Power resource scheduling and load balancing optimization method and system combined with big data

Abstract

The application discloses a power resource scheduling and load balancing optimization method and system combined with big data, and comprises the following steps: constructing a power demand prediction model, and obtaining a variety of power demand prediction result sets of each target area in a future time period; inputting real-time collection data into the power demand prediction model, and obtaining a prediction result with the highest matching degree; calculating the power load coefficient of each target area; constructing a regional load difference matrix, and obtaining a power resource dispatching-out area set and a dispatching-in area set; dividing a scheduling period into multiple time periods, and scheduling power resources through incremental power transfer; inputting the real-time collection data after scheduling into the power demand prediction model to repeat the above steps until the median value in the obtained regional load difference matrix is less than a scheduling threshold value. The application has the advantages that: accurate power demand prediction is realized through real-time data dynamic matching, and the power distribution between regions is optimized based on the load difference matrix and incremental scheduling, so that the power grid operation efficiency and stability are significantly improved.

Description

Technical Field

[0001] This invention relates to dynamic scheduling technology, and in particular to a method and system for optimizing power resource scheduling and load balancing by combining big data. Background Technology

[0002] Power resource dispatch and load balancing optimization are key technologies in the operation of modern power systems, aiming to achieve a dynamic balance between power production, transmission, and consumption. With the large-scale integration of renewable energy sources such as wind and solar power, the power system faces unprecedented challenges: these clean energy sources exhibit significant intermittency and volatility, and their power generation output is highly uncertain due to weather conditions. Simultaneously, the demand side is also showing new trends, including the randomness of electric vehicle charging loads, the diversity of user electricity consumption behavior, and load surges caused by extreme weather events. These factors collectively lead to a significant increase in the uncertainty of grid operation.

[0003] Current power resource dispatching and load balancing optimization methods mainly rely on simple load forecasting models and lack deep integration and dynamic adjustment of real-time grid data. These methods typically employ static forecasting models, which cannot respond to rapid changes in power demand and load in real time, resulting in poor timeliness of dispatching decisions and difficulty in optimizing the real-time allocation of power resources. Furthermore, traditional methods often fail to effectively distinguish the actual demand of different regions when dealing with regional load differences in complex power systems, leading to resource waste or uneven grid load. Moreover, many methods lack continuous feedback and optimization mechanisms, making them unable to adapt to changes in grid conditions and demand, resulting in long-term operational instability. Summary of the Invention

[0004] To improve existing power resource dispatching and load balancing optimization methods, this paper presents a method and system that integrates big data for power resource dispatching and load balancing optimization. This method accurately predicts power demand through big data analysis and LSTM deep learning models, dynamically adjusts power dispatching based on real-time grid data, and optimizes regional load distribution. It utilizes incremental power transfer and greedy algorithms to improve dispatching efficiency, reduce resource waste, and ensure stable grid operation.

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

[0006] Power resource scheduling and load balancing optimization methods that combine big data include:

[0007] A power demand forecasting model is built based on historical electricity consumption data, and a set of multiple electricity demand forecast results for various target regions in the future time period is generated.

[0008] Based on real-time data collected from power grid nodes in each target region, the data is input into the power demand forecasting model to obtain the forecast result with the highest matching degree among multiple sets of power demand forecast results.

[0009] The power load factor of each target region is calculated based on the electricity demand forecast results with the highest matching degree for each target region.

[0010] Based on the power load coefficient of each target region, a regional load difference matrix is ​​constructed to obtain the set of power resource outflow regions and the set of power resource inflow regions.

[0011] Based on the acquired set of outgoing and incoming regions, the scheduling cycle is divided into multiple time periods, and power resource scheduling is carried out through incremental power transfer.

[0012] Based on the real-time collected data of each target region after power resource dispatch, the data is input into the power demand forecasting model and the above steps are repeated until the median of the obtained regional load difference matrix is ​​less than the dispatch threshold.

