Data center-based integrated electricity purchase and sale intelligent analysis and decision system and method

Abstract

The application provides a kind of based on data in the station's integrated electricity purchase and sale intelligent analysis and decision system and method, it is related to data analysis technical field.The application includes: data in the station layer, for the data from multiple business systems is aggregated, and data is cleaned, integrated processing, form unified data resource;Business service layer is based on microservice architecture construction, contains multiple business function microservices;Multiple business function microservices call unified data resource;Service access layer is used for user request, and carries out routing, authentication and flow control;Front-end application layer is used to provide visual interactive interface to user, at least including business board, data analysis report and prediction report generation.The application provides integrated electricity purchase and sale intelligent analysis and decision method based on the above system implementation.The application realizes the real-time monitoring, accurate prediction and intelligent decision of electricity purchase and sale business.

Description

Technical Field

[0001] This invention belongs to the field of data analysis technology, and specifically relates to an intelligent analysis and decision-making system and method for integrated power purchase and sales based on a data middle platform. Background Technology

[0002] With the deepening of power market reforms, power grid companies are shifting their business focus from traditional transmission, distribution, and sales to integrated power purchase and sales operations. Against this backdrop, companies are placing higher demands on the lean management of their power purchase and sales businesses. Currently, mainstream business analysis models have the following limitations:

[0003] Data on electricity purchase, sales, load, and costs are scattered across multiple independent systems such as marketing, data collection, and transactions. The data standards are inconsistent, making it difficult to conduct correlation analysis and gain a holistic understanding.

[0004] Existing analyses are mostly based on fixed reports and post-event statistics, lacking the ability to make forward-looking predictions and real-time perceptions of market trends and operational risks, and thus cannot effectively support rapid decision-making such as purchasing and bidding.

[0005] Traditional statistical analysis tools struggle to handle the complex nonlinear relationships and time-series characteristics in power data, resulting in insufficient accuracy and reliability of prediction models, and they are not deeply integrated with specific power purchase and sale business rules (such as cost transmission and policy impact).

[0006] Traditional monolithic application architectures struggle to support rapidly changing business needs and the processing of massive amounts of data, while microservices lack sufficient collaboration, fault tolerance, and elastic scaling capabilities.

[0007] Therefore, there is an urgent need for an integrated power purchase and sales analysis and decision support system that can integrate multi-source data, possess intelligent predictive capabilities, and adapt to rapid business changes. Summary of the Invention

[0008] Therefore, the technical problem to be solved by the present invention is to provide an intelligent analysis and decision-making system and method for integrated power purchase and sale based on a data platform, which can realize real-time monitoring, accurate prediction and intelligent decision-making of power purchase and sale business.

[0009] In a first aspect, this invention discloses an intelligent analysis and decision-making system for integrated power purchase and sales based on a data middle platform, the system comprising:

[0010] The data platform layer is used to aggregate data from multiple business systems, clean and integrate the data to form a unified data resource.

[0011] The business service layer is built on a microservice architecture and contains multiple business function microservices; these multiple business function microservices call unified data resources.

[0012] The service access layer is used for user requests and performs routing, authorization authentication, and rate limiting.

[0013] The front-end application layer is used to provide users with a visual interactive interface, including at least business dashboards, data analysis reports, and forecast reports.

[0014] Furthermore, the data platform layer includes:

[0015] Based on pre-configured CDC change data capture rules, incremental data is extracted in real time from the multiple business systems to form a first dataset; the first dataset includes electricity data and electricity price data;

[0016] In the source layer, the first dataset is cleaned, and the cleaning includes: applying an outlier detection and smoothing algorithm based on box plot statistics to the electricity data, and applying term mapping and normalization based on a business rule dictionary to the electricity price data.

[0017] In the shared layer, dimensional modeling is used to build a standardized model with consistent dimensions;

[0018] At the analysis layer, based on the standardized model, data integration is performed through the Spark distributed computing engine to generate the unified data resource.

