Design and optimization strategy of electricity marketing information system supported by cloud computing platform

Abstract

This paper provides an in-depth discussion on the comprehensive requirements analysis, design implementation, algorithm optimization, and experimental evaluation of an electric power marketing information system, aiming to build a modern information system that is efficient, secure, and user-friendly. In the requirements analysis phase, the importance of business process optimization, data management analysis, security compliance, system integration and scalability is emphasized, while the diversified needs of end customers are considered. For the design and implementation part, the system architecture is based on microservices and cloud-native technologies to ensure high performance and security; and modularized development is achieved through Spring Boot, Vue.js and other technology stacks. For algorithm optimization, LSTM is used for power demand prediction and anomaly detection by combining integrated learning and self-encoder, which improves the prediction accuracy and anomaly identification capability. Experimental evaluation shows that the system demonstrates good performance, security and scalability in cloud computing environment, and the cost-effectiveness is significantly better than traditional deployment.

Introduction

With the continuous growth of global energy demand and the gradual opening of the power market, electric power enterprises are facing unprecedented competitive pressure and complex challenges. Under the wave of digital transformation, power marketing, as a bridge connecting power production and consumption, has a direct impact on the market competitiveness and customer service capability of enterprises through its informatization and intelligence level. The traditional power marketing information system is limited by problems such as fixed hardware resources, poor scalability, and limited data processing capacity, which makes it difficult to meet the rapidly changing needs of the current market. Especially in the context of the rapid development of big data and artificial intelligence technology, electric power enterprises are in urgent need of a new generation of information systems that can efficiently process massive data, provide accurate marketing strategies, and enhance customer satisfaction (Yanikara et al. 2020).

The emergence of cloud computing platform provides new possibilities for upgrading the power marketing information system. Cloud computing, with its flexible resource scheduling, powerful data processing capabilities, and low-cost operation and maintenance advantages, has become one of the key technologies to promote the digital transformation of the power industry. By migrating the power marketing information system to the cloud, companies can make full use of the cloud platform’s elastic computing, big data analysis, artificial intelligence services and other functions to achieve data-driven precision marketing, personalized services and efficient energy management, thus gaining an advantage in the fierce market competition. The impact pattern of cloud computing on the power marketing system is shown in Fig. 1.

Fig. 1

Impact model of cloud computing on power marketing system

In recent years, scholars at home and abroad have conducted extensive research on the application of cloud computing in the power industry. Literature shows that cloud computing technology shows great potential in enhancing power system flexibility, reducing operation and maintenance costs, and enhancing data processing capability. For example, some studies focus on the application of cloud computing platforms in smart grids, exploring how remote monitoring, fault diagnosis and preventive maintenance of power systems can be realized through cloud platforms (Du et al. 2018). Other studies have deeply analyzed the role of cloud computing in power demand side management, such as using cloud computing platforms for big data analysis to optimize load forecasting and scheduling (Shang et al. 2022). However, there are relatively few existing studies on the specific design and optimization of power marketing information systems, especially on how to maximize system effectiveness by combining specific algorithms, and how to improve customer experience and service quality while ensuring data security, and there is still much room for research in these areas. Therefore, this research will focus on the following aspects: (1) analyzing the applicability of different cloud service models (IaaS, PaaS, SaaS) in power marketing information systems, and designing cloud architectures suitable for the characteristics of power marketing business. (2) In-depth study of the application of algorithms (e.g., time series prediction, machine learning classification algorithms) in power consumption prediction, customer segmentation, personalized recommendation, assessing the performance of algorithms, and exploring the path of algorithm optimization.

