Electric power marketing fusion management method and platform based on intelligent data platform

Abstract

The invention discloses an electric power marketing fusion management method and platform based on an intelligent data center, and relates to the technical field of data management. The method comprises the following steps: acquiring electric power data of a target main body through an intelligent data medium station, and respectively performing basic anomaly analysis and load change anomaly analysis to obtain a basic data anomaly degree and a change data anomaly degree; calculating the abnormal degree of fusion data and configuring basic load prediction resources; calculating a load fluctuation degree and average power data based on the historical power data, obtaining a load prediction weight and a data record weight, and adjusting the basic load prediction resource to obtain a final load prediction resource; and dynamically calling a plurality of agents in the load prediction agent sequence based on the load prediction resource to carry out integrated prediction, obtaining a predicted load and completing data record management. According to the invention, closed-loop management of data quality evaluation, dynamic allocation of prediction resources and load prediction is realized, and the accuracy and efficiency of power marketing management are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of data management technology, specifically to a method and platform for integrated management of power marketing based on an intelligent data platform. Background Technology

[0002] In power management, the accuracy of load forecasting directly impacts the efficiency of power resource allocation and the effectiveness of marketing strategies. Traditional forecasting methods largely rely on statistical analysis of historical load data, failing to fully consider the intrinsic correlation between real-time data quality and load change trends. Power data may exhibit numerical anomalies during the acquisition phase due to metering equipment malfunctions or communication interference, while on the consumer side, load fluctuations may be caused by changes in user behavior or unforeseen events.

[0003] Existing technical solutions typically treat data quality inspection and load forecasting as two separate processes, lacking a unified anomaly assessment framework. This results in the prediction model being unable to adaptively adjust its computing resources and prediction strategies when data anomalies occur, causing the prediction results to deviate from the actual situation. Summary of the Invention

[0004] This invention addresses the technical problems in existing technologies, such as the simplistic assessment of power data anomalies, rigid resource allocation for forecasting, and the difficulty in balancing load forecasting accuracy with data management efficiency. It provides a power marketing integrated management method and platform based on an intelligent data middleware platform.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: In a first aspect, the present invention provides a power marketing integrated management method based on an intelligent data platform, comprising: Through the intelligent data platform, power data of target entities within the power system are collected, basic anomaly analysis of power data is performed, and the degree of anomaly of basic data is obtained. Based on the historical predicted load in the previous time period, load change anomaly analysis is performed on the power data to obtain the anomaly degree of the change data. Based on the anomaly degree of the basic data and the anomaly degree of the change data, configure the basic load forecasting resources for load forecasting based on the power data; Based on the historical power data of the target entity, load forecasting weights and data recording weights are allocated, the basic load forecasting resources are adjusted, load forecasting resources are obtained, the forecast load is predicted, and data recording management is performed.

[0006] Secondly, this invention provides a power marketing integrated management platform based on an intelligent data middleware platform, comprising: The data acquisition and basic anomaly analysis module is used to collect power data of target entities within the power system through an intelligent data platform, and to perform basic anomaly analysis of the power data to obtain the anomaly degree of the basic data. The load change anomaly analysis module is used to perform load change anomaly analysis on the power data based on the historical predicted load in the previous time period to obtain the anomaly degree of the change data. The basic forecasting resource configuration module is used to configure the basic load forecasting resources for load forecasting based on the power data, according to the basic data anomaly degree and the change data anomaly degree. The load forecasting and data recording management module is used to allocate load forecasting weights and data recording weights based on the historical power data of the target entity, adjust the basic load forecasting resources, obtain load forecasting resources, forecast the load, and manage data recording.

[0007] Thirdly, the present invention provides an electronic device, comprising: Memory, used to store computer software programs; The processor is used to read and execute the computer software program, thereby realizing the power marketing integration management method based on the intelligent data platform described in the first aspect.

[0008] Fourthly, the present invention provides a non-transitory computer-readable storage medium, comprising: a computer software program stored in the storage medium, wherein the computer software program, when executed by a processor, implements the power marketing integration management method based on an intelligent data platform as described in the first aspect.

