Power distribution cabinet intelligent configuration method and device based on power utilization curve analysis
By collecting and analyzing electricity consumption curves and utilizing dynamic time warping and prediction models, the problem of unsuitability for load changes in distribution cabinet configuration has been solved, achieving accurate prediction and dynamic parameter adjustment, thereby improving the operating efficiency and intelligence level of the distribution cabinet.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-25
- Publication Date
- 2026-03-13
AI Technical Summary
Existing distribution cabinet configuration schemes fail to fully consider the dynamic changes in electrical load, leading to false triggering of overload protection, overheating of lines, or equipment damage, and failing to meet the reliability and economic requirements of smart grids.
By collecting basic power, environmental, and equipment status data, historical and real-time power consumption curves are generated. Relevant curves are selected using dynamic time warping methods, a benchmark prediction curve is constructed, and a prediction model is matched and adjusted to parameters for future prediction periods. This solves technical problems and improves the operating efficiency and intelligence level of the distribution cabinet.
It improves the accuracy of load forecasting, reduces ineffective energy consumption and idle capacity, avoids overload risks, and enhances the operating efficiency and safety of the distribution cabinet.
Smart Images
Figure CN121663343A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of intelligent configuration technology for power distribution cabinets, and in particular to an intelligent configuration method and device for power distribution cabinets based on power consumption curve analysis. Background Technology
[0002] In the current operation and management of distribution cabinets, parameter configuration often relies on prior experience or fixed design standards, without fully considering the dynamic characteristics of power load changes. For example, industrial workshops have high start-up and shutdown frequencies for production equipment, resulting in drastic fluctuations in power load. Fixed overload protection thresholds are prone to false triggering during sudden load increases, leading to production interruptions. Residential communities exhibit significant seasonal power consumption characteristics, with air conditioning loads surging in summer and heating loads rising in winter. Traditional configurations cannot predict and adjust capacity allocation in advance, easily causing line overheating or equipment damage.
[0003] Furthermore, with the advancement of smart grid construction, users' demands for the reliability, economy, and intelligence of power supply are constantly increasing. The technological gaps in traditional distribution cabinet configuration schemes regarding dynamic adjustment and accurate forecasting have become a key bottleneck restricting the efficient operation of the power system. Against this backdrop, there is an urgent need for an intelligent configuration method capable of accurate load forecasting and dynamic parameter adjustment to address the inherent shortcomings of traditional schemes and improve the operational efficiency and intelligence level of distribution cabinets. Summary of the Invention
[0004] Therefore, it is necessary to provide a method and device for intelligent configuration of power distribution cabinets based on power consumption curve analysis, which can improve the operating efficiency of power distribution cabinets and address the aforementioned technical problems.
[0005] In a first aspect, this application provides a method for intelligent configuration of power distribution cabinets based on electricity consumption curve analysis, the method comprising: Collect basic power data, environmental data, and equipment status data, and generate historical power consumption curves and real-time power consumption curves based on the basic power data, environmental data, equipment status data, and preset prediction period. The historical power consumption curve reflects the power consumption trend in the past prediction period, and the real-time power consumption curve reflects the dynamic changes in power consumption at the current moment. Using the real-time electricity consumption curve as the baseline sequence, each curve in the historical electricity consumption curve is used as a candidate sequence. The similarity between the baseline sequence and each candidate sequence is calculated using the dynamic time warping method, and multiple historical electricity consumption curves that are highly correlated with the real-time electricity consumption curve are selected. For the selected historical electricity consumption curves, the load average value is taken according to the corresponding time to obtain the benchmark prediction value for each time and construct the benchmark prediction curve. Then, according to the preset prediction period, the corresponding prediction model is matched. Based on the deviation between the actual load data in the historical period and the predicted data of the corresponding benchmark prediction curve, the deviation prediction value in the future prediction period is obtained through the matching prediction model, and the final load prediction value is obtained based on the deviation prediction value and the benchmark prediction value. Based on the final load forecast, adjust the key parameters in the distribution cabinet.
[0006] In one embodiment, the step of using the real-time electricity consumption curve as a reference sequence, taking each curve in the historical electricity consumption curve as a candidate sequence, calculating the similarity between the reference sequence and each candidate sequence using a dynamic time warping method, and filtering out multiple historical electricity consumption curves that are highly correlated with the real-time electricity consumption curve includes: Using the real-time electricity consumption curve as the reference sequence, each curve in the historical electricity consumption curve is used as a candidate sequence. Data preprocessing is performed on the reference sequence and the candidate sequence to remove invalid data points and fill in missing data. Based on the preprocessed baseline sequence and candidate sequence, a distance matrix is constructed to quantitatively characterize the degree of difference between the data points corresponding to the real-time electricity consumption curve and the historical electricity consumption curve. The optimal matching path is found from the distance matrix by using a dynamic time warping method, and the overall similarity metric between the real-time electricity consumption curve and the historical electricity consumption curve is calculated. Based on the overall similarity metric, multiple historical electricity consumption curves that are highly correlated with the real-time electricity consumption curve are selected.
