Method and device for optimizing intelligent scheduling algorithm of power distribution center

Abstract

The invention relates to the technical field of power system automation, and particularly provides a power distribution center intelligent scheduling algorithm optimization method and device, and the method comprises the following steps: S1, constructing a multi-source data fusion preprocessing module; s2, an improved Transform prediction model is constructed, and accurate prediction of load and power supply output is realized; s3, constructing a self-adaptive multi-objective optimization scheduling model; and S4, constructing a dynamic response adjustment module. Compared with the prior art, the method has the advantages that the improved time sequence prediction model and the adaptive multi-objective optimization algorithm can be fused, and accurate prediction, efficient scheduling and dynamic adjustment of the power distribution system are realized.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of power system automation, and particularly provides a power distribution center intelligent scheduling algorithm optimization method and device. BACKGROUND

[0002] With large-scale access of new loads such as distributed photovoltaic, energy storage equipment and electric vehicle charging piles, the operation state of the power distribution system presents the characteristics of strong randomness and high fluctuation, and the traditional power distribution scheduling algorithm based on experience rules or static models has been difficult to meet the multi-objective operation requirements of system safety, economy and environmental protection.

[0003] The current power distribution center algorithm has the following key problems:

[0004] Insufficient load and power prediction accuracy: the traditional time series prediction method (such as ARIMA) cannot effectively capture the nonlinear correlation characteristics of distributed power output and new loads, and the prediction error is large, which leads to the disconnection of the scheduling scheme and the actual operation state;

[0005] Low multi-objective optimization efficiency: power distribution scheduling needs to meet multiple objectives such as "reducing network loss, reducing voltage deviation and improving renewable energy consumption rate", but the existing algorithm mostly uses the weighted summation method to convert multiple objectives into a single objective for solving, which is easy to appear "optimal solution bias", and the solving process converges slowly, which is difficult to adapt to real-time scheduling requirements;

[0006] Weak dynamic response capability: when the power distribution system has a sudden fault (such as line overload and distributed power off-grid), the traditional algorithm cannot quickly adjust the scheduling strategy, which easily causes safety risks such as voltage out-of-limit and equipment overload.

[0007] Therefore, it is urgent to design a power distribution center algorithm optimization scheme with high-precision prediction capability, efficient multi-objective optimization capability and fast dynamic response capability to improve the operation reliability and economy of the power distribution system. SUMMARY

[0008] The application is aimed at the deficiencies of the prior art, and provides a practical power distribution center intelligent scheduling algorithm optimization method.

[0009] The further technical task of the application is to provide a power distribution center intelligent scheduling algorithm optimization device which is reasonable in design and safe and applicable.

[0010] The technical scheme adopted by the application to solve the technical problems is:

[0011] The power distribution center intelligent scheduling algorithm optimization method is based on deep learning and multi-objective optimization, and has the following steps:

[0012] S1, a multi-source data fusion preprocessing module is constructed;

[0013] S2, build an improved Transformer prediction model to realize accurate prediction of load and power output;

[0014] S3, build an adaptive multi-objective optimization scheduling model;

[0015] S4, build a dynamic response adjustment module.

[0016] Further, in step S1, including:

[0017] S1-1, collect multi-source data of power distribution system;

[0018] S1-2, pre-process the collected data.

[0019] Further, in step S1-1, including:

[0020] Distributed power data: photovoltaic power station output, wind power output and energy storage equipment charging and discharging state;

[0021] Load data: real-time power of residential load, industrial load and electric vehicle charging pile load;

[0022] Grid operation data: line current / voltage, transformer load rate and node power factor;

[0023] Environmental data: light intensity, wind speed, temperature and precipitation weather data.

[0024] Further, in step S1-2, including:

[0025] Outlier processing: use an anomaly detection model based on the Isolation Forest algorithm to identify and correct mutation values and missing values in the data;

[0026] Data normalization: map data of different dimensions to the [0, 1] interval through Min-Max standardization to avoid the influence of data magnitude difference on model training;

[0027] Feature selection: based on Mutual Information, filter features strongly related to load / power output.

[0028] Further, in step S2, including:

[0029] On the basis of the traditional Transformer encoder, introduce the time attention mechanism to strengthen the model's ability to capture the time sequence dependence of load / power output;

[0030] Add a weather feature embedding layer to convert light and wind speed weather data into a high-dimensional feature vector and fuse it with load time sequence features;

[0031] The pre-processed multi-source data is used as the training set, and the AdamW optimizer and the root mean square error (RMSE) loss function are used to train the improved Transformer model.

[0032] The trained model is used to output the load prediction curve and the distributed power output prediction curve for a future period of time, with a prediction error controlled within 5%.