[0009] The beneficial effects of this invention are: Compared to existing technologies, this invention first establishes a dual anomaly analysis mechanism to comprehensively assess both static data quality and dynamic load changes, providing a more reliable data foundation for forecasting. Secondly, it dynamically adjusts forecasting resource allocation based on anomaly analysis results, ensuring precise matching between resource allocation and data quality status, significantly improving resource utilization efficiency. Thirdly, it introduces a weighted allocation strategy based on historical electricity consumption characteristics, enabling differentiated processing for different user groups. Finally, through an intelligent agent-integrated collaborative mechanism of forecasting and dynamic resource allocation, it improves computational efficiency while maintaining forecast accuracy. The entire solution forms a complete technical closed loop from data quality assessment to forecast execution, effectively solving the technical challenge of balancing forecast accuracy and operational efficiency in power marketing management. Attached Figure Description

[0010] Figure 1 A flowchart illustrating the integrated management method for power marketing based on an intelligent data platform provided by this invention; Figure 2This is a schematic diagram of the structure of the power marketing integration management platform based on an intelligent data middleware provided by the present invention.

[0011] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention; Figure 4 This is a schematic diagram of the structure of a computer-readable storage medium provided by the present invention.

[0012] In the attached diagram, the components represented by each number are as follows: The system includes a data acquisition and basic anomaly analysis module 11, a load change anomaly analysis module 12, a basic forecast resource allocation module 13, a load forecast and data recording management module 14, an electronic device 500, a memory 510, a processor 520, a first computer program 511, a computer-readable storage medium 600, and a second computer program 611. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0015] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0016] Example 1, as Figure 1As shown, this embodiment of the invention provides a power marketing integrated management method based on an intelligent data platform, including: S10: Through the intelligent data platform, collect power data of target entities within the power system, conduct basic power data anomaly analysis, and obtain the basic data anomaly degree; Specifically, through an intelligent data platform, power data of target entities within the power system is collected, and basic anomaly analysis of the power data is performed to obtain the anomaly degree of the basic data, including: Power data of target entities within the power system is collected through an intelligent data platform; Acquire historical power data for the target entity over multiple time periods. The deviation of the power data from multiple historical power data is calculated, and the mean is calculated to obtain the anomaly degree of the basic data.

[0017] First, power data from target entities within the power system is collected through an intelligent data platform. This intelligent data platform is a unified data processing platform integrating data collection, consolidation, governance, and service capabilities. As the core data hub, it is responsible for selectively acquiring real-time or near-real-time power data from metering devices, electricity consumption information collection systems, or relevant databases within the power system for designated target entities. Target entities within the power system include a single designated electricity user, a group of users, or a specific power supply area. This collection step is the foundational step in data acquisition, and the collected power data includes active power, reactive power, voltage, current, or electricity consumption values.

[0018] Secondly, historical electricity data for the target entity over multiple time periods is acquired to provide a reliable historical benchmark for subsequent anomaly analysis. Specifically, similar electricity data for the target entity over multiple consecutive time periods needs to be extracted from the historical data storage unit of the intelligent data platform. The time periods can be set according to the business scenario, such as hours, days, or months. The selection of multiple time periods should be statistically significant, reflecting the target entity's normal electricity consumption levels and fluctuation ranges.

[0019] Finally, the deviation between the current power data and multiple historical power data sets is calculated, and the arithmetic mean of all deviations is taken to obtain the basic data anomaly rate. The deviation can be calculated as the absolute value or a relative percentage of the difference between the current data and each historical data point for the same period. Then, all calculated individual deviations are summed and divided by the total number of historical data sets to obtain a comprehensive mean, i.e., the basic data anomaly rate. The value of this basic data anomaly rate directly characterizes the degree of deviation of the current power data from its historical normal level, thus providing a quantitative basis for assessing the reliability of the data itself.

[0020] For example, if the currently collected daily electricity consumption is 150 kWh, and the selected historical data is the daily electricity consumption sequence of the past seven days, with values ​​of 142 kWh, 138 kWh, 145 kWh, 155 kWh, 148 kWh, 152 kWh, and 140 kWh respectively, the relative percentage deviation between the current value and each historical value is first calculated using the formula: |current value - historical value| / historical value × 100%. The resulting seven relative percentage deviations are 5.63%, 8.70%, 3.45%, 3.23%, 1.35%, 1.32%, and 7.14%. Summing these seven relative percentage deviations yields 31.82%, which, divided by the number of historical data points (7), gives approximately 4.55%. This arithmetic mean of 4.55% is defined as the baseline data anomaly of the current daily electricity consumption data, characterizing the degree of deviation of the current data from historical normal levels.