[0007] In one embodiment, the step of obtaining a deviation prediction value for the future prediction period based on the deviation between the actual load data in the historical period and the prediction data corresponding to the benchmark prediction curve, using a matched prediction model, and obtaining the final load prediction value based on the deviation prediction value and the benchmark prediction value includes: The difference between the actual load data and the predicted data corresponding to the benchmark prediction curve within the historical period is calculated to obtain the historical deviation at each time point, and the historical deviation is arranged in chronological order to form a deviation sequence. The deviation sequence is input into the matching prediction model to obtain the deviation prediction value at each time point in the future prediction period. The final load forecast value is calculated by adding the deviation forecast value to the baseline forecast value.
[0008] In one embodiment, inputting the deviation sequence into a matching prediction model to obtain the deviation prediction values at each time point within the future prediction period includes: Trend features were extracted from the deviation sequence according to different time dimensions; Based on the extracted trend features, the type of change trend of the deviation sequence is determined, and the type of change trend includes upward trend and downward trend; The deviation sequence and trend analysis results are input into the matching prediction model to obtain the deviation prediction values at each time point within the future prediction period.
[0009] In one embodiment, after arranging the historical deviations in chronological order to form a deviation sequence, the method further includes: Outliers in the deviation sequence are identified using statistical anomaly detection methods. Based on the cause of the outlier, the corresponding outlier correction strategy is used to replace the outlier.
[0010] In one embodiment, matching the corresponding prediction model according to a preset prediction period includes: When the preset prediction period is a short-term prediction period, an LSTM prediction model with the ability to capture short-term features of time series data is matched. The model input is the recent deviation sequence, and the output is the deviation prediction value in the future short-term prediction period. When the preset prediction period is the medium-term prediction period, the GRU prediction model that combines periodic features is matched. The model input is the medium-term deviation sequence, and the output is the deviation prediction value within the future medium-term prediction period. When the preset forecast period is a long-term forecast period, the ARIMA forecast model, which focuses on capturing long-term trends, is matched. The model input is a long-term deviation sequence, and the output is the predicted deviation value within the future long-term forecast period.
[0011] In one embodiment, after adjusting the key parameters in the distribution cabinet based on the final load forecast value, the method further includes: After parameter adjustment, the actual load data of the power distribution cabinet is collected in real time; The load forecasting error is calculated based on the collected data. Based on the load forecasting error, the corresponding error level is determined, and the corresponding parameter optimization strategy is executed. Establish a parameter configuration database to record different parameter configuration schemes and load forecasting errors, forming reusable parameter configuration templates.
[0012] Secondly, this application also provides an intelligent configuration device for power distribution cabinets based on electricity consumption curve analysis. The device includes: The data acquisition module is used to collect basic power data, environmental data, and equipment status data, and generate historical power consumption curves and real-time power consumption curves based on the basic power data, environmental data, equipment status data, and a preset prediction period. The historical power consumption curve reflects the power consumption trend in the past prediction period, and the real-time power consumption curve reflects the dynamic changes in power consumption at the current moment. The candidate sequence filtering module is used to take the real-time electricity consumption curve as the reference sequence, take each curve in the historical electricity consumption curve as a candidate sequence, calculate the similarity between the reference sequence and each candidate sequence through the dynamic time warping method, and filter out multiple historical electricity consumption curves that are highly related to the real-time electricity consumption curve. The prediction curve construction module is used to take the load average at the corresponding time point from multiple selected historical electricity consumption curves, obtain the benchmark prediction value at each time point, construct the benchmark prediction curve, and match the corresponding prediction model according to the preset prediction period. The load forecasting module is used to obtain the deviation forecast value for the future forecast period based on the deviation between the actual load data in the historical period and the forecast data corresponding to the benchmark forecast curve, through a matching forecast model, and to obtain the final load forecast value based on the deviation forecast value and the benchmark forecast value. The parameter adjustment module is used to adjust the key parameters in the distribution cabinet according to the final load forecast value.
[0013] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps: Collect basic power data, environmental data, and equipment status data, and generate historical power consumption curves and real-time power consumption curves based on the basic power data, environmental data, equipment status data, and preset prediction period. The historical power consumption curve reflects the power consumption trend in the past prediction period, and the real-time power consumption curve reflects the dynamic changes in power consumption at the current moment. Using the real-time electricity consumption curve as the baseline sequence, each curve in the historical electricity consumption curve is used as a candidate sequence. The similarity between the baseline sequence and each candidate sequence is calculated using the dynamic time warping method, and multiple historical electricity consumption curves that are highly correlated with the real-time electricity consumption curve are selected. For the selected historical electricity consumption curves, the load average value is taken according to the corresponding time to obtain the benchmark prediction value for each time and construct the benchmark prediction curve. Then, according to the preset prediction period, the corresponding prediction model is matched. Based on the deviation between the actual load data in the historical period and the predicted data of the corresponding benchmark prediction curve, the deviation prediction value in the future prediction period is obtained through the matching prediction model, and the final load prediction value is obtained based on the deviation prediction value and the benchmark prediction value. Based on the final load forecast, adjust the key parameters in the distribution cabinet.