[0019] Furthermore, the business dashboard includes:

[0020] The configurable graphical orchestrator allows users to customize dashboard layouts and indicator components through drag-and-drop functionality.

[0021] Real-time indicator data is monitored based on preset rules, and a visual alarm is automatically triggered when any indicator exceeds a preset threshold.

[0022] In response to a user's click on any summary data point in the dashboard, the user is redirected to the corresponding business function microservice.

[0023] Furthermore, the business function microservices include an operational assessment and prediction microservice, which includes:

[0024] The rules configuration module is used to set the boundary conditions for electricity sales and transaction volume;

[0025] The prediction generation module, based on historical electricity purchase and sales data, automatically generates prediction results for electricity sales volume, average electricity price, and electricity cost composition for a preset future period through a preset prediction model.

[0026] The prediction review module is used to compare the prediction results with the actual data, calculate the prediction accuracy, and provide the reasons for the deviation.

[0027] Furthermore, the electricity sales forecast in the forecast generation module adopts the following model:

[0028]

[0029] in, t represents the predicted value at time point t; B is the shift factor; p is the non-seasonal autoregressive order; d is the non-seasonal differencing order; q is the non-seasonal moving average order; P is the seasonal autoregressive order; D is the seasonal differencing order; Q is the seasonal moving average order; s is the seasonal cycle length. θq(B) is a non-seasonal AR term polynomial; θq(B) is a non-seasonal MA term polynomial. For seasonal AR term polynomials; It is a seasonal MA-term polynomial; This is the random error term; This is a policy shock intensity index quantified based on policy documents. Meteorological influencing factors calculated based on meteorological data. , They are respectively , The corresponding regression coefficients.

[0030] Furthermore, the average electricity price prediction in the prediction generation module adopts a machine learning regression model, and the training process is performed using the following loss function:

[0031]

[0032] in, n The number of samples; y i For the first i The historical average electricity price of a sample; According to the first i Features of a sample xi The predicted average electricity price;

[0033] This is a rigid cost floor calculated based on electricity purchase costs, transmission and distribution prices, and government funds. This is the constraint strength hyperparameter.

[0034] Furthermore, the prediction generation module uses different models to perform combined predictions, with each model having a weight. Calculated dynamically using the following formula:

[0035]

[0036] in, represents the importance coefficient of historical performance, and is a hyperparameter between 0 and 1; Accuracy (M) i ) For the model M i Historical accuracy; Accuracy (M) j ) For the model M j Historical accuracy; N The total number of models participating in the combination; Based on real-time business scenarios as the model The scene adaptability score is calculated.

[0037] Furthermore, the system architecture is deployed on a cloud platform, the business service layer adopts containerized orchestration, and the microservices use Nacos as a service registration and discovery component for service registration and subscription.

[0038] Microservices use RESTful calls and integrate the Sentinel component to achieve dependency isolation and circuit breaking / degradation. When a microservice call has a long response time or a high failure rate, the system automatically degrades to returning pre-prepared cached data.

[0039] Furthermore, the containerized orchestration utilizes Kubernetes' HorizontalPodAutoscaler to automatically scale the number of business service layer Pod instance replicas based on the query per second (QPS) or the stack length of the data platform layer task queue.

[0040] The service registration and discovery component works in conjunction with Kubernetes' Service mechanism to achieve weight-based traffic scheduling and namespace-based multi-tenant service isolation.

[0041] In a second aspect, this invention discloses an intelligent analysis and decision-making method for integrated power purchase and sales based on a data middle platform, the method comprising:

[0042] Aggregate data from multiple business systems, clean and integrate the data to form a unified data resource;

[0043] Multiple business function microservices are built based on a microservice architecture; the multiple business function microservices call a unified data resource, including at least: electricity purchase and sale management microservice, electricity purchase and sale cost monitoring microservice, average electricity purchase and sale price analysis microservice, and business investigation and forecasting microservice.