[0021] S20: Based on the historical predicted load in the previous time period, perform load change anomaly analysis on the power data to obtain the anomaly degree of the change data; Specifically, based on the historical predicted load for the previous time period, load change anomaly analysis is performed on the power data to obtain the anomaly degree of the change data, including: Obtain historical forecast load for the previous time period, where historical forecast load includes power data for the next forecast time period; The degree of difference between the power data and the historical predicted load is calculated to obtain the degree of data anomaly.

[0022] First, the historical predicted load generated in the previous time period is obtained. This historical predicted load is an estimate of the future power data for the current time period, based on the available data and information at a specific point in the past, using a specific forecasting method. It represents an expectation of the current power situation based on historical knowledge. Specifically, this historical predicted load data can be obtained from the prediction result repository built into the intelligent data platform. This prediction result repository continuously records and maintains all prediction results generated by the load forecasting module in each time period. In the specific implementation, after the load forecast for a time period is completed, the corresponding prediction result, along with its timestamp and other metadata, is persistently stored in the prediction result repository. When performing anomaly analysis for the current period, the corresponding historical predicted load value can be retrieved from the prediction result repository based on the target entity identifier and the specified parameters of the previous time period.

[0023] Secondly, the difference between the currently collected power data and the historical predicted load is calculated to obtain the anomaly degree of the change data. This calculation process quantifies the degree of deviation between the actual measured value and the historical expected value using mathematical methods. In specific implementation, either the absolute difference method or the relative percentage difference method can be used for numerical calculation. The absolute difference method directly calculates the absolute value of the difference between the two values, while the relative percentage difference method multiplies the ratio of the absolute difference to the historical predicted load by 100%. The calculated difference degree is defined as the anomaly degree of the change data, which objectively characterizes the degree of consistency between the actual change in power load and the historical predicted trajectory. When the anomaly degree of the change data is at a high level, it indicates that the actual development trend of power load deviates from the expected pattern established based on historical knowledge. This state may originate from abnormal changes in electricity consumption behavior or unforeseen emergencies.

[0024] For example, suppose that at the end of the previous time period, the historical forecast load for the current period was 200 kW. The actual power data collected now is 235 kW. Using the relative percentage difference method, the data anomaly rate is calculated as |235 - 200| / 200 × 100% = 17.5%. This 17.5% data anomaly rate clearly quantifies the extent to which the actual load exceeds the expected load.

[0025] S30: Based on the anomaly degree of the basic data and the anomaly degree of the change data, configure the basic load forecasting resources for load forecasting based on the power data; Specifically, based on the anomaly degree of the basic data and the anomaly degree of the change data, basic load forecasting resources for load forecasting based on the power data are configured, including: The anomaly score of the fused data is calculated based on the anomaly scores of the basic data and the anomaly scores of the changed data. The anomaly degree of the fused data is used as the basic load forecasting coefficient for load forecasting based on the power data, and is used as the basic load forecasting resource.

[0026] First, the calculated anomalies of the baseline data and the variable data are merged to obtain the merged data anomaly. This fusion calculation uses a weighted summation method, assigning different weight coefficients to the two anomalies to reflect their relative importance. The weight coefficient of the baseline data anomaly reflects the reliability of the data itself, while the weight coefficient of the variable data anomaly reflects the stability of the load change trend. Specifically, the weight coefficients are set according to the electricity consumption characteristics of the target subject and the business analysis needs. For example, for residential users with stable electricity consumption behavior, the variable data anomaly is given a higher weight of 0.7; for industrial users, both baseline data anomalies and variable data anomalies are considered, with a weight of 0.5 for each.

[0027] The anomaly score of the fused data is equal to the product of the anomaly score of the basic data and its weighting coefficient, plus the product of the anomaly score of the variable data and its weighting coefficient. This weighted calculation of the fused data anomaly score comprehensively characterizes the anomalies in power data across both static quality and dynamic trend dimensions.

[0028] For example, based on the aforementioned calculated basic data anomaly rate of 4.55% and variable data anomaly rate of 17.5%, and using a balanced weight allocation scheme set for industrial user scenarios (i.e., a weighting coefficient of 0.5 for both basic and variable data anomalies), the fused data anomaly rate is calculated using a weighted summation formula.

[0029] The anomaly rate of the merged data is calculated as 4.55% × 0.5 + 17.5% × 0.5 = 11.025%. This calculation result comprehensively reflects the combined impact of data quality anomalies and load change anomalies.

[0030] Furthermore, the anomaly degree of the fused data is used as the basic load forecasting coefficient, and then as the basic load forecasting resource.

