Artificial Intelligence-Based Digital Simulation Decision Analysis Method and System for Distribution Network

Abstract

The present invention discloses a digital simulation decision analysis method and system for a distribution network based on artificial intelligence, including obtaining an initial loss data set according to the topological structure of the distribution network and a standardized data set; establishing an impedance loss model according to the initial loss data set and the impedance values of each branch to judge the loss type of the branch; performing loss pattern recognition on the initial loss data set and the loss type of the branch to obtain a loss spatial distribution pattern; inputting the current impedance value of the branch into a power flow prediction model to obtain a predicted active power value; establishing a load distribution optimization model according to the predicted active power value and the loss spatial distribution pattern to obtain an optimal load distribution scheme. The present invention performs dynamic simulation of the distribution network through digital simulation, combines artificial intelligence for real-time decision optimization, and performs verification feedback correction through digital simulation to form a closed loop of perception, decision-making, and verification, reducing the network loss of the distribution network and improving the safety and stability of the operation of the distribution network.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network decision - making analysis, and particularly to a distribution network digital simulation decision - making analysis method and system based on artificial intelligence. Background Art

[0002] With the rapid development of smart grids and the large - scale access of distributed energy, the load volatility and operation complexity of distribution networks have increased significantly, and the load distribution optimization methods for distribution networks are facing severe challenges. The existing distribution network optimization technologies mainly include offline optimization based on static power flow, rule - driven load transfer strategies, and single - model prediction methods, etc. Although these methods can all achieve load distribution, there are still certain limitations. On the one hand, the offline optimization based on static power flow constructs a fixed topology model through historical data and uses heuristic algorithms such as genetic algorithms to generate load distribution schemes. However, this method has a long optimization cycle and cannot respond to load mutations in real - time. Therefore, the scheme has hysteresis and is not suitable for the optimization of distribution networks in dynamic scenarios. On the other hand, most of the rule - driven load transfer strategies rely on manual experience to set load rate thresholds and transfer the load of overloaded branches through switch operations, thus lacking a global optimization perspective and being prone to falling into the trap of local optimization, leading to a chain problem of "overload - transfer - new overload". In addition, in single - model prediction, load prediction and optimization decision - making are usually separated, resulting in the cumulative error of prediction errors and the lack of fusion of multi - dimensional data, there is a problem of data islands, leading to the optimization scheme being prone to deviate from the actual working conditions.

[0003] Aiming at the problems such as hysteresis verification, local optimization trap, and data island problems existing in the prior art, there is an urgent need for a data - driven, closed - loop dynamic, and branch - level refined distribution network optimization scheme. Summary of the Invention

[0004] To solve the above - mentioned technical problems, the present invention provides a distribution network digital simulation decision - making analysis method and system based on artificial intelligence, so as to solve the problems of high hysteresis, weak global optimization ability, low resource utilization rate, etc. in the prior art, and achieve the technical effects of reducing the network loss of the distribution network and improving the load balance of the distribution cabinet.

[0005] In the first aspect, the present invention provides a distribution network digital simulation decision - making analysis method based on artificial intelligence, and the method includes:

[0006] Obtain the power data of the distribution network, and perform data pre - processing on the power data to generate a standardized data set, where the standardized data set includes the branch ID, impedance value, active power, and reactive power of each branch of the distribution network;

[0007] According to the topological structure of the distribution network and the standardized data set, an initial loss data set is obtained through power flow calculation. The initial loss data set includes the loss values, voltage deviations, and power factors of each branch.

[0008] According to the initial loss data set and the impedance values of each branch, an impedance loss model is established, and based on the current impedance values of each branch and the impedance loss model, the loss types of each branch of the distribution network are judged.

[0009] An initial loss data set and the loss types of each branch are used to perform loss pattern recognition by using a preset algorithm to obtain the loss spatial distribution pattern of each branch.

[0010] The current impedance values of each branch are input into a pre-constructed power flow prediction model to obtain the predicted active power values of each branch.

[0011] Based on the predicted active power values of each branch and the loss spatial distribution pattern, a load distribution optimization model is established, and the particle swarm optimization algorithm is used to solve it to obtain the optimal load distribution scheme.

[0012] Further, the step of obtaining the initial loss data set through power flow calculation according to the topological structure of the distribution network and the standardized data set includes:

[0013] According to the topological structure of the distribution network, a node-branch incidence matrix is constructed.

[0014] Based on the node-branch incidence matrix and the standardized data set, an improved forward-backward sweep method is used for power flow calculation to obtain branch currents and node voltages.

[0015] According to the branch current and the corresponding branch resistance, the loss value is calculated.

[0016] According to the node voltage and the nominal voltage of the distribution network, the voltage deviation is calculated.

[0017] According to the active power and reactive power of each branch, the power factor is calculated.

[0018] According to the branch ID, the loss value, the voltage deviation, and the power factor of each branch, an initial loss data set is formed.

[0019] Further, the step of establishing an impedance loss model according to the initial loss data set and the impedance values of each branch, and judging the loss types of each branch of the distribution network based on the current impedance values of each branch and the impedance loss model includes:

[0020] A sliding time window is used to extract the time series characteristics of the impedance values of each branch to obtain an impedance feature matrix.

[0021] Use the least squares method to perform data fitting on the impedance characteristic matrix and the loss values in the initial loss dataset to obtain the impedance loss model for each branch;

[0022] Obtain the current impedance value of each branch, input it into the corresponding impedance loss model, obtain the first predicted loss value of each branch, and extract the loss sensitivity coefficient of each branch from the corresponding impedance loss model;

[0023] If the first predicted loss value is greater than the loss threshold, or if the loss sensitivity coefficient is greater than the coefficient threshold, mark the corresponding branch as a high-loss branch and output the high-loss branch mark;

[0024] Among them, the impedance loss model is expressed by the following formula:

[0025]

[0026] In the formula, P loss represents the first predicted loss value, a represents the loss sensitivity coefficient, R represents the impedance value, and b represents the correction coefficient.