[0014] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps: Collect basic power data, environmental data, and equipment status data, and generate historical power consumption curves and real-time power consumption curves based on the basic power data, environmental data, equipment status data, and preset prediction period. The historical power consumption curve reflects the power consumption trend in the past prediction period, and the real-time power consumption curve reflects the dynamic changes in power consumption at the current moment. Using the real-time electricity consumption curve as the baseline sequence, each curve in the historical electricity consumption curve is used as a candidate sequence. The similarity between the baseline sequence and each candidate sequence is calculated using the dynamic time warping method, and multiple historical electricity consumption curves that are highly correlated with the real-time electricity consumption curve are selected. For the selected historical electricity consumption curves, the load average value is taken according to the corresponding time to obtain the benchmark prediction value for each time and construct the benchmark prediction curve. Then, according to the preset prediction period, the corresponding prediction model is matched. Based on the deviation between the actual load data in the historical period and the predicted data of the corresponding benchmark prediction curve, the deviation prediction value in the future prediction period is obtained through the matching prediction model, and the final load prediction value is obtained based on the deviation prediction value and the benchmark prediction value. Based on the final load forecast, adjust the key parameters in the distribution cabinet.
[0015] In summary, this application includes the following beneficial technical effects: First, basic power data, environmental data, and equipment status data are collected to ensure that the power consumption curves fully reflect the actual power consumption scenario. Using the real-time power consumption curves as a benchmark sequence, a dynamic time warping method is used to select multiple historical power consumption curves that are highly correlated with the real-time power consumption curves. This overcomes the limitations of traditional fixed-window matching and eliminates interference from irrelevant data. By quantitatively analyzing the deviation between the actual load and the benchmark prediction curve within historical periods, the potential deviation magnitude in future periods can be predicted in advance. This allows for correction based on the benchmark prediction value, improving the accuracy of load prediction. Based on the final load prediction value, key parameters within the distribution cabinet are adjusted to reduce ineffective energy consumption, idle capacity, and overload risks, thereby improving the operating efficiency of the distribution cabinet while ensuring its safe operation. Attached Figure Description
[0016] Figure 1 This is a flowchart illustrating a method for intelligent configuration of power distribution cabinets based on electricity consumption curve analysis in one embodiment. Figure 2 This is a flowchart illustrating a power distribution cabinet intelligent configuration method based on electricity consumption curve analysis in another embodiment. Figure 3 This is a structural block diagram of a power distribution cabinet intelligent configuration device based on power consumption curve analysis in one embodiment. Detailed Implementation
[0017] This invention provides a method and device for intelligent configuration of power distribution cabinets based on power consumption curve analysis.
[0018] The embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the accompanying drawings and embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.
[0019] In the description of the embodiments disclosed in this invention, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc., may refer to different or the same objects. Other explicit and implicit definitions may also be included below.
[0020] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the intelligent configuration method for power distribution cabinets based on electricity consumption curve analysis in this invention includes: The S100 collects basic power data, environmental data, and equipment status data, and generates historical and real-time power consumption curves based on the basic power data, environmental data, equipment status data, and preset prediction cycles.
[0021] Specifically, basic power data, environmental data, and equipment status data are collected to ensure that the power curves comprehensively reflect actual power consumption. Basic power data includes voltage, current, power, and energy consumption, collected in real-time by sensors built into the distribution cabinet. Environmental data covers ambient temperature, humidity, and light intensity, collected by environmental sensors deployed in the distribution cabinet or power consumption area. Equipment status data includes the operating status of electrical equipment (e.g., on / off, standby / full load) and cumulative operating time, obtained through the distribution cabinet's communication interface. After data collection, historical and real-time power consumption curves are generated. For historical power consumption curves, data from N time periods consistent with the preset prediction period (e.g., if the preset prediction period is 24 hours, then 24-hour power consumption data from each day of the past 30 days are selected) are selected and integrated into multiple historical power consumption curves in chronological order to reflect power consumption trends within past periods. Real-time power consumption curves select real-time data from one prediction period prior to the current moment (e.g., if the preset prediction period is 24 hours, then real-time power consumption data from the 24 hours before the current moment) to reflect the dynamic changes in power consumption at the current moment.
[0022] S200 uses the real-time electricity consumption curve as the baseline sequence, takes each curve in the historical electricity consumption curve as a candidate sequence, calculates the similarity between the baseline sequence and each candidate sequence through the dynamic time warping method, and selects multiple historical electricity consumption curves that are highly correlated with the real-time electricity consumption curve.
[0023] Specifically, since electricity consumption curves from different historical periods may have time axis offsets, traditional Euclidean distance calculations cannot accurately measure curve similarity. Therefore, a dynamic time warping method is adopted. This dynamic time warping method is an algorithm used to measure the similarity between two time series that may have different lengths. It non-linearly stretches or compresses the sequences on the time axis to make the two sequences as aligned as possible in shape, thereby obtaining the minimum cumulative distance. Specifically, the real-time electricity consumption curve is used as the baseline sequence, and each curve in the historical electricity consumption curves is used as a candidate sequence. A distance matrix is constructed based on the baseline sequence and the candidate sequences. An optimal matching path is found from the distance matrix, where the sum of the distances of all elements on the path is minimized. The minimum sum of distances is the overall similarity measure between the real-time electricity consumption curve and the historical electricity consumption curve. A similarity threshold is set to filter out multiple historical electricity consumption curves that are highly correlated with the real-time electricity consumption curve.