[0031] For example, when the calculated anomaly rate of the fused data is 11.025%, the basic load forecasting coefficient is set to 0.11025, meaning the basic load forecasting resource is 0.11025. This basic load forecasting resource directly participates in determining the resource allocation strategy for subsequent load forecasting tasks. A higher basic load forecasting resource value indicates a higher level of data anomaly, correspondingly requiring more computational resources to be allocated during the forecasting process for more in-depth data cleaning and model validation, or the invocation of more complex forecasting algorithms to address data uncertainty. Therefore, the basic load forecasting resource, as a key control variable, enables the technical objective of dynamically allocating forecasting resources based on data anomaly conditions.

[0032] S40: Based on the historical power data of the target entity, load forecasting weight and data recording weight are allocated, the basic load forecasting resources are adjusted, load forecasting resources are obtained, the forecast load is predicted, and data recording management is performed.

[0033] First, based on the historical power data of the target entity, load forecasting weights and data recording weights are allocated, and the basic load forecasting resources are adjusted to obtain load forecasting resources, including: Obtain the historical power data sequence of the target entity and calculate the power load fluctuation. Calculate the mean of the historical power data sequence to obtain average power data; Based on the power load fluctuation and average power data, load forecast weights and data recording weights are allocated to obtain load forecast weights and data recording weights. Calculate the ratio of load forecast weight to data record weight, adjust the calculated basic load forecast resources, and obtain the load forecast resources.

[0034] First, obtain the historical electricity data sequence of the target entity within a preset historical period. The preset historical period refers to the time span on which the weighting calculation is based. It is set according to the analysis needs of the specific business scenario and the electricity consumption characteristics of the target entity. For example, for daily load forecasting, it can be set to the most recent 30 days, and for weekly load forecasting, it can be set to the most recent 12 weeks. The historical electricity data sequence is obtained through the historical database query interface of the intelligent data platform. This database stores electricity consumption data at various time granularities that have been cleaned and standardized.

[0035] Furthermore, the fluctuation of electricity load is calculated, which can be characterized by statistical measures such as standard deviation or coefficient of variation. Standard deviation is calculated by taking the square root of the sum of the squares of the differences between each data point in the historical electricity data series and the series mean, directly reflecting the absolute dispersion of the data. Coefficient of variation is calculated by dividing the standard deviation by the series mean, yielding a relative fluctuation index that eliminates the influence of dimensions. In practice, standard deviation is suitable for comparison among different entities with similar electricity consumption levels, while coefficient of variation is more appropriate for groups of entities with significantly different electricity consumption scales to fairly assess the strength of their load fluctuations. Simultaneously, the arithmetic mean of the historical electricity data series is calculated to obtain average electricity data reflecting the typical electricity consumption level of the entity.

[0036] Furthermore, based on the calculated power load fluctuation and average power data, the weights for load forecasting and data recording are allocated.

[0037] Specifically, based on the power load fluctuation and average power data, load forecasting weights and data recording weights are allocated to obtain the load forecasting weights and data recording weights, including: Obtain the maximum power load fluctuation and maximum power data for similar entities; Calculate the ratio of the power load fluctuation and average power data of the target entity to the maximum power load fluctuation and maximum power data to obtain the fluctuation coefficient and power coefficient; Based on the fluctuation coefficient and power coefficient, the load forecast weight and data recording weight are calculated and assigned.

[0038] First, benchmark data needs to be obtained, namely the maximum power load fluctuation and maximum power consumption data of similar entities within the target entity's category. Similar entities refer to user groups with similar electricity consumption characteristics or belonging to the same industry classification. The maximum power load fluctuation represents the upper limit of load fluctuation within this group, while the maximum power consumption data characterizes the group's maximum electricity consumption scale. Specifically, the maximum power load fluctuation and maximum power consumption data of similar entities are obtained by traversing the historical data of all members within the group and taking the maximum value of the corresponding statistics.

[0039] Secondly, the fluctuation coefficient and power coefficient of the target entity are calculated. Specifically, the fluctuation coefficient is obtained by dividing the power load fluctuation of the target entity by the maximum power load fluctuation of similar entities. This fluctuation coefficient ranges from 0 to 1; the lower the value, the more stable the load change of the target entity. Simultaneously, the power coefficient is obtained by dividing the average power data of the target entity by the maximum power data of similar entities. This power coefficient also ranges from 0 to 1; the higher the value, the more significant the electricity consumption of the target entity is within the group.