[0027] Further, the step of using a preset algorithm to perform loss pattern recognition on the initial loss dataset and the loss type of each branch to obtain the loss spatial distribution pattern of each branch includes:

[0028] Use the PCA dimensionality reduction algorithm to perform dimensionality reduction processing on the initial loss dataset to obtain the initial loss dataset after dimensionality reduction processing;

[0029] Use the K-means clustering algorithm to perform clustering analysis on the initial loss dataset after dimensionality reduction processing to obtain the load type of each branch, and the load type includes industrial load, commercial load, residential load, and mixed load;

[0030] Determine the loss spatial distribution pattern of each branch according to the load type and the loss type of each branch.

[0031] Further, the step of inputting the current impedance value of each branch into a pre-constructed power flow prediction model to obtain the power flow distribution prediction result includes:

[0032] Use a sliding time window to extract the time series characteristics of the current impedance value of each branch within a preset time period to obtain impedance time series characteristics;

[0033] Perform static feature extraction on the initial loss dataset to obtain network state reference features;

[0034] Input the impedance time series characteristics and the network state reference characteristics into a pre-constructed power flow prediction model to obtain the predicted results of the power flow distribution of each branch. The power flow prediction model is constructed using a long short-term memory neural network model.

[0035] Further, the step of establishing a load distribution optimization model according to the predicted active power values of each branch and the loss spatial distribution pattern includes:

[0036] Calculate the second loss prediction value of each branch according to the predicted active power value of each branch;

[0037] Determine the loss weight according to the loss spatial distribution pattern of each branch;

[0038] Perform a weighted sum of the second loss prediction values according to the loss weight to obtain the network loss of the distribution network;

[0039] Calculate the load rate standard deviation according to the branch load rate of each branch;

[0040] Construct a load distribution optimization model with the minimization of the network loss and the load rate standard deviation as the objective function and the voltage deviation and the branch load rate as the constraint conditions.

[0041] Further, the step of determining the loss weight according to the loss spatial distribution pattern of each branch includes:

[0042] Set the corresponding initial weight according to the load type of each branch;

[0043] Modify the initial weight according to the loss type of each branch to obtain the loss weight.

[0044] Further, after the step of obtaining the optimal load distribution plan, it further includes:

[0045] Obtain the real-time monitoring data of the distribution network, and judge whether to trigger path transfer according to the real-time monitoring data. The real-time monitoring data includes the real-time branch load rate and the real-time voltage;

[0046] If path transfer is triggered, determine all transfer paths of the branch to be transferred according to the topological structure of the distribution network;

[0047] Use power flow calculation to obtain the transferred load and the loss reduction corresponding to each transfer path, and select the transfer path corresponding to the maximum loss reduction as the optimal transfer path;

[0048] Modify the optimal load distribution plan according to the transferred load and the optimal transfer path to obtain the modified optimal load distribution plan.

[0049] Further, the step of determining whether to trigger path transfer according to the real-time monitoring data includes:

[0050] According to the real-time branch load rate, using a sliding time window mechanism, calculate the load change rate between adjacent windows. If the load change rate is greater than the change rate threshold, trigger path transfer;

[0051] Or,

[0052] According to the real-time voltage obtained from each sampling, using a sliding time window mechanism, calculate the voltage deviation within the current window to obtain multiple voltage deviation values;

[0053] Compare each voltage deviation value with the deviation threshold. If the number of voltage deviation values greater than the deviation threshold is greater than the quantity threshold, trigger path transfer.

[0054] In a second aspect, the present invention provides a digital simulation decision analysis system for a distribution network based on artificial intelligence. The system includes:

[0055] A data processing module, configured to obtain power data of the distribution network, perform data preprocessing on the power data, and generate a standardized data set. The standardized data set includes the branch ID, impedance value, active power, and reactive power of each branch of the distribution network;

[0056] According to the topological structure of the distribution network and the standardized data set, obtain an initial loss data set through power flow calculation. The initial loss data set includes the loss value, voltage deviation, and power factor of each branch;

[0057] A loss analysis module, configured to establish an impedance loss model according to the initial loss data set and the impedance value of each branch, and judge the loss type of each branch of the distribution network according to the current impedance value of each branch and the impedance loss model;

[0058] Use a preset algorithm to perform loss mode recognition on the initial loss data set and the loss type of each branch to obtain the loss spatial distribution mode of each branch;

[0059] A power flow prediction module, configured to input the current impedance value of each branch into a pre-constructed power flow prediction model to obtain the predicted active power value of each branch;

[0060] A load distribution module, configured to establish a load distribution optimization model according to the predicted active power value of each branch and the loss spatial distribution mode, and use a particle swarm algorithm to solve it to obtain an optimal load distribution plan.

[0061] The present invention provides a method and system for digital simulation decision analysis of a distribution network based on artificial intelligence. By establishing an impedance loss model and quantifying the sensitivity of impedance to loss, the present invention can accurately locate high-loss branches. Through a power prediction model that fuses multi-variable features, the accuracy of power flow prediction can be improved, providing data support for forward-looking decisions. And through a multi-objective particle swarm algorithm and a dynamic weight adjustment mechanism, the optimal allocation of the load of the distribution network can be achieved. The present invention can effectively reduce the network loss of the distribution network and improve the load balance degree, thereby ensuring the safety and stability of the operation of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 is a schematic flow chart of a method for digital simulation decision analysis of a distribution network based on artificial intelligence in an embodiment of the present invention;

[0063] Figure 2 is a schematic structural diagram of a system for digital simulation decision analysis of a distribution network based on artificial intelligence in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0064] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0065] Please refer to Figure 1 , a method for digital simulation decision analysis of a distribution network based on artificial intelligence proposed in the first embodiment of the present invention, which includes steps S10 to S60:

[0066] Step S10, obtain the power data of the distribution network, and perform data preprocessing on the power data to generate a standardized data set, where the standardized data set includes the branch ID, impedance value, active power, and reactive power of each branch of the distribution network;

[0067] Step S20, according to the topological structure of the distribution network and the standardized data set, obtain an initial loss data set through power flow calculation, where the initial loss data set includes the loss value, voltage deviation, and power factor of each branch;

[0068] Step S30, establish an impedance loss model according to the initial loss data set and the impedance value of each branch, and judge the loss type of each branch of the distribution network according to the current impedance value of each branch and the impedance loss model;

[0069] Step S40, using a preset algorithm to perform loss pattern recognition on the initial loss data set and the loss type of each branch to obtain a loss spatial distribution pattern of each branch;

[0070] Step S50, inputting the current impedance value of each branch into a pre-built power flow prediction model to obtain the active power prediction value of each branch;

[0071] Step S60, establishing a load distribution optimization model according to the active power prediction value of each branch and the loss spatial distribution pattern, and using a particle swarm algorithm to solve it to obtain an optimal load distribution solution.