[0033] Further, in step S3, it includes:

[0034] S3-1, target 1 economy: minimize distribution system loss, formula:

[0035]

[0036] Where I is the current of line i, R is the resistance of line i, and Δt is the scheduling time interval. i i

[0037] Target 2 safety: minimize node voltage deviation, formula:

[0038]

[0039] Where U is the actual voltage of node j, and U is the rated voltage of node j. j j0

[0040] Target 3 environmental protection: maximize renewable energy consumption rate, formula:

[0041]

[0042] Where P is the total renewable energy output, and P is the conventional power output. ren conv

[0043] S3-2, design constraints:

[0044] Power balance constraint:

[0045] ΣP in = ΣP out + P loss ;

[0046] Device capacity constraint: line current ≤ rated current, transformer load rate ≤ 80%;

[0047] Energy storage device constraint: energy storage SOC ∈ [20%, 80%], charge and discharge power ≤ rated power;

[0048] S3-3, design constraints:

[0049] ​​​​​​Based on the non-dominated sorting genetic algorithm NSGA-Ⅱ, an adaptive crossover and mutation operator is introduced: the crossover probability and mutation probability are dynamically adjusted according to the convergence degree of the population, so that the algorithm is prevented from falling into local optimum;

[0050] Adopt the congestion degree sorting and elite reservation strategy to screen out the uniformly distributed Pareto optimal solution set for the dispatcher to select according to the actual demand.

[0051] Further, in step S4, comprising:

[0052] S4-1, real-time monitoring of the operation state of the power distribution system, triggering the dynamic adjustment mechanism when a sudden fault occurs;

[0053] S4-2, based on the real-time prediction data of step S2 and the optimization model of step S3, adopting the rolling time domain optimization strategy, dividing the 24-hour scheduling period into multiple 15-minute sub-periods, updating the optimization objective and constraint conditions every 15 minutes, and quickly generating the adjusted scheduling scheme;

[0054] S4-3, outputting the scheduling instruction: controlling the charge and discharge power of the energy storage device, adjusting the output of the distributed power supply, switching the tie switch, realizing the rapid recovery of the fault, and ensuring the stable operation of the system.

[0055] The power distribution center intelligent scheduling algorithm optimization device comprises at least one memory and at least one processor;

[0056] The at least one memory is used to store a machine readable program;

[0057] The at least one processor is used to call the machine readable program and execute the power distribution center intelligent scheduling algorithm optimization method.

[0058] Compared with the prior art, the power distribution center intelligent scheduling algorithm optimization method and device have the following outstanding beneficial effects:

[0059] After the application of the present application, the prediction error is reduced by more than 30% compared with the traditional method, the convergence speed of the improved NSGA-Ⅱ algorithm is improved by 25%, the network loss is reduced by 8% to 12%, the voltage deviation is less than or equal to 3%, and the renewable energy consumption rate is greater than or equal to 90%, which significantly improves the reliability, economy and environmental protection of the power distribution system. DETAILED DESCRIPTION

[0060] In order for those skilled in the art to better understand the scheme of the present application, the present application will be further described in detail below in combination with specific embodiments. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0061] A preferred embodiment is given below:

[0062] Embodiment 1:

[0063] The power distribution center intelligent scheduling algorithm optimization method in this embodiment is based on deep learning and multi-objective optimization, and has the following steps:

[0064] S1, construct a multi-source data fusion preprocessing module;

[0065] Including:

[0066] S1-1, collect multi-source data of the power distribution system, including:

[0067] Distributed power data: photovoltaic power station output, wind power output, and energy storage device charging and discharging state;

[0068] Load data: real-time power of residential load, industrial load, and electric vehicle charging pile load;

[0069] Power grid operation data: line current / voltage, transformer load rate, and node power factor;

[0070] Environmental data: light intensity, wind speed, temperature, and precipitation meteorological data.

[0071] S1-2, pre-process the collected data, including:

[0072] Outlier processing: use an anomaly detection model based on the Isolation Forest algorithm to identify and correct mutation values and missing values in the data;

[0073] Data normalization: map data of different dimensions to the [0, 1] interval through Min-Max standardization to avoid the influence of data magnitude differences on model training;

[0074] Feature selection: based on mutual information entropy (Mutual Information), select features that are strongly related to load / power output (such as the mutual information value of light intensity and photovoltaic output ≥0.8), and reduce the input dimension of the model.

[0075] S2, construct an improved Transformer prediction model to achieve accurate prediction of load and power output;

[0076] Including:

[0077] On the basis of the traditional Transformer encoder, introduce a time attention mechanism to strengthen the model's ability to capture the time sequence dependence of load / power output;

[0078] A new weather feature embedding layer is added to convert weather data such as light and wind speed into high-dimensional feature vectors, which are fused with load time series features to improve prediction accuracy.