[0040] Finally, weights are calculated based on the fluctuation coefficient and the power coefficient. Specifically, the fluctuation coefficient and the power coefficient are added together to obtain a total coefficient. Then, the ratio of the power coefficient to the total coefficient is calculated to obtain the load forecast weight; the ratio of the fluctuation coefficient to the total coefficient is calculated to obtain the data record weight.

[0041] For example, assuming an industrial user has a fluctuation coefficient of 0.3 and an electricity coefficient of 0.7, the sum of the coefficients is 1.0. The load forecasting weight equals the electricity coefficient divided by the sum of the coefficients, i.e., 0.7 / 1.0 = 0.7. The data recording weight equals the fluctuation coefficient divided by the sum of the coefficients, i.e., 0.3 / 1.0 = 0.3. This allocation mechanism reflects the following technical characteristics: the load forecasting weight is mainly related to the electricity coefficient, because entities with large electricity consumption have a more significant impact on the system load, and their forecasting accuracy is more important; the data recording weight is mainly related to the fluctuation coefficient, because entities with large load fluctuations have more abnormal patterns and change characteristics in their operating data, which have higher value for recording and analysis. By normalizing the two coefficients into weight allocation, resource balance between the two management objectives of load forecasting and data recording is ensured.

[0042] Specifically, load forecasting weights measure the proportion of current subject data in the load forecasting task, while data recording weights characterize the importance of that subject data in historical record management. Further, after obtaining the specific load forecasting weights and data recording weights, the ratio between them is calculated. This ratio is used as an adjustment coefficient and multiplied by the previously calculated basic load forecasting resources. Through this adjustment calculation, load forecasting resources tailored to the electricity consumption characteristics of a specific target subject are finally obtained. Load forecasting resources equal the basic load forecasting resources multiplied by the ratio of load forecasting weights to data recording weights.

[0043] For example, assuming the load forecast weight is 0.7, the data record weight is 0.3, and the basic load forecast resource is 0.11025, then the load forecast resource = basic load forecast resource × (load forecast weight ÷ data record weight) = 0.11025 × (0.7 ÷ 0.3) ≈ 0.257.

[0044] In summary, the adjusted load forecasting resources take into account both the abnormal conditions of the data itself and the electricity consumption characteristics of the main users, providing optimized resource guarantees for the subsequent generation of high-precision load forecasts.

[0045] Furthermore, the forecast load is obtained, and data recording and management are performed, including: Obtain a sequence of load forecasting agents, wherein the sequence of load forecasting agents includes multiple load forecasting agents, each of which is trained using a different set of sample power data and a set of future sample predicted loads; Based on the load prediction resources and the number of load prediction agents within the load prediction agent sequence, the number of agents to be invoked is determined. The number of load prediction agents is randomly selected and invoked. The power data is input, and the prediction output is used to obtain multiple individual predicted loads. The average value is calculated to obtain the predicted load.

[0046] First, a pre-built sequence of load forecasting agents is obtained. This sequence contains multiple independent load forecasting agents, each trained using machine learning methods and trained with different sets of sample electricity data and corresponding sets of sample predicted loads, thus ensuring differentiated forecasting characteristics among the agents.

[0047] For example, the load forecasting agent can employ a deep neural network architecture, consisting of a time-series feature extraction layer, a context fusion layer, and a load forecasting output layer. The time-series feature extraction layer receives a standardized electricity data sequence, which includes multi-dimensional time-series features such as historical load values, temperature data, and date type. This layer uses a one-dimensional convolutional neural network combined with a long short-term memory network structure, with a kernel width of 3. The number of hidden layer units is dynamically adjusted according to the length of the input sequence, and PReLU is selected as the activation function to enhance nonlinear representation capabilities. The context fusion layer employs a multi-head self-attention mechanism with 8 heads, capturing the dependencies between different time steps through weighted aggregation. The output layer uses a fully connected structure with a linear activation function to map the fused features to load forecast values ​​for a specific future time period.

[0048] Key hyperparameters during training include: an initial learning rate of 0.0005, a maximum of 200 training epochs, a batch size of 32, and a gradient clipping threshold of 1.0. The learning rate employs a cosine annealing scheduling strategy, decaying to 0.8 times its original value after every 50 training epochs. The number of training epochs ensures the model fully learns the temporal dynamic characteristics of power load, while the batch size balances training stability and computational efficiency. Furthermore, supervised learning is employed, collecting sample power data sequences from a historical power database as the input sample set, and simultaneously acquiring the corresponding actual load values ​​to form a sample label set. The input sample set and the corresponding label sample set are divided into training, validation, and test sets in an 8:1:1 ratio. Before training, the input data is standardized to eliminate the influence of unit dimensions.