[0013] Preferably, the step of constructing a power demand forecasting model based on historical electricity consumption data and generating a set of multiple electricity demand forecasting results for each target region in the future time period specifically includes:

[0014] Data preprocessing is performed based on the acquired historical data, and time series features and spatial features are constructed.

[0015] The acquired feature data is input into an LSTM-based deep model for training, and the accuracy of the model is evaluated using cross-validation.

[0016] Based on the trained model, the electricity demand of each target region at different time periods is predicted, including short-term demand forecast and long-term demand forecast.

[0017] Based on the prediction results output by the model, a set of electricity demand prediction results for different confidence intervals is generated.

[0018] Preferably, the step of collecting real-time data from power grid nodes in each target region, inputting it into the electricity demand forecasting model, and obtaining the forecast result with the highest matching degree from a set of multiple electricity demand forecast results specifically includes:

[0019] Real-time power data for each target area is collected through a real-time monitoring system at power grid nodes, including current power consumption, power grid frequency, voltage, and power transmission status.

[0020] Preprocessing is performed on the collected real-time power data, and dynamic features are constructed.

[0021] The constructed feature vector is projected onto the pre-trained feature encoder, and dual matching is performed in both time and space dimensions. Specifically, this includes comparing the morphological similarity between the real-time sequence and the predicted curve through dynamic time warping, and calculating the cosine similarity between the real-time topological feature vector and the predicted scene.

[0022] Based on the matching degree between real-time data and multiple sets of prediction results, the prediction result with the highest similarity to the current power grid state is selected.

[0023] Preferably, the calculation of the power load factor for each target region based on the electricity demand forecast result with the highest matching degree for each target region specifically includes:

[0024] Based on the electricity consumption data for each time period in the time series data of the prediction results and the total number of prediction periods, the average load and maximum load within that time period are obtained.

[0025] By inputting the calculated average load and maximum load into the load factor formula, the power load factor can be obtained.

[0026] Calculate the power load factor for each target region to obtain a time series dataset of the power load factor.

[0027] Preferably, the step of constructing a regional load difference matrix based on the power load coefficient of each target region to obtain the set of power resource exporting regions and the set of power resource importing regions specifically includes:

[0028] Based on the acquired time series dataset of electricity load coefficients, the comprehensive load average for all target regions is calculated;

[0029] Based on the comprehensive load average, the difference between the power load of each target area and the comprehensive load average is calculated, and a regional load difference matrix is ​​constructed.

[0030] Based on the difference values ​​in the regional load difference matrix, a difference judgment threshold is set to classify and identify the regions that are transferred out and regions that are transferred in.

[0031] Based on the degree of difference, the regions being transferred out and the regions being transferred in are sorted, with positive numbers representing regions being transferred in and negative numbers representing regions being transferred out, thus establishing a priority sequence for scheduling and docking.

[0032] Preferably, the step of dividing the scheduling cycle into multiple time periods based on the acquired set of outgoing and incoming regions, and scheduling power resources through incremental power transfer specifically includes:

[0033] The power dispatch cycle is divided into hours, days, and weeks, and adjusted according to actual power demand.

[0034] Based on the constructed scheduling and docking priority sequence, the regions to be transferred are sorted from largest to smallest according to the difference value, and the regions to be transferred out are sorted from smallest to largest. The two target regions with the closest absolute difference value are selected.

[0035] Based on the pre-defined scheduling cycle, the sending region will provide power during the specified time period, and the receiving region will receive power during the specified time period.

[0036] Based on load differences and changes in electricity demand, the amount of electricity dispatched in each time period is dynamically adjusted using a greedy algorithm.

[0037] Real-time monitoring of power grid operation, actual load and dispatch effectiveness in various regions, and retraining of the power demand forecasting model through a feedback mechanism to obtain stable forecast results.