Overview of cloud computing and power marketing information system

Cloud computing technology foundations and service models

Cloud computing, as a revolutionary computing resource delivery model, has a technical foundation that covers virtualization technology, distributed computing, parallel processing, automated management, and many other aspects, which together form the cornerstone of efficient cloud computing operation. Specifically: virtualization technology is the core of cloud computing, which allows physical resources (e.g., servers, storage devices) to be abstracted into logical resources, so that multiple users can share the same physical resources without interfering with each other. vMware vSphere and KVM are widely used virtualization platforms in the industry. Distributed computing and storage improves the reliability and processing power of the system by splitting the computational tasks to be executed on multiple computers over a network and storing the data on multiple nodes. Hadoop and Spark are representative frameworks for realizing large-scale data processing (Lee and Tseng 2023). Cloud computing service models are categorized into three main types depending on the level of service provided: Infrastructure as a Service (IaaS), such as Amazon AWS, Microsoft Azure, provides a lease of computing, storage and network infrastructure on which users deploy and run their own software. Platform as a Service (PaaS), e.g. Google App Engine, provides a platform for application development, testing and deployment, where users do not have to manage the underlying hardware and operating system. Software as a Service (SaaS), e.g., Salesforce, Office 365, provides applications directly to end-users, which users can access through a browser or other lightweight client (He et al. 2020).

Basic framework and functional requirements of electricity marketing information system (EMIS)

PMIS is built on top of several key modules to form a highly integrated information processing and business support platform (Olmstead et al. 2020). PMIS is shown in Fig. 2. In the framework, the customer relationship management (CRM) module is responsible for maintaining customer files, service records and interaction history, and customizing personalized service strategies through customer satisfaction surveys; the billing and account management module ensures accurate calculation of electricity charges and efficient account processing; the load management module forecasts electricity demand based on historical data analysis, guides demand-side management strategies, and balances the load on the grid; the market analysis module uses advanced data analysis techniques to analyze market dynamics; and the market analysis module uses advanced data analysis techniques to analyze market dynamics. The market analysis module utilizes advanced data analysis techniques to analyze market dynamics and provide a basis for the scientific formulation of marketing strategies; and the customer service and technical support module enhances customer experience and satisfaction through a multi-channel response mechanism.

Fig. 2

PMIS framework

At the level of functional requirements, PMIS focuses on several key aspects: first, strengthening data integration and management capabilities to ensure efficient data collection and integration as well as the timeliness and accuracy of the information (Kaszynski et al. 2023); second, introducing intelligent analytical tools to conduct in-depth customer behavioral analysis, load forecasting, and anomaly monitoring with the help of big data and machine learning technologies to assist management in decision-making (Williams and Green 2022); and further, focusing on system integration and interoperability to achieve smooth interfacing of internal modules and with external systems (finance, ERP systems, etc.) to promote business process automation; at the same time, in view of the sensitivity of the data, the system must implement stringent security measures and privacy policies, such as data encryption and access control (Copping et al. 2018); finally, emphasize the system design’s flexibility and scalability to flexibly adapt to market changes and support continuous business expansion (Wu et al. 2022). Together, these requirements form the core framework for PMIS development, driving the transformation of electric utilities to a smarter and more efficient service model.