[0044] Accept user requests and perform routing, authorization, and rate limiting;

[0045] Provide users with a visual interactive interface, including at least business dashboards, data analysis reports, and generated forecast reports;

[0046] The operational survey and prediction microservice adopts any of the prediction models described above.

[0047] Beneficial effects:

[0048] This invention achieves deep integration and standardization of multi-source heterogeneous data through a data platform, breaking down data silos.

[0049] By deeply coupling the prediction model with the specific scenarios, rules, and constraints of the electricity purchase and sale business, a domain-specific high-precision and high-reliability prediction model has been formed.

[0050] This invention not only provides historical data analysis, but also provides forward-looking data support for core decisions such as esports bidding and business investigation through intelligent prediction and simulation.

[0051] This invention is based on cloud-native and microservice architecture, and has good elastic scaling, high availability and rapid iteration capabilities, which can adapt to the rapidly changing needs of the power market. Attached Figure Description

[0052] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0053] Figure 1 This is a diagram of the integrated intelligent analysis and decision-making system for purchasing and selling electricity according to Embodiment 1 of the present invention.

[0054] Figure 2 This is a data middleware layer architecture diagram of Embodiment 1 of the present invention. Detailed Implementation

[0055] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. The principles and features of the present invention are described below with reference to the accompanying drawings. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other. The embodiments given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0056] To continuously improve the efficiency of electricity fee management and enhance the management services and bidding decision-making capabilities for electricity purchase and sale, this invention constructs an integrated platform system for electricity purchase and sale based on a data middleware platform. It enables accurate and rapid statistical analysis of electricity purchase and sale-related businesses, builds a defense against operational risks, reduces the burden on grassroots staff from the perspective of data analysis and automated reporting, and provides auxiliary decision-making and digital support for building a new type of power system.

[0057] To improve the efficiency of electricity management, the ability to make purchasing decisions and bids, the ability to analyze operational benefits, and the digital transformation of marketing management, a data platform is used to aggregate data from marketing, development, and transaction business systems. Big data technology is used to achieve fast, stable, and accurate data statistics, processing, and analysis, providing support for economic decision-making analysis, feedback, and management.

[0058] Example 1

[0059] This embodiment provides an integrated intelligent analysis and decision-making system for electricity purchase and sales based on a data middle platform. As shown in Table 1, this system is based on a microservice architecture and is mainly divided into five layers: front-end, service access layer, business service layer, PaaS, IaaS, as well as security protection, operation and maintenance monitoring, application construction, etc., supporting the technical implementation of application architecture and data architecture.

[0060] The system architecture is deployed on a cloud platform, and the business service layer adopts containerized orchestration; the microservices use Nacos as a service registration and discovery component for service registration and subscription.

[0061] Microservices use RESTful calls and integrate the Sentinel component to achieve dependency isolation and circuit breaking / degradation. When a microservice call has a long response time or a high failure rate, the system automatically degrades to returning pre-prepared cached data.

[0062] The containerized orchestration tool is Kubernetes, which uses HorizontalPodAutoscaler to automatically scale the number of business service layer Pod instance replicas based on the query per second (QPS) or the stack length of the data middle platform layer task queue.

[0063] The service registration and discovery component works in conjunction with Kubernetes' Service mechanism to achieve weight-based traffic scheduling and namespace-based multi-tenant service isolation.

[0064] Table 1 System Architecture Layers and Components

[0065] Serial Number level describe Components / Tools 1 front end Responsible for interacting with users, and can use various front-end frameworks and tools, such as React and Vue.js, depending on the specific application requirements. React, Vue.js 2 Service Access Layer This is responsible for handling requests from the frontend, performing operations such as routing, rate limiting, and permission verification. Common API gateways include Spring Cloud Gateway and Zuul. SpringCloudGateway, Zuul 3 Business service layer It consists of a series of independent business function services, each of which is an independent Spring Boot application. Nacos is used for service registration and discovery, Ribbon for load balancing, and Hystrix for circuit breaker protection. SpringBoot, Nacos, Ribbon, Hystrix 4 PaaS layer Provide container orchestration services (such as Kubernetes) for automated deployment and management of microservices. Use monitoring and log management tools (such as Prometheus and ELKStack) to monitor the running status and performance of microservices. Kubernetes, Prometheus, ELKStack 5 IaaS layer To provide infrastructure resources such as computing, storage, and networking, you can use IaaS services provided by cloud service providers (such as Alibaba Cloud). These services can help quickly deploy and manage microservices, and ensure high availability and scalability. Alibaba Cloud IaaS service