[0024] In this embodiment, by using real-time electricity consumption curves as a reference sequence, and utilizing dynamic time warping to calculate the similarity between the reference sequence and each historical electricity consumption curve, and selecting multiple highly correlated historical curves, the correlation of the model input data can be significantly improved, noise interference can be reduced, and thus the accuracy of load forecasting can be improved.
[0025] S300 takes the load average value at the corresponding time for multiple selected historical electricity consumption curves, obtains the benchmark prediction value for each time, constructs the benchmark prediction curve, and matches the corresponding prediction model according to the preset prediction period.
[0026] Specifically, the selected historical electricity consumption curves are statistically integrated to eliminate random fluctuations in individual curves. The average load power at corresponding times (e.g., the 10th minute and 20th minute of each curve) is taken to obtain the baseline prediction value. For example, if the loads of the three historical curves at 14:00 are 120kW, 125kW, and 118kW respectively, the baseline prediction value is (120+125+118) / 3=121kW. The baseline prediction values at each time point are connected in chronological order to form a baseline prediction curve reflecting the average electricity consumption trend. At the same time, according to the preset prediction period, a prediction model with the ability to capture corresponding time-series characteristics is matched. The longer the period, the more the model focuses on long-term trends; the shorter the period, the more the model focuses on short-term fluctuations.
[0027] In this embodiment, by taking the load average of multiple historical electricity consumption curves at the same time, noise and abnormal fluctuations in a single curve can be effectively suppressed, making the baseline prediction value closer to the statistical characteristics of the actual load; according to the preset prediction period, the corresponding prediction model is matched, realizing flexible model switching.
[0028] S400 uses a matching prediction model to obtain the deviation prediction value for the future prediction period based on the deviation prediction value and the baseline prediction value, according to the deviation prediction value and the baseline prediction value.
[0029] Specifically, the baseline forecast curve reflects the average trend and needs further correction to correct the deviation between the forecast data and the actual load. First, extract the actual load data within the historical period (such as the actual electricity consumption record of a certain 24 hours in the past) and the forecast data of the corresponding baseline forecast curve, calculate the deviation at each time point (deviation = actual load - baseline forecast value), then form a deviation sequence according to the time sequence of the deviations, input it into the matching forecast model, and obtain the deviation forecast value at each time point in the future forecast period (such as predicting that the deviation at 14:00 in the next 24 hours is +5kW, which means that the actual load may be 5kW higher than the baseline forecast value). Finally, add the deviation forecast value at each time point to the baseline forecast value, that is, the final load forecast value = baseline forecast value + deviation forecast value. This value takes into account both the average trend and the dynamic deviation.
[0030] In this embodiment, the deviation between the actual load data within the historical period and the predicted data of the corresponding benchmark prediction curve is used as an independent prediction target, so that the final load prediction can be corrected based on the benchmark prediction, thereby reducing the overall error.
[0031] S500 adjusts key parameters within the distribution cabinet based on the final load forecast.
[0032] Specifically, the key parameters to be adjusted include overload protection threshold, capacity allocation ratio, and heat dissipation control parameters. The overload protection threshold is set at 1.2 times the final load forecast value to avoid false triggering due to sudden load increases. The capacity allocation ratio is allocated according to the proportion of the final load forecast value of each circuit. The heat dissipation control parameters are adjusted based on the final load forecast value and the rated capacity. If the final load forecast value is higher than 80% of the rated capacity, the cooling fan speed will be increased in advance to avoid the cabinet temperature from becoming too high.
[0033] In one embodiment, such as Figure 2 As shown, S200 includes: S210: Using the real-time electricity consumption curve as the reference sequence, each curve in the historical electricity consumption curve is used as a candidate sequence. Data preprocessing is performed on the reference sequence and candidate sequences to remove invalid data points and fill in missing data. S220: Based on the preprocessed baseline sequence and candidate sequence, a distance matrix is constructed to quantitatively characterize the degree of difference between the corresponding data points of the real-time electricity consumption curve and the historical electricity consumption curve. S230 uses the dynamic time warping method to find the optimal matching path from the distance matrix and calculates the overall similarity metric between the real-time electricity consumption curve and the historical electricity consumption curve. S240, based on the overall similarity metric, selects multiple historical electricity consumption curves that are highly correlated with the real-time electricity consumption curve.