[0038] Furthermore, a power resource scheduling and load balancing optimization system combining big data is proposed, including:

[0039] Data acquisition module: The data acquisition module is mainly responsible for collecting key operational data such as electricity consumption, voltage, current, and frequency of power grid nodes in each target area in real time;

[0040] Data processing module: The data processing module is mainly used to clean, normalize and extract features from historical electricity consumption data to provide standardized input for subsequent modeling;

[0041] Electricity demand forecasting module: The electricity demand forecasting module is mainly used to make short-term and long-term forecasts of electricity demand in the future based on the LSTM model, and generate a set of forecast results under multiple confidence intervals.

[0042] Similarity matching module: The similarity matching module is mainly used to perform dual matching of real-time collected data and prediction set with time series and spatial features, and select the optimal prediction result under the current state;

[0043] Load factor calculation module: The load factor calculation module is mainly used to calculate the average load and maximum load of each target area based on the selected prediction results, and to obtain the load factor time series;

[0044] Regional Difference Analysis Module: The regional difference analysis module is mainly used to construct a regional load difference matrix, identify the regions where power resources are transferred out and transferred in, and establish a scheduling priority list based on load differences;

[0045] Power resource dispatching module: The power resource dispatching module is mainly used to divide the dispatching cycle and use a greedy algorithm to formulate the optimal incremental power dispatching strategy to complete the transfer of power from the sending area to the receiving area;

[0046] Feedback module: The feedback module is mainly used to evaluate the scheduling effect and update the power forecasting model based on the scheduling results and real-time operation data;

[0047] Processor: The processor is mainly used for the calculation process of each formula and the construction calculation process of each model.

[0048] Compared with the prior art, the advantages of the present invention are:

[0049] This method achieves accurate power demand forecasting through a deep learning model based on LSTM and dynamic matching of real-time data, thus ensuring the accuracy of dispatching decisions. Secondly, based on power load coefficients and regional load difference matrices, the method intelligently identifies regions for power outflow and inflow, and achieves precise power dispatching through incremental power transfer. By dividing the dispatching cycle into different phases and dynamically adjusting the power transfer amount using a greedy algorithm, the flexibility and real-time performance of dispatching are further improved. Finally, the system's continuous feedback and retraining mechanism ensures the stability and long-term effectiveness of the power demand forecasting model. This method can significantly reduce the waste of power resources, optimize the distribution of power load between regions, improve the operating efficiency and stability of the power grid, and promote green and sustainable energy management. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the method proposed in this invention;

[0051] Figure 2 This is a schematic diagram of the electricity demand forecasting model proposed in this invention;

[0052] Figure 3 This is a schematic diagram illustrating the prediction result matching proposed in this invention;

[0053] Figure 4 This is a schematic diagram illustrating the acquisition of the power load factor proposed in this invention;

[0054] Figure 5 This is a schematic diagram illustrating the division of regions for data transfer in and out proposed in this invention;

[0055] Figure 6 This is a schematic diagram of the power dispatching method proposed in this invention;

[0056] Figure 7 This is an architecture diagram of the electronic devices in this solution;

[0057] Figure 8 This is a schematic diagram of the computer-readable storage medium structure in this scheme. Detailed Implementation

[0058] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0059] A power resource dispatching and load balancing optimization system that integrates big data includes:

[0060] Data acquisition module: The data acquisition module is mainly responsible for collecting key operational data such as electricity consumption, voltage, current, and frequency of power grid nodes in each target area in real time;

[0061] Data processing module: The data processing module is mainly used to clean, normalize and extract features from historical electricity consumption data to provide standardized input for subsequent modeling;

[0062] Electricity demand forecasting module: The electricity demand forecasting module is mainly used to make short-term and long-term forecasts of electricity demand in the future based on the LSTM model, and generate a set of forecast results under multiple confidence intervals.