[0072] The present invention performs digital simulation on the ring network structure and power data of the distribution network, and performs flow analysis on the power and loss of the distribution network, thereby optimizing the load distribution plan in the distribution network. First, the power data of the distribution network is obtained to generate a standardized data set. In this embodiment, the sources of power data can be divided into three types. One is the impedance value obtained by the impedance monitoring device at the feeder segmentation point, and its data format is: (timestamp, branch ID, impedance value Ω / km); the second is the historical power circulation data from the data acquisition and monitoring control system SCADA, including branch active power, reactive power, etc.; the third is the monitoring data from the smart meter, load point ID, active power and reactive power, etc. Then, these power data are preprocessed, including outlier filtering, missing value filling and time alignment. For example, the Z-score is calculated for the power data, and outliers with absolute values ​​greater than 3 are removed. When the impedance data exceeds the historical mean ±50%, it is marked as invalid. For data with consecutive missing time points of ≤3, linear interpolation is used to fill in the gaps. In addition, since the time granularity of smart meters (15 minutes) may be longer than that of impedance monitoring devices and SCADA systems (5 minutes), it is also necessary to time-align the smart meter data through cubic spline interpolation, and organize the preprocessed data to obtain a standardized data set. The power data in the data set includes at least the timestamp, branch ID, impedance value, active power and reactive power of each branch of the distribution network.

[0073] After obtaining the standardized data set, combined with the topological structure of the distribution network, the power flow simulation analysis of the initial loss of the distribution network can be realized. The specific analysis steps include:

[0074] According to the topological structure of the distribution network, a node-branch association matrix is ​​constructed;

[0075] According to the node-branch association matrix and the standardized data set, an improved forward-backward substitution method is used to perform power flow calculation to obtain branch current and node voltage;

[0076] Calculating a loss value according to the branch current and the corresponding branch resistance;

[0077] Calculate the voltage deviation based on the node voltage and the nominal voltage of the distribution network;

[0078] Calculate the power factor based on the active power and reactive power of each branch;

[0079] Form an initial loss data set based on the branch ID, the loss value, the voltage deviation, and the power factor of each branch.

[0080] In this embodiment, the topological structure of the distribution network includes node connection relationships and branch types. The branch types include feeders, transformers, distributed power nodes, etc. In view of the characteristics of the high resistance / reactance ratio of the distribution network, combined with the standardized data set, an improved forward-backward sweep method is used for power flow calculation. Specifically, according to the distribution network topology, the root node (feeder starting point), end nodes, and branch hierarchy relationships are marked, and the parameters are initialized. If there is a loop network (such as Node4 → Node5 → Node6 → Node4), the loop is temporarily disconnected and converted into a radial structure for iteration. The initialized parameters include setting the initial voltage value, and extracting the active power and reactive power of the nodes from the standardized data set. If the node contains a distributed power source (such as photovoltaic), it is processed according to the PQ / PV node type.

[0081] When performing power flow calculation, first calculate the branch circuit through the forward process, that is, push back from the end node to the root node to obtain the branch current. If there are sub-branches in the branch, accumulate the sub-branch currents to obtain the branch current. Then update the node voltage through the backward substitution process, that is, push forward from the root node to the end node to obtain the node voltage. If there is a loop network, perform loop correction, restore the disconnected loop, calculate the loop voltage difference, eliminate the deviation through the compensation current, and at the same time set the convergence conditions, such as the maximum voltage difference between adjacent nodes is less than the preset threshold, or the maximum number of iterations is reached, etc. Finally, obtain the voltage deviation based on the difference between the node voltage and the nominal voltage, calculate the branch power based on the node voltage and the branch current, and calculate the active power loss and reactive power loss based on the difference between the branch head power and the branch end power, and calculate the power factor based on the active power loss and the reactive power loss. Among them, when calculating the power loss, the derivation of the difference between the branch head power and the branch end power finally converts the active power loss into the product relationship between the square of the branch current and the resistance, and converts the reactive power loss into the product relationship between the square of the branch current and the reactance. The specific derivation process can refer to the conventional loss value calculation steps and will not be elaborated here.

[0082] The initial loss data set can be obtained through the above power flow calculation. The initial loss data set includes the branch ID extracted from the topological data, the loss value determined according to the active power loss, the voltage deviation, and the power factor.

[0083] In this embodiment, an improved forward-backward substitution method is adopted, combined with loop processing and distributed power support, which significantly improves the convergence and efficiency of power flow calculation in the distribution network. At the same time, it accurately outputs the initial loss dataset, providing reliable data input for subsequent decision-making analysis.

[0084] After obtaining the initial loss dataset, by combining the impedance of each branch and establishing an impedance loss model, the loss types of each branch are analyzed. The specific steps include:

[0085] Use a sliding time window to extract the time series characteristics of the impedance values of each branch to obtain an impedance feature matrix;

[0086] Use the least squares method to fit the impedance feature matrix and the loss values in the initial loss dataset to obtain the impedance loss model of each branch;

[0087] Obtain the current impedance value of each branch, input it into the corresponding impedance loss model to obtain the first predicted loss value of each branch, and extract the loss sensitivity coefficient of each branch from the corresponding impedance loss model;

[0088] If the first predicted loss value is greater than the loss threshold, or if the loss sensitivity coefficient is greater than the coefficient threshold, mark the corresponding branch as a high-loss branch and output the high-loss branch mark.