PMIS use cases

The development of smart grid technology has facilitated the widespread use of remote automatic meter reading (AMR) systems, significantly improving the accuracy and efficiency of data collection (Saghravanian et al. 2020). For example, by deploying an AMR system in combination with GIS (geographic information system) and big data analysis, a power company in China has realized real-time power monitoring and anomaly detection for millions of users, significantly reducing the error rate and cost of manual meter reading (Beni and Sheikh-El-Eslami 2021). According to Soofiabadi and Foroud (2022), an electric utility company implemented a personalized customer communication strategy through the CRM module of PMIS, including sending regular electricity consumption suggestions, energy saving solutions, and satisfaction surveys, which effectively improved customer satisfaction and loyalty. Combined with data analysis, the enterprise was able to respond quickly to customer needs and provide customized services, thereby maintaining an edge in a competitive market. With the increasing application of big data in power marketing, the risk of data leakage and privacy violation increases (Cui et al. 2022). Ensuring the secure transmission, storage, and compliant use of massive customer data has become a key challenge for PMIS. Integration between different business modules and interoperability with external systems often require complex interface design and data standardization work (Kara et al. 2022). How to achieve efficient and flexible system integration while ensuring data consistency is a major challenge in PMIS implementation. The rapid iteration of technology requires PMIS to be constantly upgraded to cope with the convergence of emerging technologies such as blockchain and artificial intelligence (Afanasyev et al. 2022). At the same time, the adaptability of employees to new technologies and professional training become important factors in improving the effectiveness of the system. Adopt advanced encryption technology, access control mechanisms and data desensitization processing, combined with regular security audits and employee security awareness training, to establish a multi-level data protection system (Guo et al. 2019). Promoting business process standardization and modular design and adopting microservice architecture can improve system flexibility and scalability and reduce integration complexity (Candra et al. 2018). Establishing a continuous technology tracking mechanism, investing in the research and development and application of emerging technologies, as well as increasing the technical training of employees to ensure that the team can quickly master and apply the latest technology (Jia et al. 2022).

To address the inadequacy in the review and analysis of existing literature, particularly concerning the application of cloud services in contemporary power systems, it is crucial to underscore the current landscape and prevailing trends that form the backdrop of our study. A comprehensive examination of recent publications reveals a growing trend towards the integration of cloud computing in the electricity sector, driven by the need for scalable resources, enhanced data management capabilities, and improved operational efficiency (Perninge and Eriksson 2018). Despite this progression, the literature highlights a disparity in the depth and breadth of cloud adoption across different segments of the power industry. Some studies emphasize successful implementations of cloud-based SCADA (Supervisory Control and Data Acquisition) systems for remote monitoring and control, while others report on the deployment of cloud platforms for advanced analytics and demand response management.

However, these accounts often lack a systematic analysis of the challenges faced during migration to the cloud, the specific security concerns inherent to cloud environments, and the measures taken to ensure uninterrupted service availability – aspects that are pivotal to our research’s validity and relevance. By bridging this gap, our study not only contributes a nuanced understanding of the practical implications of cloud services in electricity systems but also offers empirical evidence of performance improvements post-optimization, thereby reinforcing the significance of our findings. Through a meticulous review augmented by our empirical results, we aim to clarify misconceptions, highlight best practices, and propose strategies for overcoming barriers to wider adoption, thereby enriching the discourse on cloud-enabled transformation in the energy sector (Wu et al. 2020).

Demand analysis of power marketing information system

This chapter will delve into the requirements of the power marketing information system to ensure that the design and optimization strategies can accurately match the business and technical requirements. Requirements analysis is the cornerstone of system development and is essential for building an efficient, user-friendly, and sustainable power marketing information system. This chapter is divided into three subsections for detailed elaboration:

Analysis of business requirements

Business Requirements Analysis (BRA) is a foundational step in building a power marketing information system, which aims to identify and define the business functionality and performance requirements that the system must meet to support the efficient operation of an electric utility, improve service quality, and facilitate business growth. This section will expand on the following key aspects in detail:

Business Process Integration and Optimization. Electricity marketing business covers a wide range of processes, including user registration, electricity consumption application, metering management, electricity bill calculation and collection, customer service and complaint handling. The system needs to support the automation and seamless integration of these processes to reduce manual operations and improve work efficiency. For example, by integrating the online application function, it simplifies the user application process and realizes the full digital management from application to power connection; in the electricity charge calculation process, the system needs to support complex billing rules (e.g., time-of-day tariffs, ladder tariffs) and quickly adapt to policy adjustments to ensure accurate billing, and the specific business hydrangea framework is shown in Fig. 3.

Fig. 3

Business requirements framework

Data management and analysis capabilities. In the context of growing massive data, power marketing information systems need to have strong data management and analysis capabilities. This includes real-time collection, storage, cleansing and analysis of customer electricity consumption data to support business decisions and market analysis. The system should support efficient retrieval of data and report generation, and provide management with key business indicators (KPIs) to monitor, such as average fault response time and tariff recovery rate. In addition, through in-depth analysis of historical data, the system should be able to provide advanced functions such as load forecasting and user behavior analysis to help companies optimize resource allocation and enhance service foresight and personalization.