[0063] Similarity matching module: The similarity matching module is mainly used to perform dual matching of real-time collected data and prediction set with time series and spatial features, and select the optimal prediction result under the current state;

[0064] Load factor calculation module: The load factor calculation module is mainly used to calculate the average load and maximum load of each target area based on the selected prediction results, and to obtain the load factor time series;

[0065] Regional Difference Analysis Module: The regional difference analysis module is mainly used to construct a regional load difference matrix, identify the regions where power resources are transferred out and transferred in, and establish a scheduling priority list based on load differences;

[0066] Power resource dispatching module: The power resource dispatching module is mainly used to divide the dispatching cycle and use a greedy algorithm to formulate the optimal incremental power dispatching strategy to complete the transfer of power from the sending area to the receiving area;

[0067] Feedback module: The feedback module is mainly used to evaluate the scheduling effect and update the power forecasting model based on the scheduling results and real-time operation data;

[0068] Processor: The processor is mainly used for the calculation process of each formula and the construction calculation process of each model.

[0069] See Figure 1 As shown, the power resource scheduling and load balancing optimization method combining big data includes:

[0070] Step 1: Construct an electricity demand forecasting model based on historical electricity consumption data, and generate a set of multiple electricity demand forecast results for each target region in the future time period;

[0071] Step 2: Based on real-time data collected from power grid nodes in each target region, input the data into the power demand forecasting model to obtain the forecast result with the highest matching degree among multiple power demand forecast result sets;

[0072] Step 3: Calculate and obtain the power load factor for each target region based on the electricity demand forecast results with the highest matching degree for each target region;

[0073] Step 4: Based on the power load coefficient of each target region, construct a regional load difference matrix to obtain the set of power resource exporting regions and the set of power resource importing regions.

[0074] Step 5: Based on the acquired set of outgoing and incoming regions, the scheduling cycle is divided into multiple time periods, and power resource scheduling is carried out through incremental power transfer;

[0075] Step Six: Based on the real-time collected data of each target region after power resource scheduling, input it into the power demand forecasting model and repeat the above steps until the median value of the obtained regional load difference matrix is ​​less than the scheduling threshold.

[0076] See Figure 2 As shown, an electricity demand forecasting model is constructed based on historical electricity consumption data, and a set of multiple electricity demand forecast results for various target regions in future time periods is generated, specifically including:

[0077] Data preprocessing is performed based on the acquired historical data, and time series features and spatial features are constructed.

[0078] The acquired feature data is input into an LSTM-based deep model for training, and the accuracy of the model is evaluated using cross-validation.

[0079] Based on the trained model, the electricity demand of each target region at different time periods is predicted, including short-term demand forecast and long-term demand forecast.

[0080] Based on the prediction results output by the model, a set of electricity demand prediction results for different confidence intervals is generated.

[0081] Specifically, the LSTM architecture includes three layers: the input layer accepts feature data of the time step, such as the power demand of the previous moment, time features, spatial features, etc., the LSTM layer sets up multiple LSTM units to learn long-term dependencies, and the output layer predicts the power demand of future moments.

[0082] Cross-validation is used to evaluate model performance and prevent overfitting. Mean Squared Error (MSE) is used as the evaluation metric, with the following formula:

[0083]

[0084] Among them, yi This is the actual value. The predicted value is N, where N is the number of historical data samples.

[0085] For short-term demand forecasting, a trained LSTM model is used to predict electricity demand for the next few hours or days. For long-term demand forecasting, external factors (such as seasonal changes, policy changes, etc.) and long-term trends need to be considered.

[0086] Based on the predicted values ​​output by the model, multiple prediction models are obtained by resampling the training data multiple times. These prediction results are combined to calculate the confidence interval of the predicted values. Prediction results with different confidence intervals are generated according to the model output. The formula for the confidence interval is:

[0087]

[0088] in, For predicted values, Let n be the standard deviation of the prediction error, and n be the sample size. This is the critical value for the standard normal distribution.