[0089] In this embodiment, the data used to establish the impedance loss model of each branch includes the impedance data in the standardized dataset and the loss data in the initial loss dataset. First, the impedance value is processed based on a sliding time window, the window length is set and features are extracted, including the impedance mean, variance, maximum value, and minimum value of each window. Then, based on the impedance value and the loss value, the least squares method is used for modeling:

[0090]

[0091] In the formula, P loss represents the first predicted loss value, a represents the loss sensitivity coefficient, R represents the impedance value, and b represents the correction coefficient.

[0092] It should be noted that in the initial model, R selects the impedance mean, and P loss selects the initial loss value. Through data fitting, with the goodness of fit greater than 0.9 and the residual standard deviation less than 5% as the verification conditions, the loss sensitivity coefficient and the correction coefficient are determined, and finally the impedance loss model is obtained.

[0093] Input the current impedance values of each branch into the corresponding impedance loss model to obtain the first predicted loss values of each branch. Then, based on the impedance loss model and the predicted loss values, determine the loss type of each branch. Here, the loss types include high-loss type, medium-loss type, and low-loss type. In this embodiment, there are two ways to determine the loss type. One is based on the predicted loss value. Different loss thresholds are preset. If the predicted loss value is greater than the high-loss threshold, it is considered that the branch is a high-loss branch. If it is less than the low-loss threshold, it is considered that the branch is a low-loss branch. Otherwise, it is a medium-loss branch.

[0094] Another determination method is based on the loss sensitivity coefficient of the impedance loss model. According to the model formula, it can be known that impedance and loss are positively correlated. However, the sensitivity of the impedance of different branches to loss is significantly different. The sensitivity of impedance to loss is the loss sensitivity coefficient in the formula. For example, for the branches in the industrial area, due to the high load rate and large current, a slight increase in impedance will cause a large increase in loss. It can be seen from its impedance loss model that the loss sensitivity coefficient is relatively large. While the load rate and current of the residential area branches are significantly smaller than those in the industrial area, so the loss sensitivity coefficient in its model is also smaller. Therefore, it can be determined whether the branch is a high-loss branch through the loss sensitivity coefficient. Similarly, different coefficient thresholds are set. When the loss sensitivity coefficient of the model of a certain branch is greater than the high coefficient threshold, it indicates that the branch is a high-loss branch and mark this branch. If it is less than the low coefficient threshold, it is considered that the branch is a low-loss branch. Otherwise, it is a medium-loss branch. In this embodiment, the key lies in the marking of high-loss branches because high-loss branches have a relatively greater impact on the load.

[0095] In this embodiment, the impedance loss model is used to accurately predict the loss value of the branch, and the loss sensitivity coefficient is used to measure the influence degree of impedance on loss, identify the key branches that contribute the most to the overall network loss, provide clear optimization objectives and priorities for the subsequent algorithm, so as to achieve efficient allocation of resources.

[0096] For the initial loss data set obtained from the above steps and the loss types of each branch, through further loss pattern recognition, to judge the loss spatial distribution pattern of each branch. The specific steps include:

[0097] Use the PCA dimensionality reduction algorithm to perform dimensionality reduction processing on the initial loss data set to obtain the initial loss data set after dimensionality reduction processing;

[0098] Use the K-means clustering algorithm to perform clustering analysis on the initial loss data set after dimensionality reduction processing to obtain the load types of each branch. The load types include industrial load, commercial load, residential load, and mixed load;

[0099] Determine the loss spatial distribution pattern of each branch according to the load type and the loss type of each branch.

[0100] In this embodiment, by performing cluster analysis on the initial loss data set, the load type of each branch is determined, and the load type is combined with the loss type to determine the loss spatial distribution pattern of the branch. Among them, the load types include industrial load, commercial load, residential load, and mixed load. Since the loss characteristics of different load types are different, for industrial loads, due to the high proportion of equipment such as motors and transformers, the power factor is low, and a large amount of reactive power compensation is required. The current is large and continuous, so it will cause high losses and relatively high voltage deviations. That is, industrial loads have the characteristics of high-persistence loads, low power factors, and small loss fluctuations but high means. For commercial loads, since equipment such as air conditioners and lighting operate concentratedly during business hours, the power factor is relatively high, but the short-term load density leads to medium to high losses. Therefore, it has the characteristics of double-peak loads, relatively high power factors, and medium loss fluctuations. For residential loads, since household appliances are dispersed and the usage time is concentrated, the power factor is close to 1, but due to long lines and low terminal voltage, the losses are low but the voltage deviation fluctuations are large. Therefore, it has the characteristics of night peaks, dispersed and fluctuating loads, and high power factors. For mixed loads, which are the superposition of multiple types of loads, their losses and voltage deviations are between those of industry and residents.

[0101] According to the differences in the losses of different loads, in this embodiment, the PCA dimensionality reduction algorithm is used to perform dimensionality reduction processing on the initial loss data set. For example, the 20-dimensional features (loss, voltage, power factor, etc.) are reduced to 5-8-dimensional principal components (retaining 85% of the variance). By extracting time-series statistics such as mean, variance, peak value, etc. as features, the periodic law of the load is captured, and the core pattern of the load behavior is extracted. Then, the K-means clustering algorithm is used to perform cluster analysis on the dimensionality-reduced principal component features, so as to cluster each branch into different load types.

[0102] After obtaining the load type of each branch, combined with the loss type of the branch determined by the impedance loss model above, the loss spatial distribution pattern of each branch based on the load type and the loss type is determined. For example, branch A is industrial load - high loss, branch B is residential load - low loss, branch C is commercial load - medium loss, etc.

[0103] In the present invention, in addition to predicting the loss value, the impedance value of each branch is also related to the active power of the branch. In this embodiment, a power flow prediction model is used to predict the active power of each branch, and its specific steps include:

[0104] Adopt a sliding time window to extract the time-series characteristics of the current impedance value of each branch within a preset duration to obtain the impedance time-series characteristics;

[0105] Static feature extraction is performed on the initial loss data set to obtain the benchmark features of the network state;

[0106] The impedance time series features and the benchmark features of the network state are input into a pre-constructed power flow prediction model to obtain the predicted results of the power flow distribution of each branch. The power flow prediction model is constructed using a long short-term memory neural network model.