[0066] The data platform integrates source layer, shared layer, and analysis layer data to generate cross-professional, multi-domain analytical results. For business interaction, a cloud-based database serves as the foundation, supporting data exchange with front-end information systems. The system architecture is as follows: Figure 1 As shown.

[0067] like Figure 2 As shown, the data platform layer includes:

[0068] Based on pre-configured CDC change data capture rules, incremental data is extracted in real time from the multiple business systems to form a first dataset; the first dataset includes electricity data and electricity price data; at the source layer, the first dataset is cleaned, and the cleaning includes: applying an outlier detection and smoothing algorithm based on box plot statistics to the electricity data, and applying term mapping and normalization based on the business rule dictionary to the electricity price data.

[0069] In the shared layer, dimensional modeling is used to build a standardized model with consistent dimensions;

[0070] At the analysis layer, based on the standardized model, data integration is performed through the Spark distributed computing engine to generate the unified data resource.

[0071] Specifically, the Canal component is used to capture CDC logs from the marketing system's database to obtain real-time electricity sales change data. Sqoop is used to periodically synchronize electricity purchase contract data from the electricity trading platform in batches on a T+1 basis.

[0072] At the source layer, the power data stream is processed in real time using Apache Flink, and a box plot algorithm is applied (that is, data points that exceed 1.5 times the interquartile range of the upper and lower quartiles are regarded as outliers and smoothed).

[0073] In the shared layer, a multidimensional data cube is built using Apache Kylin. The core fact tables include the "Daily Electricity Sales Fact Table" (dimensions: time, user, industry; measures: electricity volume, electricity cost) and the "Electricity Purchase Cost Fact Table" (dimensions: time, power plant, transaction type; measures: contracted electricity volume, settlement price).

[0074] In the analysis layer, a "wide table for matching electricity purchase and sale days" is generated through pre-computed SparkSQL tasks and directly provided to the business service layer for querying.

[0075] The front-end application layer is developed using the Vue.js framework. The business dashboard uses Echarts for graphical rendering. When the "Monthly Electricity Purchase and Sale Profit" indicator on the dashboard decreases by more than 10% month-on-month and triggers an alarm, the user clicks on the indicator, and the system automatically redirects to the details page of the electricity purchase and sale cost monitoring microservice, and automatically filters out the current month and the region with the largest month-on-month decrease for focused analysis.

[0076] The business dashboard uses a configurable graphical orchestrator, allowing users to customize the dashboard layout and indicator components through drag-and-drop.

[0077] Real-time indicator data is monitored based on preset rules, and a visual alarm is automatically triggered when any indicator exceeds a preset threshold.

[0078] In response to a user's click on any summary data point in the dashboard, the user is redirected to the corresponding business function microservice.

[0079] The business service layer uses the Spring Boot framework to develop various microservices. All microservice API interfaces are registered with the gateway in accordance with the OpenAPI 3.0 specification. Microservices make declarative RESTful calls through the Feign client and integrate Sentinel. A QPS threshold of 100 is set for the prediction microservice. When the number of calls exceeds this threshold, the system automatically breaks the circuit and returns the last successful prediction result to ensure that the UI does not freeze.

[0080] The business function microservices include the business assessment and prediction microservices, which include:

[0081] The rule configuration module is used to set the boundary conditions for electricity sales and trading. Based on the statistical time, the rule configuration module maintains the boundary conditions for electricity sales and trading, providing a basis for verification of subsequent forecast results.