[0049] The sample sequences in the training set are used as input, with the corresponding actual load values ​​as supervision signals. The network parameters are iteratively optimized using the backpropagation algorithm combined with the AdamW optimizer. The smoothed average absolute percentage error is used as the loss function. During training, the model performance is monitored using a validation set. When the validation set loss function value no longer decreases for 15 consecutive rounds and the prediction error is lower than a preset threshold of 3.5%, training is terminated, the optimal model parameters are saved, and the converged load prediction agent is obtained.

[0050] Each agent learns differentiated load change patterns from different subsets of data, providing a diverse basis for subsequent ensemble predictions.

[0051] Furthermore, based on the load forecasting resources calculated in the aforementioned steps, and combined with the total number of available agents in the load forecasting agent sequence, the specific number of agents to be invoked is determined through a preset mapping rule. A higher load forecasting resource value indicates more abundant resources allocated to the forecasting task, and consequently, a greater number of agents can be invoked. For example, the corrected load forecast can be multiplied by the number of load forecasting agents, and the result rounded up; the upper limit for the number of agents invoked is the total number of load forecasting agents.

[0052] For example, assuming the load forecasting resource is 0.257 and the number of load forecasting agents is 20, then the number of agents called = 20 × 0.257 = 5.14, rounded up to 6. Therefore, the final number of agents called is determined to be 6.

[0053] Furthermore, after determining the number of agents to be invoked, a specified number of load forecasting agents are randomly selected from the sequence of load forecasting agents. The currently collected power data is input in parallel into the selected load forecasting agents, and each agent independently performs forecast calculations and outputs its individual predicted load value. The individual predicted loads output by all invoked load forecasting agents are collected, and their arithmetic mean is calculated. This average value is used as the final predicted load result.

[0054] Furthermore, after obtaining the predicted load, key data related to this prediction are written to a historical database for persistent storage, completing data record management. The recorded data includes input power data, the identifier of the invoked agent, the predicted load of each agent, the final predicted load value, and key parameters in the calculation process, providing complete data support for subsequent prediction accuracy analysis and model optimization.

[0055] In summary, the embodiments of this application have at least the following technical effects: Compared to existing technologies, this application firstly achieves unified collection and multi-dimensional anomaly analysis of power data through an intelligent data platform. This platform assesses both the fundamental anomaly of the data itself and the anomaly of load variation data, establishing a more comprehensive data quality assessment system. Secondly, it dynamically allocates basic load forecasting resources based on the results of dual anomaly analysis, ensuring precise matching between forecasting resource allocation and data quality status, avoiding resource waste and accuracy degradation caused by blindly forecasting in the event of data anomalies. Thirdly, it introduces a weighting allocation mechanism based on historical electricity consumption characteristics, automatically calculating load forecasting weights and data record weights according to load fluctuations and average power data, enabling differentiated resource allocation among different entities and improving resource utilization efficiency. Finally, by constructing a sequence of intelligent load forecasting agents and dynamically determining the number of calls based on load forecasting resource coefficients, an integrated forecasting strategy effectively improves the accuracy and robustness of load forecasting. Ultimately, this forms a complete closed loop from data quality assessment and dynamic resource allocation to intelligent forecasting execution, improving the refinement level and operational efficiency of power marketing management.

[0056] Example 2, as Figure 2 As shown, based on the same inventive concept as the power marketing integration management method based on an intelligent data platform provided in Embodiment 1, this embodiment of the invention also provides a power marketing integration management platform based on an intelligent data platform, including: The data acquisition and basic anomaly analysis module 11 is used to collect power data of target entities within the power system through the intelligent data platform, and to perform basic anomaly analysis of the power data to obtain the anomaly degree of the basic data. The load change anomaly analysis module 12 is used to perform load change anomaly analysis on the power data based on the historical predicted load in the previous time period to obtain the anomaly degree of the change data. The basic forecasting resource configuration module 13 is used to configure the basic load forecasting resources for load forecasting based on the power data according to the basic data anomaly degree and the change data anomaly degree. The load forecasting and data recording management module 14 is used to allocate load forecasting weights and data recording weights based on the historical power data of the target entity, adjust the basic load forecasting resources, obtain load forecasting resources, forecast the load, and perform data recording management.