[0034] Specifically, data preprocessing is performed on the baseline and candidate sequences: invalid data points (such as data exceeding the normal load range by 10 times due to sensor failure) are identified and removed using the Laida criterion; missing data (such as data missing for 5 consecutive minutes due to communication interruption) are supplemented using linear interpolation to ensure the integrity and validity of the sequence data; the Euclidean distance d(i,j)=|Q(i)-C(j)| between each data point Q(i) of the baseline sequence and each data point C(j) of the candidate sequence is calculated to form an m×n distance matrix, and each element in the matrix is quantized. The degree of difference between the corresponding data points of the two curves is analyzed. The optimal matching path is found from the distance matrix using a dynamic time warping method. Specifically, a path is found from the top left to the bottom right corner of the distance matrix that satisfies boundary constraints and monotonic constraints (the path extends from left to right and from top to bottom; that is, if a point on the path is (i,j), the next point can only be (i+1,j), (i,j+1), or (i+1,j+1), avoiding reverse matching on the time axis), and minimizes the sum of distances of all elements on the path. This minimum sum of distances is the overall similarity metric of the two curves. A similarity threshold is set, and historical electricity consumption curves with an overall similarity metric less than the threshold are judged as highly correlated. A preset number (10-20) of historical electricity consumption curves that are highly correlated with the real-time electricity consumption curve are selected.
[0035] In one embodiment, based on the deviation between actual load data and the corresponding baseline forecast curve forecast data within a historical period, a predicted deviation value for the future forecast period is obtained through a matched forecast model. The final load forecast value is then obtained based on the predicted deviation value and the baseline forecast value, including: The difference between the actual load data and the predicted data of the corresponding benchmark forecast curve within the historical period is calculated to obtain the historical deviation at each time point. The historical deviations are then arranged in chronological order to form a deviation sequence. The deviation sequence is input into the matching forecast model to obtain the deviation prediction value at each time point within the future forecast period. The deviation prediction value is added to the benchmark forecast value to calculate the final load forecast value.
[0036] Specifically, to ensure that historical deviations accurately reflect load fluctuation patterns, the alignment principle must be strictly adhered to. This means the length and granularity of the historical period must be completely consistent with the preset prediction period. For example, if the preset prediction period is 24 hours and the granularity is 1 minute, the selected historical period must also be 24 hours, and the actual historical load must be accurate to the minute, corresponding one-to-one with the minute-by-minute predicted value of the benchmark prediction curve. This avoids errors in deviation calculation due to inconsistent time granularity. Specifically, historical periods corresponding to the screened similar historical electricity consumption curves are selected. The actual load data for each moment within each historical period is obtained, and the difference is calculated between this data and the predicted data for the corresponding moment in the benchmark prediction curve (historical deviation = actual historical load value - benchmark predicted value). Then, the historical deviations within each historical period are arranged in chronological order, resulting in a continuous deviation sequence that prioritizes "highly similar historical periods and time order within the same period." This sequence reflects the patterns of historical load fluctuations. The deviation sequence is input into the matching prediction model to obtain the predicted deviation values for each moment in the future prediction period. These predicted deviation values are then added to the benchmark predicted value to obtain the final load prediction value.
[0037] In this embodiment, the historical average trend is used as a basis, and possible future fluctuations are superimposed to make the prediction results closer to the random changes in actual electricity consumption scenarios.
[0038] In one embodiment, inputting the deviation sequence into a matching prediction model to obtain the predicted deviation values for each time point within the future prediction period includes: Trend features are extracted from the deviation sequence according to different time dimensions; based on the extracted trend features, the type of change trend of the deviation sequence is determined, including upward trend and downward trend; the deviation sequence and trend analysis results are input into the matching prediction model to obtain the deviation prediction value at each time point in the future prediction period.
[0039] Specifically, trend features are extracted from the deviation sequence, and the direction and magnitude of the deviation change are analyzed according to time dimensions (such as 5 minutes, 10 minutes, and 1 hour) to determine the type of trend of the deviation sequence. If the deviation increases in three consecutive time dimensions, the trend is determined to be an upward trend. If the deviation decreases in three consecutive time dimensions, the trend is determined to be a downward trend. If the above conditions are not met, it is determined to be a stationary trend (the deviation has no obvious direction of change). The prediction model will output the deviation prediction value according to the stationary fluctuation law. Trend analysis results, as one of the input features of the prediction model, help the model adjust prediction weights. For short-term predictions (using the LSTM prediction model), if the trend analysis result shows an upward trend, the model will increase the weight of recent upward deviation data, making the predicted deviation value more inclined to continue the upward trend. For medium-term predictions (using the GRU prediction model), if the trend analysis result shows a downward trend in a certain period, the model will combine the periodic characteristics of that period to strengthen the prediction of downward deviations. For long-term predictions (using the ARIMA prediction model), if the trend analysis result shows a long-term upward trend, the model will retain this trend information during differencing to avoid losing long-term variation patterns due to stabilization processing. By combining the trend analysis results with the deviation sequence as input, the output accuracy of the prediction model is improved.
[0040] In one embodiment, after arranging the historical deviations in chronological order to form a deviation sequence, the method further includes: Outliers in the deviation sequence are identified using statistical anomaly detection methods; based on the cause of the outliers, corresponding outlier correction strategies are used to replace the outliers.
[0041] Specifically, since outliers may exist in the deviation sequence (such as sudden large deviations caused by equipment failures within historical periods), outliers need to be identified through statistical anomaly detection methods, such as box plot-based outlier detection, which identifies data exceeding 1.5 times the interquartile range as outliers. Corresponding correction strategies are adopted according to the cause of the outlier. If it is a one-time anomaly caused by equipment failure, the average of the deviations at adjacent time points is used for replacement. If it is an anomaly caused by data acquisition errors, linear interpolation is used for replacement to ensure the reliability of the deviation sequence.