[0089] See Figure 3 As shown, based on real-time data collected from power grid nodes in each target region, this data is input into the electricity demand forecasting model to obtain the forecast result with the highest matching degree from a set of various electricity demand forecast results. Specifically, this includes:

[0090] Real-time power data for each target area is collected through a real-time monitoring system at power grid nodes, including current power consumption, power grid frequency, voltage, and power transmission status.

[0091] Preprocessing is performed on the collected real-time power data, and dynamic features are constructed.

[0092] The constructed feature vector is projected onto the pre-trained feature encoder, and dual matching is performed in both time and space dimensions. Specifically, this includes comparing the morphological similarity between the real-time sequence and the predicted curve through dynamic time warping, and calculating the cosine similarity between the real-time topological feature vector and the predicted scene.

[0093] Based on the matching degree between real-time data and multiple sets of prediction results, the prediction result with the highest similarity to the current power grid state is selected.

[0094] Specifically, by slicing time series data through a sliding window, dynamic features between the current moment and historical moments are constructed, and real-time data is projected onto a pre-trained feature encoder that can map the input high-dimensional feature space to a low-dimensional space.

[0095] For each set of prediction results, the similarity with the current real-time power grid state is calculated, specifically including:

[0096] The morphological similarity between real-time sequence data and the prediction result set data is obtained through dynamic time warping matching, using the following formula:

[0097]

[0098] Among them, Q i For real-time sequence data, R j The set of prediction results data;

[0099] When calculating topological cosine similarity, the real-time topological vector g t With the predicted scenario h (k) The matching calculation formula is:

[0100]

[0101] The formula for considering the similarity of current direction is:

[0102]

[0103] Based on dual matching across both temporal and spatial dimensions, a joint similarity score is calculated using the following formula:

[0104]

[0105] The weights α+β+γ=1 are obtained through optimization of historical power grid events.

[0106] See Figure 4 As shown, the calculation of the power load factor for each target region based on the electricity demand forecast results with the highest matching degree specifically includes:

[0107] Based on the electricity consumption data for each time period in the time series data of the prediction results and the total number of prediction periods, the average load and maximum load within that time period are obtained.

[0108] By inputting the calculated average load and maximum load into the load factor formula, the power load factor can be obtained.

[0109] Calculate the power load factor for each target region to obtain a time series dataset of the power load factor.

[0110] Specifically, suppose the electricity consumption for a certain period in the time series is P. t (t represents the time step. Within a specific time period, the electricity consumption sequence is [P1, P2, ..., P...).) n ], where n is the number of time steps within that period, and the average load can be calculated using the following formula:

[0111]

[0112] Among them, L avg P represents the average load during this period. t Power consumption at each time step;

[0113] The load factor is used to measure electricity demand. It represents the ratio of maximum load to average load over a period of time; a smaller load factor means greater electricity demand, and a larger load factor means lower electricity demand. The formula is:

[0114]

[0115] Where LF is the load factor, L extmax L represents the maximum load during that period. extavg This represents the average load during that period.

[0116] For multiple target regions, we can calculate the power load factor for each region using the method described above, generating a time series dataset of the load factor for each region, denoted as:

[0117] P extregion1 =[P 1,1 ,P 1,2 ,…,P 1,n ],P extregion2 =[P 2,1 ,P 2,2 ,…,P 2,n ],…

[0118] Among them, P extregion1 P extregion2 The data represents the electricity demand sequence for each region. For each region, the average load and maximum load within the time period are calculated. Then, the load factor is calculated using the formula above, and finally, the time series dataset of the load factor for each region can be obtained.

[0119] See Figure 5 As shown, based on the power load coefficient of each target region, a regional load difference matrix is ​​constructed to obtain the set of power resource exporting regions and the set of power resource importing regions, specifically including:

[0120] Based on the acquired time series dataset of electricity load coefficients, the comprehensive load average for all target regions is calculated;

[0121] Based on the comprehensive load average, the difference between the power load of each target area and the comprehensive load average is calculated, and a regional load difference matrix is ​​constructed.