[0082] The forecast generation module, based on historical electricity purchase and sales data, automatically generates forecast results for electricity sales volume, average electricity price, and electricity cost composition for a preset future period through a pre-set forecast model. It compares the forecast data with various types of electricity sales volume, average electricity price, basic electricity cost, government funds, power regulation fees, time-of-use fees, and tiered residential electricity rates, according to the power supply unit and review time. This analysis assesses the forecast accuracy, supports business personnel in analyzing the causes of forecast deviations, and improves the accuracy of forecasts for operational assessments.

[0083] The forecasting and review module compares the forecast results with actual data, calculates the forecast accuracy, and provides the reasons for deviations. Based on the power supply unit and review time, the module compares various types of electricity sales volume, average electricity price, basic electricity fee, government funds, power regulation fees, time-of-use fees, and tiered residential electricity fees with the forecast data, analyzes the forecast accuracy, supports business personnel in analyzing the reasons for forecast deviations, and improves the accuracy of forecasts regarding operational assessments.

[0084] The operational survey and forecasting microservice is implemented as follows:

[0085] The specialized SARIMA model is as follows:

[0086]

[0087] in, t represents the predicted value at time point t; B is the shift factor; p is the non-seasonal autoregressive order; d is the non-seasonal differencing order; q is the non-seasonal moving average order; P is the seasonal autoregressive order; D is the seasonal differencing order; Q is the seasonal moving average order; s is the seasonal cycle length. θq(B) is a non-seasonal AR term polynomial; θq(B) is a non-seasonal MA term polynomial. For seasonal AR term polynomials; It is a seasonal MA-term polynomial; This is the random error term; This is a policy shock intensity index quantified based on policy documents. Meteorological influencing factors calculated based on meteorological data. , represents the corresponding regression coefficient.

[0088] Seasonal and non-seasonal are two different periodic patterns that models use to capture in time series data.

[0089] Non-seasonal patterns (p, d, q):

[0090] It captures the dependencies between months. For example, sales in January this year may affect sales in February this year (the Spring Festival effect manifested in adjacent months).

[0091] What is eliminated is the long-term growth or decline trend of electricity sales (for example, due to economic development, the overall electricity consumption grows steadily every year).

[0092] Seasonal patterns (P, D, Q)s, where s=12):

[0093] It captures a fixed pattern year after year. For example:

[0094] Every year in July and August, there is a peak in electricity consumption due to air conditioning.

[0095] Every year in February, due to the Spring Festival holiday, industrial and commercial electricity consumption decreases, creating a low point in electricity demand.

[0096] When the model learns to predict electricity sales in August next year, data from August this year, August last year, and so on are crucial references.

[0097] A specialized SARIMA model is used for electricity sales forecasting. Taking monthly forecasting as an example, the period s=12. policy The quantification method for (t) is as follows: when the government issues a notice on orderly electricity use in summer, the load impact is comprehensively assessed based on the scope of enterprises involved and the voltage reduction ratio, and a value of -0.03 is assigned. weather (t) is the difference between the predicted "degree-days" within the forecast period and the historical average for the same period. The model parameters (p,d,q) are determined using grid search and the AIC criterion on 5 years of historical data.

[0098] The above model has two specific, business-quantified exogenous variables I policy(t) and H weather This upgrades the model from an "introspective" model that relies solely on historical data to an "insightful" model that can perceive and respond to changes in the external environment.

[0099] The model uses coefficients and It dynamically learns and quantifies the specific impact of policies and weather on electricity sales, thus enabling automatic and accurate adjustment of forecast values ​​when policies are released or extreme weather is predicted.

[0100] The specialized GBDT model employs a machine learning regression model, and the training process uses the following loss function:

[0101]

[0102] in, n The number of samples; y i For the first i The historical average electricity price of a sample; According to the first i Features of a sample xi The predicted average electricity price;

[0103] This is a rigid cost floor calculated based on electricity purchase costs, transmission and distribution prices, and government funds. This is the constraint strength hyperparameter.

