy neural network, and use the fuzzy neural network to quickly output pre-decision of scheduling strategy.

[0148] The real-time random scheduling strategy acquisition module is used to take the pre-decision value as the initial value for the solver to optimize, and then use the solver to accurately solve the active distribution network random optimization power scheduling model in the scheduling model construction module to obtain the real-time random scheduling strategy of the active distribution network.

[0149] It should be understood that the active distribution network real-time random dispatching system with fuzzy neural network pre-decision in the embodiments of the present invention can realize all the technical solutions in the above method embodiments. The functions of each functional module can be specifically implemented according to the methods in the above method embodiments. The specific implementation process can be referred to the relevant descriptions in the above embodiments, which will not be repeated here.

[0150] The present invention also provides a computer device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, wherein when the programs are executed by the processors, they implement the steps of the active distribution network real-time random scheduling method for pre-decision fusion of fuzzy neural networks as described above.

[0151] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the active distribution network real-time random scheduling method for pre-decision fusion of fuzzy neural networks as described above.

[0152] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus (systems), computer devices, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention 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.

[0153] This invention is described with reference to a flowchart of a method according to embodiments of the invention. It should be understood that each step in the flowchart and combinations thereof 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 device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing device, generate instructions for implementing the process. Figure 1 A device for a function specified in one or more processes.

[0154] 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 The function specified in one or more processes.

[0155] 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 Steps of a specified function in one or more processes.

Claims

1. A real-time stochastic dispatching method for active distribution networks integrating fuzzy neural network pre-decision analysis, characterized in that, Includes the following steps: Step 1: Based on the Euler equations, re-derive the power flow model of the active distribution network, including the network loss equation, active power balance equation, reactive power balance equation, voltage drop equation, and line current constraints. The derived model is equivalent to the original power flow model. Step 2: Treat the predicted power values ​​of photovoltaic output and load as deterministic variables, and the prediction error as an uncertain variable. Decompose the actual power of photovoltaic output and load into two parts: predicted power value and prediction error. The uncertainty of photovoltaic output and the randomness of load are described as the uncertainty of the corresponding prediction error. The probability distribution of the uncertain variable is obtained through a data-driven method. Step 3: Taking into account the power supply cost of the upstream substation, the distribution network loss, the dispatch cost of the distributed energy storage system (ESS), and the dispatch cost of the static var compensator (SVC), construct the objective function of the active distribution network stochastic optimization power dispatch model with the goal of minimizing cost. Its constraints include power flow constraints, safe operation constraints, substation injected power constraints, photovoltaic unit operation constraints, ESS operation constraints, SVC constraints, and flexible load constraints. Step 4: The uncertainty of the active distribution network stochastic optimization power dispatch model is described by integrating a fuzzy neural network, and historical data is used to train the fuzzy neural network. The forward propagation process of the fuzzy neural network includes a data input layer, a membership function calculation layer, a rule generation layer, a normalization layer, and an output layer. The back propagation process uses the negative gradient descent algorithm to optimize the center point, width vector of the membership function, and connection weights of the output layer to obtain the parameters of the fuzzy neural network. Step 5: Input the real-time status data of the active distribution network, including real-time power prediction values ​​and data-driven prediction errors, into the trained fuzzy neural network, and use the fuzzy neural network to quickly output the pre-decision of the scheduling strategy. Step 6: Use the pre-decision value as the initial value for the solver to find the optimal value, and then use the solver to accurately solve the active distribution network stochastic power dispatch model in Step 3 to obtain the real-time stochastic dispatch strategy of the active distribution network.

2. The method according to claim 1, characterized in that, Step 1 specifically includes: For the network loss equation, the original network loss function formula is expressed as: In the formula: P mn and Q mn These represent the active and reactive power flowing through line m and n, respectively. and These represent the active and reactive power losses on line mn, respectively, and I. mn R is the current flowing through line mn. mn and X mn These represent the resistance and reactance of line mn, respectively, U m Let be the voltage amplitude at node m; For the active power balance equation and the reactive power balance equation, the active power balance and reactive power balance of node m are expressed as follows: In the formula: These represent the active power injected into node m from three types of distributed resources: substations, distributed photovoltaics, and distributed energy storage. These represent the reactive power injected into node m from the substation, distributed photovoltaic system, and static var compensator, respectively. The active and reactive power of the load at node m are P, respectively. km Q km These represent the active and reactive power injected into node m at line km, respectively, in P. mn Q mn These represent the active and reactive power flowing out of node m on line mn, respectively. For the set of routes; For the voltage drop equation, the initial definition of voltage drop is: In the formula: Z mn The impedance of the line is mn, i.e., Z mn =R mn +jX mn V m V n Let V be the complex voltage at nodes m and n. m =|V m |∠θ m V n =|V n |∠θ n S mn S represents the line apparent power. mn =P mn +jQ mn ; and S mn and V m The conjugate complex number; using Euler's formula to rewrite the voltage phase angle difference and through mathematical derivation of the real and imaginary parts, the voltage drop formula after voltage phase angle relaxation is obtained: Line current constraints: In the formula: This represents the upper limit of the current amplitude flowing through line mn.