[0122] Based on the difference values ​​in the regional load difference matrix, a difference judgment threshold is set to classify and identify the regions that are transferred out and regions that are transferred in.

[0123] Based on the degree of difference, the regions being transferred out and the regions being transferred in are sorted, with positive numbers representing regions being transferred in and negative numbers representing regions being transferred out, thus establishing a priority sequence for scheduling and docking.

[0124] Specifically, the comprehensive load average is obtained by weighting the load factors of all regions. For the m-th time period, the formula for calculating the comprehensive load average of all target regions is as follows:

[0125]

[0126] in, The average comprehensive load for the m-th time period is... Let be the load factor of the i-th region during the m-th time period;

[0127] Calculate the difference between the power load and the overall average load for each target region. The difference represents the degree of deviation between the load factor of each region and the overall average load. For the i-th region, the formula for calculating the difference in the m-th time period is:

[0128]

[0129] in, Let be the difference value of the i-th region in the m-th time period. Let be the load factor of the i-th region during the m-th time period. Let m be the average comprehensive load during the m-th time period;

[0130] Based on the obtained difference values ​​of each target region, a regional load difference matrix is ​​constructed. By setting a difference threshold, regions with high load and low load are divided. According to the degree of load difference between regions, the regions to be transferred in and the regions to be transferred out are sorted. The regions to be transferred in and the regions to be transferred out will be prioritized according to the size of their load difference. Positive numbers represent regions to be transferred in, and negative numbers represent regions to be transferred out.

[0131] See Figure 6 As shown, based on the acquired set of outgoing and incoming regions, the scheduling cycle is divided into multiple time periods, and power resource scheduling is carried out through incremental power transfer, specifically including:

[0132] The power dispatch cycle is divided into hours, days, and weeks, and adjusted according to actual power demand.

[0133] Based on the constructed scheduling and docking priority sequence, the regions to be transferred are sorted from largest to smallest according to the difference value, and the regions to be transferred out are sorted from smallest to largest. The two target regions with the closest absolute difference value are selected.

[0134] Based on the pre-defined scheduling cycle, the sending region will provide power during the specified time period, and the receiving region will receive power during the specified time period.

[0135] Based on load differences and changes in electricity demand, the amount of electricity dispatched in each time period is dynamically adjusted using a greedy algorithm.

[0136] Real-time monitoring of power grid operation, actual load and dispatch effectiveness in various regions, and retraining of the power demand forecasting model through a feedback mechanism to obtain stable forecast results.

[0137] Specifically, power dispatch cycles can be divided into hours, days, and weeks to adapt to power demand at different time scales. Within each cycle, the power dispatch volume will be optimized based on demand and the grid's carrying capacity. Hourly dispatch is suitable for short-term power load fluctuations and is typically used for peak-valley regulation; daily dispatch is adjusted according to daily power demand and is suitable for daily power demand forecasting; weekly dispatch is suitable for long-term load forecasting, considering seasonality, weekend effects, etc. The formula for the dynamic adjustment rule is:

[0138]

[0139] Where ΔT is the magnitude of the scheduling cycle adjustment, used to determine whether the scheduling cycle needs to be re-divided, and α is a sensitivity coefficient used to control the sensitivity of the adjustment. L is the predicted load value for time period t. t-1 The actual load value for time period t-1 is used; when ΔT > 5%, cycle re-division is triggered.

[0140] Within each dispatch cycle, based on the load differences of each region, the sending region provides power and the receiving region receives power. If dispatch is required within a certain period of time, the power allocation between the sending and receiving regions will be dynamically adjusted according to the load differences.

[0141] Using a greedy algorithm, the power dispatch quantity is adjusted according to the differences in power demand in each time period. By selecting the current optimal dispatch quantity each time, the power demand in each cycle is reasonably met.

[0142] By monitoring the electricity demand and actual load in various regions in real time and continuously adjusting the power dispatch strategy based on the feedback results, the electricity demand forecasting model needs to be retrained after each cycle to adapt to new demand changes.