[0104] For predicting average electricity sales prices, a specialized GBDT model (using the XGBoost library) is implemented. Feature vectors x i This includes: average electricity purchase price, comprehensive line loss rate, policy adjustment factor (0 indicates no policy, 1 indicates the introduction of a new energy electricity price subsidy policy), and the proportion of electricity consumption by large industrial users. total The formula for calculating (xi) is: average electricity purchase price / (1 - line loss rate) + average transmission and distribution price + 0.015 (where 0.015 is the estimated government fund and surcharges and necessary profit margin). The hyperparameter λ is set to 0.7 through cross-validation.

[0105] The prediction generation module uses different models to perform combined predictions, with each model having specific weights. Calculated dynamically using the following formula:

[0106]

[0107] in, represents the importance coefficient of historical performance, and is a hyperparameter between 0 and 1; Accuracy (M) i ) For the modelM i Historical accuracy; Accuracy (M) j ) For the model M j Historical accuracy; N The total number of models participating in the combination; Based on real-time business scenarios as the model The scene adaptability score is calculated.

[0108] Preferably, for combined prediction, it can be set that: when I policy When the absolute value of (t) is greater than 0.02, it is determined to be a policy-sensitive period, and the scenario fit score of the GBDT model is [not specified]. S gbdt (t) (correspond S i (t) (representing the GBDT model) is set to 0.8, and the S of the SARIMA model is... sarima (t) (corresponding to) S i (t) (This indicates the SARIMA model) is set to 0.2; otherwise, during the steady-state period, S... sarima (t) (set to 0.7, S_gbdt(t) set to 0.3. Harmonic parameter α is set to 0.6.)

[0109] The business function microservices of this invention also include a power purchase analysis module, a power sales analysis module, a power purchase and sales matching module, a power purchase and sales fee analysis module, and a payment status analysis module.

[0110] The electricity purchase analysis module, based on the selected statistical month and period (monthly or annual), provides visualized graphical displays of electricity purchase type distribution, new power plant additions, and electricity purchase trends through dimensions such as electricity purchase volume, year-on-year growth, percentage, and changing trends. Data tables display the year-on-year and month-on-month growth, percentage, percentage change, and contribution rate of electricity purchases by different types of power generation enterprises for the current month and cumulatively.

[0111] The electricity sales analysis module allows users to filter and query data based on power supply unit, statistical month, and statistical period (monthly or annual). It provides a visual display of electricity sales and year-on-year growth across multiple dimensions, including city / region, user type, direct transaction status, and monthly trends. Data tables are used to display monthly and annual electricity sales data by region, category, user type, and industry, showing current year, last year, year-on-year, month-on-month growth, percentage, percentage change, contribution rate, and driving force changes.

[0112] The electricity purchase and sales matching module allows filtering and querying based on statistical time and period (monthly or annually). It provides historical electricity consumption monitoring and analysis, including purchase and sales trend analysis and detailed purchase and sales data, along with a visual display of trend changes and percentage changes. Data tables show the monthly and cumulative purchase volume and percentage for each electricity category. Single-dimensional and multi-dimensional filtering are supported.

[0113] The electricity purchase and sale fee analysis module allows users to filter and query data by statistical month and statistical period (monthly or annually). It provides a visual display of monthly changes in electricity purchase and sale fees and transmission and distribution prices across multiple dimensions, including purchase type, direct transaction of electricity sales fees, and transmission and distribution prices by voltage level. Data tables show the current year, last year, year-on-year, growth amount, percentage, percentage change, electricity sales volume, permitted crude revenue, and monthly and annual changes in permitted revenue electricity price levels, categorized by type of electricity payable, electricity receivable by region, and transmission and distribution price levels.