[0079] The improved Transformer model is trained using the pre-processed multi-source data as the training set, AdamW optimizer and root mean square error (RMSE) loss function.

[0080] The trained model is used to output 24-hour load prediction curves and distributed power output prediction curves, with a prediction error controlled within 5%.

[0081] S3, build an adaptive multi-objective optimization scheduling model;

[0082] Including:

[0083] S3-1, determine the optimization objective function:

[0084] Objective 1 (economic efficiency): minimize distribution system loss, formula:

[0085]

[0086] Where I is the current of line i, R is the resistance of line i, and Δt is the scheduling time interval. i i

[0087] Objective 2 (safety): minimize node voltage deviation, formula:

[0088]

[0089] Where U is the actual voltage of node j, and U is the rated voltage of node j. j j0

[0090] Objective 3 (environmental protection): maximize renewable energy consumption rate, formula:

[0091]

[0092] Where P is the total renewable energy output, and P is the conventional power output. ren conv

[0093] S3-2, design constraints:

[0094] Power balance constraint:

[0095] ΣP in = ΣP out + P loss .

[0096] ​​​​​​Device capacity constraints: line current ≤ rated current, transformer load rate ≤ 80%;

[0097] Energy storage device constraints: energy storage SOC (state of charge) ∈ [20%, 80%], charge and discharge power ≤ rated power.

[0098] S3-3, improve multi-objective optimization algorithm:

[0099] Based on the non-dominated sorting genetic algorithm (NSGA-Ⅱ), the adaptive crossover and mutation operator is introduced: according to the convergence degree of the population, the crossover probability (0.6-0.9) and the mutation probability (0.01-0.05) are dynamically adjusted to avoid the algorithm falling into local optimum;

[0100] Adopt the crowding degree sorting and elite reservation strategy to select the uniformly distributed Pareto optimal solution set for the dispatch personnel to select according to the actual demand.

[0101] S4, construct a dynamic response adjustment module;

[0102] Including:

[0103] S4-1, real-time monitoring of power distribution system operation state, when sudden failure (such as line overload, voltage limit) occurs, trigger the dynamic adjustment mechanism;

[0104] S4-2, based on the real-time prediction data of step S2 and the optimization model of step S3, adopt the rolling time domain optimization strategy, divide the 24-hour scheduling period into multiple 15-minute sub-periods, update the optimization objective and constraint conditions every 15 minutes, and quickly generate the adjusted scheduling scheme;

[0105] S4-3, output scheduling instructions: control the charge and discharge power of energy storage devices, adjust the output of distributed power, switch the tie switch, realize the rapid recovery of fault, and ensure the stable operation of the system.

[0106] Example 2:

[0107] Take the application of a certain city 10kV distribution center as an example:

[0108] Data collection and preprocessing: collect the load data of 20 nodes under the distribution center, the output data of 5 photovoltaic power stations, the operation data of 3 energy storage devices and the local weather data, process the outliers through the isolation forest algorithm, and select 6 key features such as light intensity and temperature;

[0109] Model training and prediction: take the data from January to June 2024 as the training set and the data in July as the test set, train the improved Transformer model, the photovoltaic output prediction RMSE is 4.2%, and the load prediction RMSE is 3.8%;

[0110] Multi-objective optimization solution: input the predicted data into the adaptive NSGA-II algorithm, and solve the Pareto optimal solution set. Select the minimum network loss scheme: network loss reduction of 10.5%, voltage deviation of 2.1%, and renewable energy consumption rate of 92.3%;

[0111] Dynamic response test: simulate an overload fault of a certain line. The dynamic response module generates an adjustment scheme within 3 minutes, controls the discharge power of the energy storage device to increase by 200kW, and reduces the line load rate from 85% to 68%, and restores the normal operation of the system.

[0112] Example 3:

[0113] The power distribution center intelligent scheduling algorithm optimization device comprises at least one memory and at least one processor.

[0114] The at least one memory is used to store a machine-readable program.

[0115] The at least one processor is used to call the machine-readable program and execute the power distribution center intelligent scheduling algorithm optimization method.

[0116] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), ready-to-program gate arrays (FPGAs), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The processor can be a microprocessor, or the processor can be any conventional processor.

[0117] The memory can be used to store computer programs and / or modules. The processor realizes various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required by a function, etc.; the data storage area can store data created by the terminal during use, etc. In addition, the memory can also include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SMC), a secure digital (SD) card, a flash memory card, at least one magnetic disk storage device, a flash memory device, or other volatile solid-state memory devices.