[0107] In this embodiment, a sliding time window is used to extract features from the current impedance values of each branch over a period of time. Assuming that the sliding time window is set to a 24-hour window with a 15-minute step size, the extracted features include 96-dimensional time series means, that is, the impedance means of each 15-minute segment, reflecting the intra-day periodic changes; 4-dimensional statistics, that is, the mean, variance, maximum value, and minimum value, describing the fluctuation characteristics of the impedance within the window, so as to obtain an impedance time series feature matrix. In this embodiment, a long short-term memory neural network model is preferably used to construct the power flow prediction model. As a model suitable for processing time series data, LSTM can use these time series features to learn and predict the future impedance change trend, and then affect the power flow distribution. The impedance time series feature matrix is input into the trained power flow prediction model to output the predicted value of the active power of the branch.

[0108] To improve the prediction accuracy, in a preferred embodiment, the present invention adds the static state of the branch to the input data of the power flow prediction model. Here, the static state is provided by the initial loss data set. Since the impedance time series feature matrix provides dynamic time change information, and the initial loss data set provides the loss and electrical state of the current network, the combination of the two can enable the model to consider both historical trends and the current actual state, improving the prediction accuracy. In this embodiment, the benchmark values of the current network state are provided through the loss values, voltage deviations, and power factors in the initial loss data set, reflecting the actual loss and electrical state of the current branch. The impedance time series features and the benchmark features of the network state are jointly used as the input data of the model, so as to obtain accurate prediction results.

[0109] In this embodiment, the future impedance trend is predicted through dynamic time series changes to drive the change of power flow distribution, and the prediction result is constrained by the static network state to conform to the current loss distribution, avoiding violating physical laws. The fusion of dynamic and static data combines historical trends and the current state, effectively improving the prediction robustness of the model.

[0110] Through the above steps, the predicted values of the active power of each branch and the loss spatial distribution pattern are determined. Then, based on the predicted values of the active power and the loss spatial distribution pattern, the load optimization distribution of the distribution network is realized by establishing a load distribution optimization model. The specific steps include:

[0111] Calculate the second loss prediction value of each branch according to the predicted active power value of each branch;

[0112] Determine the loss weight according to the loss spatial distribution pattern of each branch;

[0113] Perform weighted summation on the second loss prediction value according to the loss weight to obtain the network loss of the distribution network;

[0114] Calculate the load rate standard deviation according to the branch load rate of each branch;

[0115] Construct a load distribution optimization model with the minimization of the network loss and the load rate standard deviation as the objective function and the voltage deviation and the branch load rate as the constraint conditions.

[0116] In this embodiment, first, calculate the second loss prediction value of each branch according to the predicted active power value of each branch. Since the branch current is determined by the active power, that is, the branch current is equal to the value obtained by dividing the square root of the sum of the squares of the active power and the reactive power by the voltage value. The voltage value can be obtained through power flow calculation. Since the power factor of branches with different load types is relatively fixed, the reactive power can be calculated through the active power and the power factor. Finally, the branch current is calculated based on the predicted active power value. And the square of the branch current is proportional to the branch loss value, that is, the active power loss. Therefore, the loss prediction value of the branch can be obtained by multiplying the square value of the branch current by the branch resistance.

[0117] In the above embodiment, an impedance loss model is provided, that is, the loss value can be calculated through impedance. In this embodiment, the reason for not using this model for prediction is that the linear model of the fitting formula is a simplified approximation of the non-linear relationship (the loss value is positively correlated with the square value of the current). In the model, it is assumed that the impedance is stable in the short term, and the loss law under the current load distribution is reflected through historical data fitting. The coefficient of the model implies the average current effect under the current load level and does not directly reflect the relationship with the current, that is, sacrificing accuracy to improve calculation efficiency. Since the model is mainly used for classifying the losses of branches, there is no accuracy requirement for its predicted value.

[0118] The load distribution optimization model in this embodiment is used for refined prediction and optimization, and needs to accurately reflect the influence of current changes on losses. Therefore, the above model for loss classification is not used, but the change in active power will cause current fluctuations, and then the instantaneous loss value of each branch is calculated according to the current fluctuations.

[0119] Then, determine the loss weights of each branch according to the loss space distribution pattern of each branch. Specifically, set the basic weight according to the load type of the branch. In this embodiment, the basic weight of industrial load is the highest, for example, set to 0.8, and the weights of other loads are lower than that of industrial load, for example, set to 0.7. Then, according to the loss type of the branch, correct the basic weight. If it is a high-loss branch, increase the weight through a correction system, for example, multiply the weight by 1.2 to strengthen the penalty for its loss. Correspondingly, for a low-loss branch, reduce the weight through a correction coefficient, for example, multiply the weight by 0.8, and keep the weight unchanged for a medium-loss branch, so as to determine the loss weights of each branch. Finally, perform weighted summation on the predicted loss values and loss weights of each branch to obtain the network loss of the distribution network. Then, set the loss weight of the network loss, for example, 0.7. This weight value is set relatively high because during the operation of the distribution network, energy loss not only means waste of energy, but also may cause problems such as line heating and equipment aging, affecting the economy and reliability of the distribution network.

[0120] Obtain the branch load factor of each branch. The load factor is the ratio of the actual load to the rated capacity, which reflects the proportion of the current load borne by the branch relative to the maximum rated capacity it can withstand. Then, calculate the standard deviation of all branch load factors. The standard deviation is a statistic used to measure the degree of dispersion of a set of data. Here, it represents the degree of dispersion of each branch load factor relative to the average load factor, that is, it reflects the uniformity of the load distribution of each branch. In the optimization objective, making the load distribution of each branch uniform also has a certain importance. If the branch load distribution is uneven, it may cause some branches to be overloaded while other branches have light loads, which will not only affect the overall operation efficiency of the distribution network, but also may lead to safety hazards such as damage to equipment in overloaded branches. Therefore, a load balancing term is added to the objective function.

[0121] Then, construct the objective function by minimizing the network loss and the standard deviation of the load factor:

[0122]

[0123] In the formula, α represents the loss weight, β represents the load factor weight, P represents the network loss of the distribution network, and L std represents the standard deviation of the load factor.