[0057] Specifically, the data acquisition and basic anomaly analysis module 11 is used for: Through an intelligent data platform, power data of target entities within the power system is collected, and basic anomaly analysis of the power data is performed to obtain the anomaly degree of the basic data, including: Power data of target entities within the power system is collected through an intelligent data platform; Acquire historical power data for the target entity over multiple time periods. The deviation of the power data from multiple historical power data is calculated, and the mean is calculated to obtain the anomaly degree of the basic data.

[0058] Specifically, the load change anomaly analysis module 12 is used for: Based on historical predicted loads from the previous time period, load change anomaly analysis is performed on the power data to obtain the anomaly degree of the change data, including: Obtain historical forecast load for the previous time period, where historical forecast load includes power data for the next forecast time period; The degree of difference between the power data and the historical predicted load is calculated to obtain the degree of data anomaly.

[0059] Specifically, the basic prediction resource allocation module 13 is used for: Based on the anomaly degree of the basic data and the anomaly degree of the change data, configure basic load forecasting resources for load forecasting based on the power data, including: The anomaly score of the fused data is calculated based on the anomaly scores of the basic data and the anomaly scores of the changed data. The anomaly degree of the fused data is used as the basic load forecasting coefficient for load forecasting based on the power data, and is used as the basic load forecasting resource.

[0060] The load forecasting and data recording management module 14 is specifically used for: Based on the historical power data of the target entity, load forecasting weights and data recording weights are allocated, and the basic load forecasting resources are adjusted to obtain load forecasting resources, including: Obtain the historical power data sequence of the target entity and calculate the power load fluctuation. Calculate the mean of the historical power data sequence to obtain average power data; Based on the power load fluctuation and average power data, load forecast weights and data recording weights are allocated to obtain load forecast weights and data recording weights. Calculate the ratio of load forecast weight to data record weight, adjust the calculated basic load forecast resources, and obtain the load forecast resources.

[0061] Specifically, based on the power load fluctuation and average power data, load forecasting weights and data recording weights are allocated to obtain the load forecasting weights and data recording weights, including: Obtain the maximum power load fluctuation and maximum power data for similar entities; Calculate the ratio of the power load fluctuation and average power data of the target entity to the maximum power load fluctuation and maximum power data to obtain the fluctuation coefficient and power coefficient; Based on the fluctuation coefficient and power coefficient, the load forecast weight and data recording weight are calculated and assigned.

[0062] Furthermore, the forecast load is obtained, and data recording and management are performed, including: Obtain a sequence of load forecasting agents, wherein the sequence of load forecasting agents includes multiple load forecasting agents, each of which is trained using a different set of sample power data and a set of future sample predicted loads; Based on the load prediction resources and the number of load prediction agents within the load prediction agent sequence, the number of agents to be invoked is determined. The number of load prediction agents is randomly selected and invoked. The power data is input, and the prediction output is used to obtain multiple individual predicted loads. The average value is calculated to obtain the predicted load.

[0063] Example 3, Figure 3 This is a schematic diagram illustrating an embodiment of the electronic device provided in this invention. For example... Figure 3As shown, this embodiment of the invention provides an electronic device 500, including a memory 510, a processor 520, and a first computer program 511 stored in the memory 510 and executable on the processor 520. When the processor 520 executes the first computer program 511, it implements the power marketing integration management method based on an intelligent data platform as described in Embodiment 1.

[0064] Example 4, Figure 4 This is a schematic diagram illustrating an embodiment of a computer-readable storage medium provided by an embodiment of the present invention. For example... Figure 4 As shown, this embodiment of the invention provides a computer-readable storage medium 600, on which a second computer program 611 is stored. When the second computer program 611 is executed by a processor, it implements the power marketing integration management method based on an intelligent data platform as described in Embodiment 1.

[0065] It should be noted that the order of the embodiments described above 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 implementations, multitasking and parallel processing are possible or may be advantageous.

[0066] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0067] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A power marketing integrated management method based on an intelligent data platform, characterized in that: The method includes: Through the intelligent data platform, power data of target entities within the power system are collected, basic power data anomaly analysis is performed, and the anomaly degree of basic data is obtained. Based on the historical predicted load in the previous time period, load change anomaly analysis is performed on the power data to obtain the anomaly degree of the change data. Based on the anomaly degree of the basic data and the anomaly degree of the change data, configure the basic load forecasting resources for load forecasting based on the power data; Based on the historical power data of the target entity, load forecasting weights and data recording weights are allocated, the basic load forecasting resources are adjusted, load forecasting resources are obtained, the forecast load is predicted, and data recording management is performed.