Security and Compliance. Given the sensitivity of data in the electric power industry, system security and compliance is a part of business requirements analysis that cannot be ignored. System design needs to comply with industry security standards and legal and regulatory requirements, such as data encrypted transmission, access control mechanisms, anti-tampering and anti-leakage measures, etc., to ensure the security of customer data and business data. At the same time, the system should support data privacy protection and follow relevant regulations such as GDPR and Cybersecurity Law to protect user privacy rights and interests (Wu et al. 2020).

System integration and interoperability. The power marketing information system does not exist in isolation, and it needs to be tightly integrated with other systems within the enterprise (e.g., financial system, CRM system) as well as external partners (e.g., banks, third-party payment platforms). Therefore, the system design should consider interface standardization and protocol compatibility, and support seamless data exchange and process synergy to enhance the coherence and efficiency of business processing.

To refine the demand analysis for the power marketing information system, it is imperative to adopt a structured methodology that clearly delineates between functional and non-functional requirements. Functional requirements encapsulate the system’s operational capabilities, such as automated business process handling, real-time data analytics for informed decision-making, and secure data transactions. Non-functional requirements, on the other hand, pertain to performance attributes like system responsiveness, scalability to accommodate growing user bases, and adherence to regulatory compliance for data privacy and security.

Priority setting within these requirements is pivotal. Immediate focus should be directed towards enhancing process automation and integration, ensuring data accuracy and security: the lifeblood of the electric utility business. Functionalities enabling quick adaptation to billing policy changes and facilitating efficient complaint resolution are pressing needs. Meanwhile, fostering a robust data analytics backbone that not only handles big data efficiently but also provides actionable insights for load forecasting and personalized services becomes a strategic long-term goal.

Non-functional priorities necessitate a reliable and responsive system infrastructure capable of maintaining high uptime, coupled with seamless interoperability with external financial and CRM systems. Robust cybersecurity measures, adhering to GDPR and local cybersecurity laws, are non-negotiables to safeguard user trust and legal compliance.

By systematically categorizing and prioritizing these requirements, the development roadmap gains clarity, allowing for phased implementation that addresses immediate operational efficiencies while laying the groundwork for future scalability and service enhancements. This structured approach ensures the power marketing information system is not just a technological solution but a strategic enabler.

Customer needs analysis

Customer demand analysis is an indispensable part of the development process of power marketing information systems, which focuses on the actual needs and expectations of end-users to ensure that the system is able to provide a service experience that meets the diverse needs of users and enhances user satisfaction and loyalty. This section will focus on several core dimensions to explore customer demand analysis in depth. The specific customer requirements framework is shown in Fig. 4.

Fig. 4

Customer requirements framework

User experience and interface friendliness. As modern users expect more and more from digital services, power marketing information systems need to provide intuitive and easy-to-use user interfaces (UI) and excellent user experience (UX). This includes a clear and concise operation process, responsive interactive feedback, and a design layout that conforms to user habits. The system interface should be adaptable to different devices and screen sizes, and support multi-language switching to ensure accessibility for all user groups. In addition, considering that some older users may be unfamiliar with the technology, the system should provide assistive features, such as voice navigation and large font mode, to enhance its usability (Hao et al. 2023).

Personalized services and customized solutions. Customers’ needs for electricity services vary widely, and the system should provide personalized service options, such as recommending energy-saving solutions and customized tariff packages based on the user’s historical electricity consumption behavior. Through data analysis, the system is able to recognize user preferences and push relevant information and services, such as notification of power outages and reminders of peak periods of electricity consumption, to enhance the relevance and value of the service. In addition, a self-service platform is provided to allow users to adjust their service plans according to their needs, increasing the sense of participation and control.