[0124] At the same time, add constraint conditions, including voltage deviation constraint and load factor constraint. Among them, the voltage deviation constraint means that the absolute difference between the node voltage value and the nominal voltage is less than the threshold value. For example, it is restricted that the voltage deviation after optimization shall not exceed 5%. The load factor constraint means that the load factor of the branch is less than the load factor constraint threshold value. The load factor constraint is mainly a constraint for high-loss branches to avoid excessive load factors.

[0125] In this embodiment, the objective function F comprehensively considers the total loss of the distribution network and the uniformity of the load rates of each branch. With the goal of minimizing the value of F, it seeks the optimal load distribution plan. The loss term emphasizes reducing the total loss to improve the economy of the distribution network and reduce energy waste, while the load balancing term focuses on making the load distribution of each branch uniform to ensure the reliability and overall operation efficiency of the distribution network. The high efficiency, economy, and reliability of the distribution network operation are achieved by outputting the optimal load distribution plan.

[0126] For the load distribution optimization model, the particle swarm optimization algorithm is used for solution. First, the particle swarm is initialized, and the distribution ratios of the loads of each branch are used as the particle position vectors. The initial values are based on the historical load rates. The initial load rate of the high-loss branch is set to a lower value (such as 0.75). Then, the fitness is calculated, and the objective function value is calculated based on the input data (loss prediction, classification label, voltage deviation). When the constraint is violated, the fitness value is multiplied by a penalty coefficient (such as 1.5). Then, the dynamic weight adjustment is carried out. The weight of the industrial branch gradually increases with the increase of the iteration times to strengthen its optimization, and the inertia weight decreases linearly. Global search is carried out in the initial stage, and local refinement is carried out in the later stage. Finally, the global optimal solution is output, that is, the optimal load distribution ratio of each branch, ensuring that under this distribution plan, the distribution network loss is the lowest and the load is balanced, so as to realize the efficient and economic operation of the distribution network.

[0127] In a preferred embodiment, after obtaining the optimal load distribution plan, the present invention also provides a method for dynamically adjusting the distribution plan. The specific steps include:

[0128] Obtain the real-time monitoring data of the distribution network, and judge whether to trigger path transfer according to the real-time monitoring data. The real-time monitoring data includes the real-time branch load rate and the real-time voltage.

[0129] If path transfer is triggered, determine all transfer paths of the branch to be transferred according to the topological structure of the distribution network.

[0130] Adopt power flow calculation to obtain the transferred load and the loss reduction corresponding to each transfer path, and select the transfer path corresponding to the maximum loss reduction as the optimal transfer path.

[0131] Modify the optimal load distribution plan according to the transferred load and the optimal transfer path to obtain the modified optimal load distribution plan.

[0132] In this embodiment, real-time monitoring data of the distribution network is obtained, including branch load rates and voltage deviations. These data are directly collected in real time by monitoring devices (such as smart meters, voltage transformers, etc.) installed on each branch of the distribution network. The monitoring devices measure and calculate the operating parameters of the branches at certain time intervals (for example, every 1 minute) to obtain real-time monitoring data. Then, based on the branch load rate or voltage deviation, it is determined whether to trigger path transfer.

[0133] Specifically, when using the load rate for determination, a sliding time window is set. At each time point, the average value of the branch load rate within the current window is calculated, and at the same time, the average value of the branch load rate within the previous time window is recorded. Based on the average load rates of adjacent windows, the load rate change amount is calculated. If the load rate change amount is greater than a threshold, such as greater than 10%, and the duration is the same as the time window length, such as both are 5 minutes, it is determined that a load rate mutation has occurred and load transfer is required. When using the voltage deviation for determination, a sliding time window is also set. The window length is determined according to the number of consecutive samplings. After each voltage value is collected, the deviation percentage from the nominal voltage is calculated. If in the consecutive samplings, each calculated voltage deviation value is greater than the offset threshold, such as greater than 5%, it is considered that a voltage fluctuation has occurred and load transfer is required.

[0134] When load transfer is triggered, based on the topological structure information of the distribution network, all possible transfer paths of the current branch are determined. These paths are composed of a series of connected branches, which can transfer the load from the current branch to other branches to achieve load redistribution. Each path needs to meet the voltage constraint and branch capacity limit. For example, for branch A, there may be multiple candidate transfer paths such as path A - branch B - branch C, path A - branch D - branch E, etc.

[0135] The transferable load amount of the current branch is limited by its load rate and the remaining capacity of adjacent branches. Calculate the difference of this branch between the current branch load rate and the allowable minimum load rate, and calculate the difference of the adjacent branch between the rated capacity of the adjacent branch and the current load rate of the adjacent branch. Take the minimum value of the two differences as the upper limit of the load transfer amount. When calculating the optimal transfer load amount, with the goal of minimizing the total loss of the distribution network, under the constraints of voltage constraints and branch capacity constraints, in the particle swarm optimization or greedy algorithm, take the transfer load amount as the optimization variable, and obtain the optimal transfer ratio that satisfies the objective function through iterative search.

[0136] After determining the transfer load for each candidate power transfer path, for each candidate power transfer path, using a power flow calculation model (such as the improved forward-backward sweep method), combined with the current real-time monitoring data (including branch load rates, voltages, etc.) and the optimized load distribution scheme, calculate the change in the total loss of the entire distribution network after transferring a portion of the load to this path. Among them, assuming that the initial loss of the distribution network before transfer is P1 and the new loss of the distribution network after transferring the load to a certain candidate power transfer path is P2, then the loss reduction of this power transfer path is (P1 - P2) / P1. Compare the loss reductions of all candidate power transfer paths and select the path with the largest loss reduction as the power transfer path to be preferentially considered. For example, if the loss reduction of path A - branch B - branch C is 3.8%, while the loss reductions of other paths are 2.5%, 3.2%, etc., then select path A - branch B - branch C as the power transfer path.