[0114] The payment status analysis module allows users to filter and query based on power supply unit, statistical month, and statistical period (monthly or annually). It visualizes and displays the changing trends of payment amounts, payment channels, and payment details for each channel. The payment amount analysis calculates payment amounts for five cities on a daily basis; the channel payment analysis calculates the payment percentage for each channel and analyzes changes in that percentage; and the payment details for each channel, including WeChat, Alipay, 95598 website payments, and bank direct debit, show the number of payments, payment amounts, and percentages.

[0115] The integrated business management of this invention includes a data collection module, a report management module, and a report management module.

[0116] The data collection module uploads offline data such as electricity purchase data and agent electricity purchase prices, enabling comparative analysis of electricity purchase and agent electricity purchase related data.

[0117] The report management module generates various statistical tables based on dimensions such as power supply unit, electricity bill month / year, and electricity sales category, including: full-caliber electricity sales details (this month), full-caliber electricity sales details (cumulative), full-caliber pricing strategy classification details (this month), full-caliber pricing strategy classification details (cumulative), quantity and fee - full industry electricity sales statistics (full quantity), and quantity and fee - full industry electricity sales statistics (market-based).

[0118] The report management module assists analysts and decision-makers in producing reports. It generates exportable and compileable analytical reports based on the report content defined by business personnel.

[0119] Example 2

[0120] This embodiment presents an integrated intelligent analysis and decision-making method for electricity purchase and sale based on a data middle platform. The method includes:

[0121] Aggregate data from multiple business systems, clean and integrate the data to form a unified data resource;

[0122] Multiple business function microservices are built based on a microservice architecture; the multiple business function microservices call a unified data resource, including at least: electricity purchase and sale management microservice, electricity purchase and sale cost monitoring microservice, average electricity purchase and sale price analysis microservice, and business investigation and forecasting microservice.

[0123] Accept user requests and perform routing, authentication, and rate limiting;

[0124] Provide users with a visual interactive interface, including at least business dashboards, data analysis reports, and forecast report generation.

[0125] The operational survey and prediction microservice adopts the prediction model described in Example 1.

[0126] This invention develops and deploys microservices as foundational service components to support integrated purchasing and sales business processes. These foundational service components, serving as the core basic services of the entire system, provide functions such as identity authentication, data exchange, and service governance. The data platform, through deep aggregation with multiple system data sources and big data computing capabilities, achieves efficient data flow and integration. The platform collects data through the data platform, enabling the acquisition, cleaning, integration, and standardization of data from multiple business systems. The data platform also provides rich data processing tools and algorithms for data classification, aggregation, and prediction capabilities, thereby enabling in-depth data mining to uncover hidden business insights and trends. Based on the value of the analyzed data, it supports the formulation of accurate business decisions. This invention further optimizes integrated purchasing and sales analysis capabilities, improves enterprise operational efficiency, and reduces the burden of analysis work.

[0127] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A smart analysis and decision-making system for integrated power purchase and sales based on a data middle platform, characterized in that, The system includes: The data platform layer is used to aggregate data from multiple business systems, clean and integrate the data to form a unified data resource. The business service layer is built on a microservice architecture and contains multiple business function microservices; these multiple business function microservices call unified data resources. The service access layer is used for user requests and performs routing, authorization authentication, and rate limiting. The front-end application layer is used to provide users with a visual interactive interface, including at least business dashboards, data analysis reports, and forecast reports. The business function microservices include an operational assessment and prediction microservice, which includes: The rules configuration module is used to set the boundary conditions for electricity sales and transaction volume; The prediction generation module, based on historical electricity purchase and sales data, automatically generates prediction results for electricity sales volume, average electricity price, and electricity cost composition for a preset future period through a preset prediction model. The prediction review module is used to compare the prediction results with the actual data, calculate the prediction accuracy, and provide the reasons for the deviation. The electricity sales forecast in the forecast generation module uses the following model: ； in, t represents the predicted value at time point t; B is the shift factor; p is the non-seasonal autoregressive order; d is the non-seasonal differencing order; q is the non-seasonal moving average order; P is the seasonal autoregressive order; D is the seasonal differencing order; Q is the seasonal moving average order; s is the seasonal cycle length. For non-seasonal AR term polynomials; θ q (B) is a non-seasonal MA-term polynomial; Φ P (B s ) is a seasonal AR term polynomial; Θ Q (B s () is a seasonal MA-term polynomial; This is the random error term; This is a policy shock intensity index quantified based on policy documents. Meteorological influencing factors calculated based on meteorological data. , They are respectively , The corresponding regression coefficients.