Multi-channel service and interaction. In order to enhance service accessibility and convenience, power marketing information systems need to support multi-channel access, including web-based, mobile applications, social media, customer service hotlines, and so on. Ensure that users can access information, conduct business and seek help through their most preferred channels. At the same time, the system should integrate intelligent customer service robots that utilize natural language processing technology to provide 24/7 instant response, effectively solve common problems, reduce the pressure on manual customer service, and improve response speed.

Transparency and trust building. Users have high requirements for the transparency of electricity billing and related policies. The system should clearly display the billing basis, tariff structure, preferential policies and other information, and provide detailed bill descriptions and analysis of historical electricity consumption data to help users understand the composition of the fees and establish a sense of trust. At the same time, a feedback mechanism is set up so that users can easily raise questions or make suggestions, respond to their concerns in a timely manner, and continuously optimize the service.

Design and realization of power marketing information system

System design

As shown in Fig. 5, the system architecture uses Vue or React to build a responsive user interface, enabling dynamic content and modular components. The business logic layer relies on microservices, such as billing and customer information, communicated through REST and gRPC, and managed centrally by the API gateway. The data layer uses a mix of MongoDB, PostgreSQL and Redis to optimize performance, and ClickHouse helps with analysis. Infrastructure is deployed on cloud platforms (EC2, ECS, Elastic Beanstalk, Kubernetes), integrating monitoring and logging to ensure stability and efficient operation and maintenance.

Fig. 5

System architecture

System architecture design

The architectural design of the power marketing information system is the foundation for the solid operation of the entire system. This section will outline the system architecture from the macro level to ensure the high availability, scalability and security of the system. The system adopts a microservice architecture based on cloud-native design principles, which is divided into the following levels: the user interface layer adopts responsive design principles to ensure good rendering on different devices and browsers, and to provide a unified and friendly interaction experience. Integrate the latest front-end technology stack, such as Vue.js or React, to achieve dynamic loading, componentized development, and improve page responsiveness and maintainability. The business logic layer decouples and modularizes business logic through microservice architecture, with each service focusing on a single responsibility, such as billing service, customer information service, and so on. Communication between services is realized through RESTful API or gRPC, and API Gateway is used for unified management, including authentication, flow limiting and routing. The data service layer is built on top of a distributed database system, using NoSQL databases (e.g., MongoDB) to handle unstructured data and relational databases (e.g., PostgreSQL) to handle structured data. Data caching technology (e.g. Redis) is introduced to improve read efficiency, and data warehouses (e.g. ClickHouse) are used for complex data analysis and report generation. The infrastructure layer relies on public or private cloud platforms, utilizing IaaS services (e.g. EC2, ECS) to provide computing resources, and PaaS services (e.g. Elastic Beanstalk, Kubernetes) to realize application hosting and automatic capacity expansion and contraction. Integrate cloud monitoring and logging services to ensure stable system operation.

This brief overview of user experience (UX) and interface design (UI) emphasizes their centrality to building an effective power marketing information system. UX design focuses on understanding user needs, preferences, and behavior patterns to create interactive processes that are intuitive, easy to use, and meet user expectations. Good UX design can reduce operational complexity, improve user satisfaction and system adoption, and thus have a positive impact on business performance.

Interface design is the visual and interactive embodiment of this process, which focuses on how to lay out information, design interface elements and interaction patterns in the most direct and effective way to facilitate users to achieve their goals. In the case of power marketing information systems, this means designing for both professionalism and friendliness, ensuring smooth operation for both technicians and consumers, such as intuitive dashboard displays, one-click service requests, and responsive designs that accommodate multiple devices.

Integrating the latest design concepts, such as flat styling, color psychology applications, and the use of animation and micro-interaction to improve feedback quality, are key to improving the overall experience. At the same time, UI/UX design is iteratively optimized through user testing, interviews and data analysis to ensure that the system not only reflects current user needs, but also foresees and adapts to future changes, providing users with a personalized and efficient use environment. In short, excellent user experience and interface design are important cornerstones to enhance the competitiveness of power marketing information systems and deepen user loyalty.