[0042] In one embodiment, matching the corresponding prediction model according to a preset prediction period includes: When the preset prediction period is a short-term prediction period, an LSTM prediction model with the ability to capture short-term features of time series data is matched. The model input is the recent deviation sequence, and the output is the predicted deviation value within the future short-term prediction period. When the preset prediction period is a medium-term prediction period, a GRU prediction model that incorporates periodic features is matched. The model input is the medium-term deviation sequence, and the output is the predicted deviation value within the future medium-term prediction period. When the preset prediction period is a long-term prediction period, an ARIMA prediction model that focuses on capturing long-term trends is matched. The model input is the long-term deviation sequence, and the output is the predicted deviation value within the future long-term prediction period.
[0043] Specifically, since the load change characteristics differ across forecast periods, corresponding forecast models need to be matched based on the preset forecast period. For short-term forecast periods (e.g., 1-6 hours), load changes are mainly characterized by sudden fluctuations, requiring the forecast model to capture short-term features of time-series data. An LSTM (Long Short-Term Memory) forecast model is suitable, as it effectively handles dependencies in short-term time-series data through gating mechanisms (input gate, forget gate, output gate). The model input is the recent deviation sequence, and the output is the predicted deviation value for the next short-term forecast period. For medium-term forecast periods (e.g., 12-24 hours), load changes exhibit intraday periodicity (e.g., morning peak, evening peak), requiring the model to incorporate periodic features and be matched with a GRU (Gated Recurrent Unit) forecast model. The model is a simplified LSTM model that can efficiently capture the periodic patterns of medium-term time series data. The model input is the medium-term deviation sequence, and the output is the predicted deviation value within the future medium-term forecast period. Load changes within the long-term forecast period (e.g., 7-30 days) are significantly affected by long-term trends, so the model needs to focus on capturing long-term trends. The matching ARIMA (Autoregressive Integral Moving Average) forecast model is used. This model uses differencing to stabilize non-stationary time series data and combines autoregressive and moving average terms to capture long-term trends. The model input is the long-term deviation sequence, and the output is the predicted deviation value within the future long-term forecast period.
[0044] In one embodiment, after adjusting the key parameters within the distribution cabinet based on the final load forecast, the method further includes: After parameter adjustment, the actual load data of the distribution cabinet is collected in real time; the load prediction error is calculated based on the collected data, the corresponding error level is determined according to the load prediction error, and the corresponding parameter optimization strategy is executed; a parameter configuration database is established to record different parameter configuration schemes and load prediction errors, forming a reusable parameter configuration template.
[0045] Specifically, after adjusting the key parameters within the distribution cabinet, actual load data for each circuit of the distribution cabinet is collected. Then, the load forecast error is calculated using the Mean Absolute Percentage Error (MAPE), as shown in the formula: Error levels are categorized based on MAPE (Modulation Error Rate), and corresponding parameter optimization strategies are executed accordingly. A structured parameter configuration database is constructed, with fields including scene label, prediction period, load characteristics (such as peak time period and fluctuation range), parameter configuration scheme, and MAPE error. When a new scene has a similarity of ≥80% with the scene label and load characteristics recorded in the database (calculated using cosine similarity), the parameter scheme of that record is directly reused, significantly shortening the debugging time.
[0046] In one embodiment, such as Figure 3 As shown, a smart configuration device for a distribution cabinet based on electricity consumption curve analysis is provided, including: a data acquisition module 10, a candidate sequence screening module 20, a prediction curve construction module 30, a load prediction module 40, and a parameter adjustment module 50, wherein: The data acquisition module 10 is used to collect basic power data, environmental data and equipment status data, and generate historical power consumption curves and real-time power consumption curves based on the basic power data, environmental data, equipment status data and preset prediction period. The historical power consumption curve reflects the power consumption trend in the past prediction period, and the real-time power consumption curve reflects the dynamic changes in power consumption at the current moment. The candidate sequence filtering module 20 is used to take the real-time electricity consumption curve as the reference sequence, take each curve in the historical electricity consumption curve as a candidate sequence, calculate the similarity between the reference sequence and each candidate sequence through the dynamic time warping method, and filter out multiple historical electricity consumption curves that are highly related to the real-time electricity consumption curve. The prediction curve construction module 30 is used to take the load average value at the corresponding time for multiple historical electricity consumption curves, obtain the benchmark prediction value at each time and construct the benchmark prediction curve, and match the corresponding prediction model according to the preset prediction period. The load forecasting module 40 is used to obtain the deviation forecast value for the future forecast period based on the deviation between the actual load data in the historical period and the forecast data of the corresponding benchmark forecast curve, through a matching forecast model, and to obtain the final load forecast value based on the deviation forecast value and the benchmark forecast value. The parameter adjustment module 50 is used to adjust the key parameters in the distribution cabinet based on the final load forecast value.