[0118] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions recorded in the above embodiments can be modified, or some or all of the technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. An optimization method for intelligent dispatching algorithms in power distribution centers, characterized in that, Based on deep learning and multi-objective optimization, the following steps are involved: S1. Construct a multi-source data fusion preprocessing module; S2. Construct an improved Transformer prediction model to achieve accurate prediction of load and power output; S3. Construct an adaptive multi-objective optimization scheduling model; S4. Construct a dynamic response adjustment module.

2. The intelligent dispatching algorithm optimization method for power distribution centers according to claim 1, characterized in that, Step S1 includes: S1-1. Collect multi-source data from the power distribution system; S1-2. Preprocess the collected data.

3. The intelligent dispatching algorithm optimization method for power distribution centers according to claim 2, characterized in that, Step S1-1 includes: Distributed power data: output of photovoltaic power plants, output of wind power, and charging and discharging status of energy storage devices; Load data: Real-time power of residential load, industrial load and electric vehicle charging pile load; Power grid operation data: line current / voltage, transformer load factor, and node power factor; Environmental data: light intensity, wind speed, temperature, and precipitation meteorological data.

4. The intelligent dispatching algorithm optimization method for power distribution centers according to claim 3, characterized in that, Step S1-2 includes: Outlier handling: An anomaly detection model based on the isolated forest algorithm is used to identify and correct abrupt and missing values ​​in the data; Data normalization: Min-Max standardization maps data of different scales to the [0,1] interval to avoid the impact of differences in data scale on model training; Feature filtering: Based on mutual information entropy, features strongly correlated with load / power output are filtered.

5. The intelligent dispatching algorithm optimization method for power distribution centers according to claim 4, characterized in that, Step S2 includes: Based on the traditional Transformer encoder, a temporal attention mechanism is introduced to enhance the model's ability to capture the temporal dependencies between load and power output. A new meteorological feature embedding layer is added to transform sunlight and wind speed meteorological data into high-dimensional feature vectors, which are then fused with load time-series features. Using preprocessed multi-source data as the training set, the improved Transformer model was trained using the AdamW optimizer and the root mean square error (RMSE) loss function. Using the trained model, the load forecast curve and distributed power output forecast curve for a future period are output, with the forecast error controlled within 5%.

6. The intelligent dispatching algorithm optimization method for power distribution centers according to claim 5, characterized in that, Step S3 includes: S3-1, Objective 1: Economic efficiency: Minimize the network loss of the power distribution system, the formula is: Among them, I i R is the current in line i. i Let be the resistance of line i, and Δt be the scheduling time interval; Objective 2: Security: Minimize node voltage deviation, as shown in the formula: Among them, U j U is the actual voltage at node j. j0 Let be the rated voltage of node j; Objective 3: Environmental friendliness: Maximize the renewable energy integration rate, the formula is: Among them, P ren For total renewable energy output, P conv It provides power to conventional power sources. S3-2, Design Constraints: Power balance constraints: ΣP in =ΣP out +P loss ; Equipment capacity constraints: line current ≤ rated current, transformer load rate ≤ 80%; Energy storage device constraints: Energy storage SOC ∈ [20%, 80%], charging and discharging power ≤ rated power; S3-3, Design Constraints: Based on the non-dominated sorting genetic algorithm NSGA-II, an adaptive crossover and mutation operator is introduced: the crossover probability and mutation probability are dynamically adjusted according to the population convergence degree to avoid the algorithm getting trapped in local optima; By employing a congestion ranking and elite retention strategy, a set of uniformly distributed Pareto optimal solutions is selected for schedulers to choose from based on actual needs.

7. The intelligent dispatching algorithm optimization method for power distribution centers according to claim 5, characterized in that, Step S4 includes: S4-1. Monitor the operating status of the power distribution system in real time, and trigger a dynamic adjustment mechanism when a sudden fault occurs; S4-2. Based on the real-time prediction data from step S2 and the optimization model from step S3, a rolling time-domain optimization strategy is adopted to divide the 24-hour scheduling cycle into multiple 15-minute sub-cycles. The optimization objective and constraints are updated every 15 minutes to quickly generate the adjusted scheduling scheme. S4-3 Output scheduling commands: Control the charging and discharging power of energy storage equipment, adjust the output of distributed power sources, and switch interconnection switches to achieve rapid fault recovery and ensure stable system operation.

8. A power distribution center intelligent dispatching algorithm optimization device, characterized in that, include: At least one memory and at least one processor; The at least one memory is used to store a machine-readable program; The at least one processor is configured to invoke the machine-readable program to perform the method according to any one of claims 1 to 7.

Images