[0137] According to the selected power transfer path, determine the branch switches that need to be operated. For example, for path A - branch B - branch C, it may be necessary to close the tie switch between branch A and branch B, and at the same time open some load switches on branch A to achieve the transfer of load from branch A to branch B and branch C. Then, according to the actual load transfer situation, correct the optimal load distribution scheme to obtain the optimized load distribution ratio of each branch.

[0138] A power distribution network digital simulation decision analysis method based on artificial intelligence provided in this embodiment. By establishing an impedance loss model and quantifying the sensitivity of impedance to loss, the present invention realizes the accurate positioning of high-loss branches. Through a power prediction model with multi-variable feature fusion, the power flow prediction accuracy is improved, providing data support for forward-looking decisions. And through a multi-objective particle swarm algorithm and a dynamic weight adjustment mechanism, the optimized distribution of the power distribution network load is realized. The present invention can effectively reduce the network loss of the power distribution network and improve the load balance degree, thus ensuring the safety and stability of the operation of the power distribution network.

[0139] Please refer to Figure 2 , based on the same inventive concept, a power distribution network digital simulation decision analysis system proposed in the second embodiment of the present invention includes:

[0140] A data processing module 10, configured to obtain power data of the power distribution network, perform data preprocessing on the power data, and generate a standardized data set. The standardized data set includes the branch ID, impedance value, active power, and reactive power of each branch of the power distribution network;

[0141] According to the topological structure of the power distribution network and the standardized data set, obtain an initial loss data set through power flow calculation. The initial loss data set includes the loss value, voltage deviation, and power factor of each branch;

[0142] A loss analysis module 20, configured to establish an impedance loss model according to the initial loss data set and the impedance values of each branch, and determine the loss types of each branch of the distribution network according to the current impedance values of each branch and the impedance loss model;

[0143] Perform loss pattern recognition on the initial loss data set and the loss types of each branch by using a preset algorithm to obtain the loss spatial distribution pattern of each branch;

[0144] A power flow prediction module 30, configured to input the current impedance values of each branch into a pre-constructed power flow prediction model to obtain the predicted active power values of each branch;

[0145] A load distribution module 40, configured to establish a load distribution optimization model according to the predicted active power values of each branch and the loss spatial distribution pattern, and solve it by using a particle swarm algorithm to obtain an optimal load distribution scheme.

[0146] The technical features and technical effects of the distribution network digital simulation decision analysis system based on artificial intelligence proposed in the embodiments of the present invention are the same as those of the method proposed in the embodiments of the present invention, and will not be elaborated here. Each module in the above-mentioned distribution network digital simulation decision analysis system based on artificial intelligence can be implemented in whole or in part by software, hardware and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory of the computer device in the form of software, so as to facilitate the processor to call and execute the operations corresponding to the above-mentioned each module.

[0147] In summary, a digital simulation decision analysis method and system for a distribution network based on artificial intelligence proposed in the embodiments of the present invention. The method obtains power data of the distribution network, performs data preprocessing on the power data to generate a standardized data set, and the standardized data set includes the branch ID, impedance value, active power, and reactive power of each branch of the distribution network; according to the topological structure of the distribution network and the standardized data set, an initial loss data set is obtained through power flow calculation, and the initial loss data set includes the loss value, voltage deviation, and power factor of each branch; an impedance loss model is established according to the initial loss data set and the impedance value of each branch, and according to the current impedance value of each branch and the impedance loss model, the loss type of each branch of the distribution network is judged; a preset algorithm is used to perform loss pattern recognition on the initial loss data set and the loss type of each branch to obtain the loss spatial distribution pattern of each branch; the current impedance value of each branch is input into a pre-constructed power flow prediction model to obtain the predicted active power value of each branch; according to the predicted active power value of each branch and the loss spatial distribution pattern, a load distribution optimization model is established, and the particle swarm algorithm is used for solution to obtain the optimal load distribution plan. The present invention quantifies the sensitivity of impedance to loss by establishing an impedance loss model, realizes the accurate positioning of high-loss branches, improves the power flow prediction accuracy through a power prediction model with multi-variable feature fusion, provides data support for forward-looking decisions, and realizes the optimal distribution of the distribution network load through a multi-objective particle swarm algorithm and a dynamic weight adjustment mechanism. The present invention can effectively reduce the network loss of the distribution network, improve the load balance degree, and thus ensure the safety and stability of the operation of the distribution network.

[0148] Each embodiment in this specification is described in a progressive manner. For parts that are the same or similar in each embodiment, they can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiment. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0149] The above-described embodiments merely represent several preferred embodiments of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and substitutions can be made, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the protection scope of the claims described above.

Claims

1. A digital simulation decision analysis method for a distribution network based on artificial intelligence, characterized in that, Including: Obtain the power data of the distribution network, perform data preprocessing on the power data to generate a standardized data set, where the standardized data set includes the branch ID, impedance value, active power, and reactive power of each branch of the distribution network; According to the topological structure of the distribution network and the standardized data set, obtain an initial loss data set through power flow calculation, where the initial loss data set includes the loss value, voltage deviation, and power factor of each branch; Based on the initial loss data set and the impedance values of each branch, establish an impedance loss model, and based on the current impedance values of each branch and the impedance loss model, determine the loss types of each branch of the distribution network; Use a preset algorithm to perform loss pattern recognition on the initial loss data set and the loss types of each branch to obtain the loss spatial distribution pattern of each branch; Input the current impedance values of each branch into a pre-constructed power flow prediction model to obtain the predicted active power values of each branch; Based on the predicted active power values of each branch and the loss spatial distribution pattern, establish a load distribution optimization model and use the particle swarm algorithm to solve it to obtain the optimal load distribution plan.

2. The method for digital simulation decision analysis of a distribution network based on artificial intelligence according to claim 1, wherein The step of obtaining the initial loss data set through power flow calculation according to the topological structure of the distribution network and the standardized data set includes: Construct a node-branch incidence matrix according to the topological structure of the distribution network; Based on the node-branch incidence matrix and the standardized data set, use the improved forward-backward substitution method to perform power flow calculation to obtain branch currents and node voltages; Calculate the loss value based on the branch current and the corresponding branch resistance; Calculate the voltage deviation based on the node voltage and the nominal voltage of the distribution network; Calculate the power factor based on the active power and reactive power of each branch; Based on the branch ID, the loss value, the voltage deviation, and the power factor of each branch, form an initial loss data set.