[0047] In one embodiment, the candidate sequence screening module 20 is further configured to use the real-time electricity consumption curve as the baseline sequence, take each curve in the historical electricity consumption curve as a candidate sequence, perform data preprocessing on the baseline sequence and candidate sequences, remove invalid data points and fill in missing data; construct a distance matrix based on the preprocessed baseline sequence and candidate sequences to quantify the degree of difference between the corresponding data points of the real-time electricity consumption curve and the historical electricity consumption curve; find the optimal matching path from the distance matrix through the dynamic time warping method, and calculate the overall similarity metric value between the real-time electricity consumption curve and the historical electricity consumption curve; and screen out multiple historical electricity consumption curves that are highly correlated with the real-time electricity consumption curve based on the overall similarity metric value.
[0048] In one embodiment, the load forecasting module 40 is further configured to calculate the difference between the actual load data and the forecast data of the corresponding benchmark forecast curve within the historical period, obtain the historical deviation amount at each time point, and arrange the historical deviation amounts in chronological order to form a deviation sequence; input the deviation sequence into the matching forecasting model to obtain the deviation forecast value at each time point within the future forecasting period; add the deviation forecast value to the benchmark forecast value to calculate the final load forecast value.
[0049] In one embodiment, the load forecasting module 40 further extracts trend features from the deviation sequence according to different time dimensions; determines the change trend type of the deviation sequence based on the extracted trend features, including upward and downward trends; and inputs the deviation sequence and trend analysis results into a matching forecasting model to obtain the deviation forecast value at each time point in the future forecasting period.
[0050] In one embodiment, the power distribution cabinet intelligent configuration device based on power consumption curve analysis further includes an outlier correction module, which is used to identify outliers in the deviation sequence through a statistical anomaly detection method; and to replace outliers according to the cause of the outlier by adopting a corresponding outlier correction strategy.
[0051] In one embodiment, the prediction curve construction module 30 is further configured to: match an LSTM prediction model capable of capturing short-term features of time-series data when the preset prediction period is a short-term prediction period, with the model input being a recent deviation sequence and the output being the predicted deviation value within the future short-term prediction period; match a GRU prediction model incorporating periodic features when the preset prediction period is a medium-term prediction period, with the model input being a medium-term deviation sequence and the output being the predicted deviation value within the future medium-term prediction period; and match an ARIMA prediction model focusing on capturing long-term trends when the preset prediction period is a long-term prediction period, with the model input being a long-term deviation sequence and the output being the predicted deviation value within the future long-term prediction period.
[0052] In one embodiment, the intelligent configuration device for distribution cabinets based on electricity consumption curve analysis further includes a parameter optimization module, which is used to collect the actual load data of the distribution cabinet in real time after parameter adjustment; calculate the load prediction error based on the collected data; determine the corresponding error level according to the load prediction error; and execute the corresponding parameter optimization strategy; establish a parameter configuration database to record different parameter configuration schemes and load prediction errors, forming a reusable parameter configuration template.
[0053] In one embodiment, this application discloses a computer device including a memory and a processor. The memory stores a computer program that can run on the processor. When the processor loads the computer program, it executes a power distribution cabinet intelligent configuration method based on power consumption curve analysis as described in the above embodiment.
[0054] In one embodiment, this application discloses a computer-readable storage medium storing a computer program, wherein when the computer program is loaded by a processor, it executes a power distribution cabinet intelligent configuration method based on power consumption curve analysis as described in the above embodiment.
[0055] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.
Claims
1. A method for intelligent configuration of power distribution cabinets based on electricity consumption curve analysis, characterized in that, include: Collect basic power data, environmental data, and equipment status data, and generate historical power consumption curves and real-time power consumption curves based on the basic power data, environmental data, equipment status data, and preset prediction period. The historical power consumption curve reflects the power consumption trend in the past prediction period, and the real-time power consumption curve reflects the dynamic changes in power consumption at the current moment. Using the real-time electricity consumption curve as the baseline sequence, each curve in the historical electricity consumption curve is used as a candidate sequence. The similarity between the baseline sequence and each candidate sequence is calculated using the dynamic time warping method, and multiple historical electricity consumption curves that are highly correlated with the real-time electricity consumption curve are selected. For the selected historical electricity consumption curves, the load average value is taken according to the corresponding time to obtain the benchmark prediction value for each time and construct the benchmark prediction curve. Then, according to the preset prediction period, the corresponding prediction model is matched. Based on the deviation between the actual load data in the historical period and the predicted data of the corresponding benchmark prediction curve, the deviation prediction value in the future prediction period is obtained through the matching prediction model, and the final load prediction value is obtained based on the deviation prediction value and the benchmark prediction value. Based on the final load forecast, adjust the key parameters in the distribution cabinet.