2. The power marketing integration management method based on an intelligent data platform according to claim 1, characterized in that, Through an intelligent data platform, power data of target entities within the power system is collected, and basic anomaly analysis of the power data is performed to obtain the anomaly degree of the basic data, including: Power data of target entities within the power system is collected through an intelligent data platform; Acquire historical power data for the target entity over multiple time periods. The deviation of the power data from multiple historical power data is calculated, and the mean is calculated to obtain the anomaly degree of the basic data.

3. The power marketing integrated management method based on an intelligent data platform according to claim 1, characterized in that, Based on historical predicted loads from the previous time period, load change anomaly analysis is performed on the power data to obtain the anomaly degree of the change data, including: Obtain historical forecast load for the previous time period, where historical forecast load includes power data for the next forecast time period; The degree of difference between the power data and the historical predicted load is calculated to obtain the degree of data anomaly.

4. The power marketing integration management method based on an intelligent data platform according to claim 1, characterized in that, Based on the anomaly degree of the basic data and the anomaly degree of the change data, configure basic load forecasting resources for load forecasting based on the power data, including: The anomaly score of the fused data is calculated based on the anomaly scores of the basic data and the anomaly scores of the changed data. The anomaly degree of the fused data is used as the basic load forecasting coefficient for load forecasting based on the power data, and is used as the basic load forecasting resource.

5. The power marketing integration management method based on an intelligent data platform according to claim 1, characterized in that, Based on the historical power data of the target entity, load forecasting weights and data recording weights are allocated, and the basic load forecasting resources are adjusted to obtain load forecasting resources, including: Obtain the historical power data sequence of the target entity and calculate the power load fluctuation. Calculate the mean of the historical power data sequence to obtain average power data; Based on the power load fluctuation and average power data, load forecast weights and data recording weights are allocated to obtain load forecast weights and data recording weights. Calculate the ratio of load forecast weight to data record weight, adjust the calculated basic load forecast resources, and obtain the load forecast resources.

6. The power marketing integrated management method based on an intelligent data platform according to claim 5, characterized in that, Based on the aforementioned power load fluctuation and average power data, load forecasting weights and data recording weights are allocated to obtain the load forecasting weights and data recording weights, including: Obtain the maximum power load fluctuation and maximum power data for similar entities; Calculate the ratio of the power load fluctuation and average power data of the target entity to the maximum power load fluctuation and maximum power data to obtain the fluctuation coefficient and power coefficient; Based on the fluctuation coefficient and power coefficient, the load forecast weight and data recording weight are calculated and assigned.

7. The power marketing integration management method based on an intelligent data platform according to claim 1, characterized in that, Forecasting loads and managing data records include: Obtain a sequence of load forecasting agents, wherein the sequence of load forecasting agents includes multiple load forecasting agents, each of which is trained using a different set of sample power data and a set of future sample predicted loads; Based on the load prediction resources and the number of load prediction agents within the load prediction agent sequence, the number of agents to be invoked is determined. The number of load prediction agents is randomly selected and invoked. The power data is input, and the prediction output is used to obtain multiple individual predicted loads. The average value is calculated to obtain the predicted load.

8. A power marketing integrated management platform based on an intelligent data middleware, characterized in that: The method for implementing the integrated power marketing management based on an intelligent data platform as described in any one of claims 1-7 includes: The data acquisition and basic anomaly analysis module is used to collect power data of target entities within the power system through an intelligent data platform, and to perform basic anomaly analysis of the power data to obtain the anomaly degree of the basic data. The load change anomaly analysis module is used to perform load change anomaly analysis on the power data based on the historical predicted load in the previous time period to obtain the anomaly degree of the change data. The basic forecasting resource configuration module is used to configure the basic load forecasting resources for load forecasting based on the power data, according to the basic data anomaly degree and the change data anomaly degree. The load forecasting and data recording management module is used to allocate load forecasting weights and data recording weights based on the historical power data of the target entity, adjust the basic load forecasting resources, obtain load forecasting resources, forecast the load, and manage data recording.

9. An electronic device, characterized in that, include: Memory, used to store computer software programs; A processor is used to read and execute the computer software program, thereby implementing the power marketing integration management method based on an intelligent data platform as described in claims 1-7.

10. A non-transitory computer-readable storage medium, characterized in that, The storage medium stores a computer software program, which, when executed by a processor, implements the power marketing integration management method based on an intelligent data platform as described in claims 1-7.

Images