3. The digital simulation decision analysis method for a distribution network based on artificial intelligence according to claim 1, wherein The step of establishing an impedance loss model based on the initial loss data set and the impedance values of each branch, and determining the loss types of each branch of the distribution network based on the current impedance values of each branch and the impedance loss model includes: Use a sliding time window to extract the time series characteristics of the impedance values of each branch to obtain an impedance feature matrix; Use the least squares method to perform data fitting on the impedance feature matrix and the loss values in the initial loss data set to obtain the impedance loss model of each branch; Obtain the current impedance value of each branch and input it into the corresponding impedance loss model to obtain the first predicted loss value of each branch, and extract the loss sensitivity coefficient of each branch from the corresponding impedance loss model; If the first predicted loss value is greater than the loss threshold, or if the loss sensitivity coefficient is greater than the coefficient threshold, mark the corresponding branch as a high-loss branch and output the high-loss branch mark; Among them, the impedance loss model is represented by the following formula: Wherein, P loss represents the first predicted loss value, a represents the loss sensitivity coefficient, R represents the impedance value, and b represents the correction coefficient.

4. The digital simulation decision analysis method for a distribution network based on artificial intelligence according to claim 1, characterized in that, The step of using a preset algorithm to perform loss pattern recognition on the initial loss data set and the loss types of each branch to obtain the loss spatial distribution pattern of each branch includes: The PCA dimensionality reduction algorithm is used to perform dimensionality reduction processing on the initial loss data set, and the initial loss data set after dimensionality reduction processing is obtained; The K-means clustering algorithm is used to perform clustering analysis on the initial loss data set after dimensionality reduction processing, and the load types of each branch are obtained. The load types include industrial load, commercial load, residential load, and mixed load; According to the load types and loss types of each branch, the loss spatial distribution pattern of each branch is determined.

5. The digital simulation decision analysis method for a distribution network based on artificial intelligence according to claim 1, characterized in that The step of inputting the current impedance value of each branch into a pre-constructed power flow prediction model to obtain a power flow distribution prediction result includes: A sliding time window is used to extract the time series characteristics of the current impedance value of each branch within a preset time period, and impedance time series characteristics are obtained; Static feature extraction is performed on the initial loss data set to obtain network state reference features; The impedance time series characteristics and the network state reference features are input into a pre-constructed power flow prediction model to obtain the power flow distribution prediction results of each branch. The power flow prediction model is constructed using a long short-term memory neural network model.

6. The digital simulation decision analysis method for a distribution network based on artificial intelligence according to claim 1, wherein The step of establishing a load distribution optimization model according to the active power prediction value and the loss spatial distribution pattern of each branch includes: According to the active power prediction value of each branch, the second loss prediction value of each branch is calculated; According to the loss spatial distribution pattern of each branch, the loss weight is determined; According to the loss weight, the second loss prediction values are weighted and summed to obtain the network loss of the distribution network; According to the branch load rate of each branch, the load rate standard deviation is calculated; Taking the minimization of the network loss and the load rate standard deviation as the objective function, and taking the voltage deviation and the branch load rate as the constraint conditions, a load distribution optimization model is constructed.

7. The method for digital simulation decision analysis of a distribution network based on artificial intelligence according to claim 6, wherein The step of determining the loss weight according to the loss spatial distribution pattern of each branch includes: Corresponding initial weights are set according to the load types of each branch; According to the loss types of each branch, the initial weights are corrected to obtain the loss weights.

8. The digital simulation decision analysis method for a distribution network based on artificial intelligence according to claim 1, characterized in that, After the step of obtaining the optimal load distribution scheme, it further includes: Obtain the real-time monitoring data of the distribution network, and judge whether to trigger path transfer according to the real-time monitoring data. The real-time monitoring data includes the real-time branch load rate and the real-time voltage; If path transfer is triggered, according to the topological structure of the distribution network, all transfer paths of the branch to be transferred are determined; Power flow calculation is used to obtain the transfer load and loss reduction corresponding to each transfer path, and the transfer path corresponding to the maximum loss reduction value is selected as the optimal transfer path; According to the transfer load and the optimal transfer path, the optimal load distribution scheme is corrected to obtain the corrected optimal load distribution scheme.

9. The artificial intelligence-based digital simulation decision analysis method for a distribution network according to claim 8, wherein The step of judging whether to trigger path transfer according to the real-time monitoring data includes: According to the real-time branch load rate, a sliding time window mechanism is used to calculate the load change rate between adjacent windows. If the load change rate is greater than the change rate threshold, path transfer is triggered; Or, According to the real-time voltage obtained by each sampling, a sliding time window mechanism is used to calculate the voltage deviation within the current window to obtain multiple voltage deviation values; Compare each voltage deviation value with a deviation threshold. If the number of voltage deviation values greater than the deviation threshold is greater than a quantity threshold, trigger path power supply transfer.

10. A digital simulation decision analysis system for a distribution network based on artificial intelligence, characterized in that, It includes: A data processing module for obtaining power data of a distribution network, preprocessing the power data to generate a standardized data set, where the standardized data set includes the branch ID, impedance value, active power, and reactive power of each branch of the distribution network; According to the topological structure of the distribution network and the standardized data set, obtain an initial loss data set through power flow calculation, where the initial loss data set includes the loss value, voltage deviation, and power factor of each branch; A loss analysis module for establishing an impedance loss model based on the initial loss data set and the impedance value of each branch, and judging the loss type of each branch of the distribution network according to the current impedance value of each branch and the impedance loss model; Use a preset algorithm to perform loss pattern recognition on the initial loss data set and the loss type of each branch to obtain the loss spatial distribution pattern of each branch; A power flow prediction module for inputting the current impedance value of each branch into a pre-constructed power flow prediction model to obtain the predicted active power value of each branch; A load distribution module for establishing an optimal load distribution model based on the predicted active power value of each branch and the loss spatial distribution pattern, and using the particle swarm algorithm to solve it to obtain an optimal load distribution scheme.

Images