[0143] Furthermore, the method according to the embodiments of this application can also be achieved by means of... Figure 7 The architecture of the electronic device shown is used to implement this. For example... Figure 7As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to a network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store the power resource scheduling and load balancing optimization method and system combining big data provided in this application. The electronic device 500 may also include a terminal interface 508. Of course, Figure 7 The architecture shown is merely exemplary and can be omitted as needed when implementing different devices. Figure 7 One or more components in the illustrated electronic device.

[0144] Figure 8 This is a schematic diagram of a computer-readable storage medium structure provided in one embodiment of this application. Figure 8 The diagram illustrates a computer-readable storage medium 600 according to one embodiment of this application. The computer-readable storage medium 600 stores computer-readable instructions. When executed by a processor, the computer-readable instructions can perform the power resource scheduling and load balancing optimization method and system incorporating big data, as described above with reference to the accompanying drawings, according to an embodiment of this application. The storage medium 600 includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0145] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0146] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0147] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A power resource scheduling and load balancing optimization method combining big data, characterized in that, include: A power demand forecasting model is built based on historical electricity consumption data, and a set of multiple electricity demand forecast results for various target regions in the future time period is generated. Based on real-time data collected from power grid nodes in each target region, the data is input into the power demand forecasting model to obtain the forecast result with the highest matching degree among multiple sets of power demand forecast results. The power load factor of each target region is calculated based on the electricity demand forecast results with the highest matching degree for each target region. Based on the power load coefficient of each target region, a regional load difference matrix is ​​constructed to obtain the set of power resource outflow regions and the set of power resource inflow regions. Based on the acquired set of outgoing and incoming regions, the scheduling cycle is divided into multiple time periods, and power resource scheduling is carried out through incremental power transfer. Based on the real-time collected data of each target region after power resource dispatch, the data is input into the power demand forecasting model and the above steps are repeated until the median of the obtained regional load difference matrix is ​​less than the dispatch threshold.

2. The power resource scheduling and load balancing optimization method combining big data as described in claim 1, characterized in that, The process of constructing a power demand forecasting model based on historical electricity consumption data and generating a set of multiple electricity demand forecast results for various target regions over future time periods specifically includes: Data preprocessing is performed based on the acquired historical data, and time series features and spatial features are constructed. The acquired feature data is input into an LSTM-based deep model for training, and the accuracy of the model is evaluated using cross-validation. Based on the trained model, the electricity demand of each target region at different time periods is predicted, including short-term demand forecast and long-term demand forecast. Based on the prediction results output by the model, a set of electricity demand prediction results for different confidence intervals is generated.

3. The power resource scheduling and load balancing optimization method combining big data as described in claim 1, characterized in that, The process of collecting real-time data from power grid nodes in each target region, inputting it into the electricity demand forecasting model, and obtaining the forecast result with the highest matching degree from a set of multiple electricity demand forecast results specifically includes: Real-time power data for each target area is collected through a real-time monitoring system at power grid nodes, including current power consumption, power grid frequency, voltage, and power transmission status. Preprocessing is performed on the collected real-time power data, and dynamic features are constructed. The constructed feature vector is projected onto the pre-trained feature encoder, and dual matching is performed in both time and space dimensions. Specifically, this includes comparing the morphological similarity between the real-time sequence and the predicted curve through dynamic time warping, and calculating the cosine similarity between the real-time topological feature vector and the predicted scene. Based on the matching degree between real-time data and multiple sets of prediction results, the prediction result with the highest similarity to the current power grid state is selected.

4. The power resource scheduling and load balancing optimization method combining big data as described in claim 1, characterized in that, The calculation of the power load factor for each target region based on the electricity demand forecast results with the highest matching degree for each target region specifically includes: Based on the electricity consumption data for each time period in the time series data of the prediction results and the total number of prediction periods, the average load and maximum load within that time period are obtained. By inputting the calculated average load and maximum load into the load factor formula, the power load factor can be obtained. Calculate the power load factor for each target region to obtain a time series dataset of the power load factor.