3. The method according to claim 1, characterized in that, In step 2, the probability distribution of the uncertain variable is obtained through a data-driven approach, including: The k-means clustering algorithm and the synchronous back-substitution elimination technique based on probabilistic distance are used to reduce the scene of the prediction data of similar days of power prediction, so as to obtain the typical scene of prediction error.

4. The method according to claim 1, characterized in that, In step 3, the objective function of the active distribution network stochastic optimization power dispatch model is as follows: Where: Ω L Ω ESS Ω SVC Ω S These are respectively the line set, ESS node set, SVC node set, and typical scenario set after data-driven processing; K T K L K E K S These are the substation power supply cost coefficient, network loss coefficient, ESS dispatch coefficient, and SVC dispatch coefficient, respectively; P T P l L P i,ESS Q i,SVC The active power output of the substation, the active power loss of the line, the charging and discharging power of the ESS node, and the reactive power output of the SVC node are all included, where P i,ESS The positive and negative signs represent the charging and discharging state of the ESS, with positive indicating discharging and negative indicating charging; π(S) is the probability corresponding to a typical scenario obtained through data-driven analysis.

5. The method according to claim 1, characterized in that, The power flow constraint is expressed as: In the formula: P mn,S and Q mn,S These represent the active and reactive power flowing through line mn in scenario S, respectively. and These represent the active and reactive power losses on line mn under scenario S, respectively, and R. mn and X mn These represent the resistance and reactance of line mn, respectively, U m,S U n,S Let m and n be the voltage amplitudes at nodes m and n respectively, in scenario S. Let m be the active power injected into the distributed photovoltaic system in scenario S. These represent the active power injected into node m of the distributed energy storage system in the substation. These represent the reactive power injected into node m from the substation, distributed photovoltaic system, and static var compensator, respectively. The active and reactive power of the load at node m are P, respectively. km,S Q km,S These represent the active and reactive power (P) injected into node m on line km under scenario S. mn,S Q mn,S These represent the active and reactive power flowing out of node m from line mn under scenario S. For the set of routes; The safety operation constraints include line transmission current constraints, line transmission power constraints, and node voltage constraints: In i min ≤V i,S ≤V i max In the formula: P mn,S Q mn,S These represent the active and reactive power transmitted by line mn under scenario S, respectively. mn,max V represents the maximum current carrying capacity of line mn. i max V i min These are the specified upper and lower limits of the node voltage. This represents the upper limit of the current amplitude flowing through line mn; The power injection constraint of the substation is expressed as follows: In the formula: P T Q T Contribute both active and reactive power to the substation. The upper and lower limits of active and reactive power output of the substation; The operating constraints of the photovoltaic unit are expressed as follows: In the formula: P PV,S For the active power output of distributed photovoltaic power in scenario S, Q PV S PV These refer to the reactive power output and capacity of the photovoltaic unit, respectively. The ESS runtime constraint is represented as follows: In the formula: P i,ESS , The charging and discharging power of the ESS, and its upper and lower limits. The ESS charge level at adjacent times. η represents the upper and lower limits of the battery capacity. loss η ESS These are the power loss coefficient and the charging / discharging efficiency, respectively. The SVC constraint is expressed as follows: In the formula: Q i,SVC , The reactive power compensation provided by SVC, and the upper and lower limits of reactive power compensation; The flexible load constraint is expressed as follows: (1-K P )*P D0 ≤P D ≤(1+K P )*P D0 (1-K Q )*Q D0 ≤Q D ≤(1+K Q )*Q D0 In the formula: P D0 Q D0 K represents the active and reactive power of the load calculated in step 2. P K Q P represents the proportion of active and reactive flexible loads. D Q D The active and reactive power of the load after flexible load control.