2. The intelligent configuration method for power distribution cabinets based on electricity consumption curve analysis according to claim 1, characterized in that, The step of using the real-time electricity consumption curve as a reference sequence, taking each curve in the historical electricity consumption curve as a candidate sequence, and calculating the similarity between the reference sequence and each candidate sequence using a dynamic time warping method to filter out multiple historical electricity consumption curves that are highly correlated with the real-time electricity consumption curve includes: Using the real-time electricity consumption curve as the reference sequence, each curve in the historical electricity consumption curve is used as a candidate sequence. Data preprocessing is performed on the reference sequence and the candidate sequence to remove invalid data points and fill in missing data. Based on the preprocessed baseline sequence and candidate sequence, a distance matrix is constructed to quantitatively characterize the degree of difference between the data points corresponding to the real-time electricity consumption curve and the historical electricity consumption curve. The optimal matching path is found from the distance matrix by using a dynamic time warping method, and the overall similarity metric between the real-time electricity consumption curve and the historical electricity consumption curve is calculated. Based on the overall similarity metric, multiple historical electricity consumption curves that are highly correlated with the real-time electricity consumption curve are selected.
3. The intelligent configuration method for power distribution cabinets based on electricity consumption curve analysis according to claim 1, characterized in that, The step of obtaining the deviation prediction value for the future prediction period based on the deviation between the actual load data in the historical period and the prediction data corresponding to the benchmark prediction curve, through a matched prediction model, and obtaining the final load prediction value based on the deviation prediction value and the benchmark prediction value includes: The difference between the actual load data and the predicted data corresponding to the benchmark prediction curve within the historical period is calculated to obtain the historical deviation at each time point, and the historical deviation is arranged in chronological order to form a deviation sequence. The deviation sequence is input into the matching prediction model to obtain the deviation prediction value at each time point in the future prediction period. The final load forecast value is calculated by adding the deviation forecast value to the baseline forecast value.
4. The intelligent configuration method for power distribution cabinets based on electricity consumption curve analysis according to claim 3, characterized in that, The step of inputting the deviation sequence into the matching prediction model to obtain the deviation prediction values at each time point within the future prediction period includes: Trend features were extracted from the deviation sequence according to different time dimensions; Based on the extracted trend features, the type of change trend of the deviation sequence is determined, and the type of change trend includes upward trend and downward trend; The deviation sequence and trend analysis results are input into the matching prediction model to obtain the deviation prediction values at each time point within the future prediction period.
5. The intelligent configuration method for power distribution cabinets based on electricity consumption curve analysis according to claim 3, characterized in that, After arranging the historical deviations in chronological order to form a deviation sequence, the method further includes: Outliers in the deviation sequence are identified using statistical anomaly detection methods. Based on the cause of the outlier, the corresponding outlier correction strategy is used to replace the outlier.
6. The intelligent configuration method for distribution cabinets based on electricity consumption curve analysis according to claim 1, characterized in that, The step of matching the corresponding prediction model according to the preset prediction period includes: When the preset prediction period is a short-term prediction period, an LSTM prediction model with the ability to capture short-term features of time series data is matched. The model input is the recent deviation sequence, and the output is the deviation prediction value in the future short-term prediction period. When the preset prediction period is the medium-term prediction period, the GRU prediction model that combines periodic features is matched. The model input is the medium-term deviation sequence, and the output is the deviation prediction value within the future medium-term prediction period. When the preset forecast period is a long-term forecast period, the ARIMA forecast model, which focuses on capturing long-term trends, is matched. The model input is a long-term deviation sequence, and the output is the predicted deviation value within the future long-term forecast period.
7. The intelligent configuration method for power distribution cabinets based on electricity consumption curve analysis according to claim 1, characterized in that, After adjusting the key parameters in the distribution cabinet based on the final load forecast, the process also includes: After parameter adjustment, the actual load data of the power distribution cabinet is collected in real time; The load forecasting error is calculated based on the collected data. Based on the load forecasting error, the corresponding error level is determined, and the corresponding parameter optimization strategy is executed. Establish a parameter configuration database to record different parameter configuration schemes and load forecasting errors, forming reusable parameter configuration templates.
8. A smart configuration device for a power distribution cabinet based on electricity consumption curve analysis, characterized in that, include: The data acquisition module is used to collect basic power data, environmental data, and equipment status data, and generate historical power consumption curves and real-time power consumption curves based on the basic power data, environmental data, equipment status data, and a preset prediction period. The historical power consumption curve reflects the power consumption trend in the past prediction period, and the real-time power consumption curve reflects the dynamic changes in power consumption at the current moment. The candidate sequence filtering module is used to take the real-time electricity consumption curve as the reference sequence, take each curve in the historical electricity consumption curve as a candidate sequence, calculate the similarity between the reference sequence and each candidate sequence through the dynamic time warping method, and filter out multiple historical electricity consumption curves that are highly related to the real-time electricity consumption curve. The prediction curve construction module is used to take the load average at the corresponding time point from multiple selected historical electricity consumption curves, obtain the benchmark prediction value at each time point, construct the benchmark prediction curve, and match the corresponding prediction model according to the preset prediction period. The load forecasting module is used to obtain the deviation forecast value for the future forecast period based on the deviation between the actual load data in the historical period and the forecast data corresponding to the benchmark forecast curve, through a matching forecast model, and to obtain the final load forecast value based on the deviation forecast value and the benchmark forecast value. The parameter adjustment module is used to adjust the key parameters in the distribution cabinet according to the final load forecast value.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.