2. The system according to claim 1, characterized in that, The data platform layer includes: Based on pre-configured CDC change data capture rules, incremental data is extracted in real time from the multiple business systems to form a first dataset; the first dataset includes electricity data and electricity price data; In the source layer, the first dataset is cleaned, and the cleaning includes: applying an outlier detection and smoothing algorithm based on box plot statistics to the electricity data, and applying term mapping and normalization based on a business rule dictionary to the electricity price data. In the shared layer, dimensional modeling is used to build a standardized model with consistent dimensions; At the analysis layer, based on the standardized model, data integration is performed through the Spark distributed computing engine to generate the unified data resource.

3. The system according to claim 1, characterized in that, The business dashboard includes: The configurable graphical orchestrator allows users to customize dashboard layouts and indicator components through drag-and-drop functionality. Real-time indicator data is monitored based on preset rules, and a visual alarm is automatically triggered when any indicator exceeds a preset threshold. In response to a user's click on any summary data point in the dashboard, the user is redirected to the corresponding business function microservice.

4. The system according to claim 1, characterized in that, The prediction generation module uses a machine learning regression model to predict the average electricity price, and the training process is performed using the following loss function: ； in, n The number of samples; y i For the first i The historical average electricity sales price of a sample; According to the first i Features of a sample xi The predicted average electricity price; This is a rigid cost floor calculated based on electricity purchase costs, transmission and distribution prices, and government funds. This is the constraint strength hyperparameter.

5. The system according to claim 4, characterized in that, The prediction generation module uses different models to perform combined predictions, with each model having specific weights. Calculated dynamically using the following formula: ； in, represents the importance coefficient of historical performance, and is a hyperparameter between 0 and 1; Accuracy (M) i ) For the model M i Historical accuracy; Accuracy (M) j ) For the model M j Historical accuracy; N The total number of models participating in the combination; Based on real-time business scenarios as the model The scene adaptability score is calculated.

6. The system according to claim 1, characterized in that, The system architecture is deployed on a cloud platform, and the business service layer adopts containerized orchestration; the microservices use Nacos as a service registration and discovery component for service registration and subscription. Microservices use RESTful calls and integrate the Sentinel component to achieve dependency isolation and circuit breaking / degradation. When a microservice call has an excessively long response time or a high failure rate, the system automatically downgrades to returning pre-prepared cached data.

7. The system according to claim 6, characterized in that, The containerized orchestration utilizes Kubernetes' HorizontalPodAutoscaler to automatically scale the number of business service layer Pod instance replicas based on the query per second (QPS) or the backlog length of the data platform layer task queue. The service registration and discovery component Nacos works in conjunction with Kubernetes' Service mechanism to achieve weight-based traffic scheduling and namespace-based multi-tenant service isolation.

8. A smart analysis and decision-making method for integrated power purchase and sales based on a data middle platform, characterized in that, The method includes: Aggregate data from multiple business systems, clean and integrate the data to form a unified data resource; Multiple business function microservices are built based on a microservice architecture; the multiple business function microservices call a unified data resource, including at least: electricity purchase and sale management microservice, electricity purchase and sale cost monitoring microservice, average electricity purchase and sale price analysis microservice, and business investigation and forecasting microservice. Accept user requests and perform routing, authorization, and rate limiting; Provide users with a visual interactive interface, including at least business dashboards, data analysis reports, and generated forecast reports; The operational survey and prediction microservice adopts the prediction model as described in any one of claims 1 to 5.

Images