6. The method according to claim 1, characterized in that, The forward propagation process of the fuzzy neural network includes: a) Data Input Layer: The number of neurons in this layer is equal to the feature dimension of the input data. That is, when the feature dimension of the data is n, the number of neurons in the input layer is n. The input data includes the predicted photovoltaic active power P. PV,pre The predicted active and reactive power P of the load D,pre and Q D,pre and its prediction error ξ in various typical scenarios PV ξ DP ξ DQ The input variables contain uncertainty; b) Membership Function Calculation Layer: Each neuron represents a fuzzy variable. The probability that each neuron in the input layer belongs to the fuzzy variable is calculated using a membership function. The number of neurons is the number of possible fuzzy conditions formed by the input variables. All neurons form a fuzzy set, and the number of nodes in this layer is the number of typical scenarios. The membership function used is a Gaussian function, as shown below: In the formula: u ij x is the output of the neuron. i For input data, c ij b is the center point of the membership function. ij is the width vector of the membership function, and m is the number of fuzzy classifications of the input, i.e., the number of typical scenarios; c) Rule generation layer: Each neuron represents a fuzzy rule. The output of this layer is the applicability of each rule, calculated using the following formula: a j =Where? ij ,j=1,2,...,m d) Normalization layer: The fitness of each rule in the rule generation layer is normalized using the following formula: e) The output layer performs weighted processing to obtain the final output. The calculation formula is as follows: In the formula: y i w is the output variable. ij is the corresponding weight, and k is the number of output variables.

7. The method according to claim 6, characterized in that, The backpropagation process of the fuzzy neural network includes: a) Calculate the negative gradient of the error E with respect to the output layer, as follows: b) Calculate the negative gradient of error E with respect to the normalization layer, rule generation layer, and membership function layer in sequence to obtain the negative gradient of error E with respect to the membership function layer: c) Error E versus weight w ij Center point c ij and width vector b ij The negative gradient is: d) The final parameter adjustment algorithm is as follows: In the formula: η, ξ, and ψ are the learning rates of the corresponding parameters, and r is the number of learning rounds of the neural network.

8. A real-time stochastic dispatching system for active distribution networks that integrates fuzzy neural network pre-decision making, characterized in that, include: The power flow model re-derivation module is used to re-derive the power flow model of the active distribution network based on the Euler equations, including network loss equations, active power balance equations, reactive power balance equations, voltage drop equations, and line current constraints. The derived model is equivalent to the original power flow model. The uncertainty description module treats the power prediction values ​​of photovoltaic output and load as deterministic variables and the prediction error as an uncertain variable. It decomposes the actual power of photovoltaic output and load into two parts: power prediction value and prediction error. The uncertainty of photovoltaic output and the randomness of load are described as the uncertainty of the corresponding prediction error, and the probability distribution of the uncertain variable is obtained through a data-driven method. The scheduling model construction module is used to comprehensively consider the power supply cost of the upstream substation, the distribution network loss, the scheduling cost of the distributed energy storage system (ESS), and the scheduling cost of the static var compensator (SVC). It constructs the objective function of the active distribution network stochastic optimization power scheduling model with the goal of minimizing costs. Its constraints include power flow constraints, safe operation constraints, substation injected power constraints, photovoltaic unit operation constraints, ESS operation constraints, SVC constraints, and flexible load constraints. The fuzzy neural network description and training module is used to describe the uncertainty of the stochastic optimal power dispatch model of the active distribution network by integrating fuzzy neural networks, and to train the fuzzy neural network using historical data. The forward propagation process of the fuzzy neural network includes a data input layer, a membership function calculation layer, a rule generation layer, a normalization layer, and an output layer. The back propagation process uses the negative gradient descent algorithm to optimize the center point, width vector of the membership function, and connection weights of the output layer to obtain the parameters of the fuzzy neural network. The pre-decision acquisition module is used to input real-time status data of the active distribution network, including real-time power prediction values ​​and data-driven prediction errors, into a trained fuzzy neural network, and use the fuzzy neural network to quickly output pre-decision of scheduling strategy. The real-time random scheduling strategy acquisition module is used to take the pre-decision value as the initial value for the solver to optimize, and then use the solver to accurately solve the active distribution network random optimization power scheduling model in the scheduling model construction module to obtain the real-time random scheduling strategy of the active distribution network.

9. A computer device, characterized in that, include: One or more processors; Memory; And one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, wherein when the programs are executed by the processors, they implement the steps of the active distribution network real-time random scheduling method for fusion fuzzy neural network pre-decision as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the active distribution network real-time random scheduling method for pre-decision fusion of fuzzy neural networks as described in any one of claims 1-7.

Images