5. The power resource scheduling and load balancing optimization method combining big data according to claim 1, characterized in that, The process of constructing a regional load difference matrix based on the power load coefficient of each target region and obtaining the set of power resource exporting regions and the set of power resource importing regions specifically includes: Based on the acquired time series dataset of electricity load coefficients, the comprehensive load average for all target regions is calculated; Based on the comprehensive load average, the difference between the power load of each target area and the comprehensive load average is calculated, and a regional load difference matrix is ​​constructed. Based on the difference values ​​in the regional load difference matrix, a difference judgment threshold is set to classify and identify the regions that are transferred out and regions that are transferred in. Based on the degree of difference, the regions being transferred out and the regions being transferred in are sorted, with positive numbers representing regions being transferred in and negative numbers representing regions being transferred out, thus establishing a priority sequence for scheduling and docking.

6. The power resource scheduling and load balancing optimization method combining big data as described in claim 1, characterized in that, The process of dividing the scheduling cycle into multiple time periods based on the acquired set of outgoing and incoming regions, and scheduling power resources through incremental power transfer, specifically includes: The power dispatch cycle is divided into hours, days, and weeks, and adjusted according to actual power demand. Based on the constructed scheduling and docking priority sequence, the regions to be transferred are sorted from largest to smallest according to the difference value, and the regions to be transferred out are sorted from smallest to largest. The two target regions with the closest absolute difference value are selected. Based on the pre-defined scheduling cycle, the sending region will provide power during the specified time period, and the receiving region will receive power during the specified time period. Based on load differences and changes in electricity demand, the amount of electricity dispatched in each time period is dynamically adjusted using a greedy algorithm. Real-time monitoring of power grid operation, actual load and dispatch effectiveness in various regions, and retraining of the power demand forecasting model through a feedback mechanism to obtain stable forecast results.

7. A power resource scheduling and load balancing optimization system combining big data, used to implement the power resource scheduling and load balancing optimization method combining big data as described in any one of claims 1-6, characterized in that, include: Data acquisition module: The data acquisition module is mainly responsible for collecting key operational data such as electricity consumption, voltage, current, and frequency of power grid nodes in each target area in real time; Data processing module: The data processing module is mainly used to clean, normalize and extract features from historical electricity consumption data to provide standardized input for subsequent modeling; Electricity demand forecasting module: The electricity demand forecasting module is mainly used to make short-term and long-term forecasts of electricity demand in the future based on the LSTM model, and generate a set of forecast results under multiple confidence intervals. Similarity matching module: The similarity matching module is mainly used to perform dual matching of real-time collected data and prediction set with time series and spatial features, and select the optimal prediction result under the current state; Load factor calculation module: The load factor calculation module is mainly used to calculate the average load and maximum load of each target area based on the selected prediction results, and to obtain the load factor time series; Regional Difference Analysis Module: The regional difference analysis module is mainly used to construct a regional load difference matrix, identify the regions where power resources are transferred out and transferred in, and establish a scheduling priority list based on load differences; Power resource dispatching module: The power resource dispatching module is mainly used to divide the dispatching cycle and use a greedy algorithm to formulate the optimal incremental power dispatching strategy to complete the transfer of power from the sending area to the receiving area; Feedback module: The feedback module is mainly used to evaluate the scheduling effect and update the power forecasting model based on the scheduling results and real-time operation data; Processor: The processor is mainly used for the calculation process of each formula and the construction calculation process of each model.

8. An electronic device, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to perform the power resource scheduling and load balancing optimization method combining big data as described in any one of claims 1-6.

9. A computer-readable storage medium storing computer-readable instructions, characterized in that, When the computer-readable instructions are executed by the processor, they implement the power resource scheduling and load balancing optimization method combining big data as described in any one of claims 1-6.

Images