Security architecture design

Security is an integral part of system design, and this section will focus on how to build a secure and reliable information system: Identity and access management adopts OAuth 2.0 and OpenID Connect standards to realize Single Sign-On (SSO) and fine-grained privilege control. JWT is utilized for stateless authentication to enhance security and efficiency. Data is encrypted during transmission using TLS/SSL protocol, and static data is stored using standard encryption algorithms such as AES256. Database field level encryption is realized to protect sensitive information. We set up network access control lists (ACLs) and security group rules to allow access to only the necessary ports and services to prevent malicious attacks. We use integrated security log collection and analysis tools, such as ELK Stack (Elasticsearch, Logstash, Kibana), to record and audit system operations, monitor the system security status in real time, and respond quickly to security events (Lin et al. 2023).

Design for performance and scalability

We use the load balancing provided by the cloud platform to disperse the traffic and automatically adjust the number of service instances according to the business volume to ensure that the system stably responds to highly concurrent accesses. We introduce message queues (e.g. RabbitMQ, Kafka) to handle asynchronous tasks, such as email sending, data synchronization, etc., to reduce system response latency and improve processing efficiency. We use Docker containerization technology, combined with Kubernetes for service orchestration, to achieve rapid deployment, rapid recovery and flexible expansion of services. With the above design, the power marketing information system aims to build a flexible, secure, and efficient business support platform to meet the diverse needs of power marketing, while reserving enough expansion space for future business growth and technology evolution.

At the practical level, the integration of new systems with existing enterprise systems (ERP, CRM, etc.) is a meticulous and highly strategic engineering process to ensure data flow, process collaboration, and business optimization. Take the case of a retail company that plans to integrate its newly developed intelligent supply chain management system (SCM) with its existing SAP ERP system and Salesforce CRM platform. The integration process first involves adopting standardized data exchange protocols such as XML or JSON, and RESTful API interfaces to bridge communication between different systems and achieve real-time synchronization of data. Second, configure middleware solutions such as IBM Sterling Integrator or MuleSoft to transform data formats and map fields to ensure that both systems understand and process information coming from each other. In terms of security, OAuth2 and SSL/TLS encryption technologies are deployed to maintain the security of data transmission. In addition, the incremental integration strategy is adopted, which is piloted in a small scale and then promoted comprehensively to gradually verify the compatibility and stability of the system and reduce the overall risk. Throughout the integration project, special emphasis was placed on user training and process reengineering to ensure a smooth transition of employees to the new system, while optimizing business processes to maximize integration benefits. Through such a step-by-step and all-round strategy, the integration of the new system with ERP and CRM not only solves the technical connection problem, but also ensures the continuity and efficiency improvement of enterprise operations, providing clear feasibility proof for stakeholders.

System implementation

Technology selection and environment setup

We choose Spring Boot as the main framework, because of its simple and fast characteristics, can quickly build RESTful API. Integrated Spring Cloud to manage microservices, using Spring Security to deal with security issues, to ensure the security of communication between services. We use Vue.js framework, combined with Vuex for state management, Element UI or Vuetify as a UI library to improve development efficiency and user experience. We utilize Webpack for module packaging to ensure effective loading of resources. We choose MySQL as the core relational database to handle core business data such as orders, users, etc. MongoDB is used to store unstructured data such as user behavior logs.Redis is used as a caching database to accelerate data access speed. We use Docker to containerize all services and dependencies to ensure the consistency of development, testing and production environments. Kubernetes (K8s) is responsible for the deployment and management of services, and the implementation of automated operation and maintenance. We integrate Jenkins or GitLab CI/CD processes to automate code build, test, and deployment, shorten development-to-live time, and improve software delivery quality (Biely et al. 2018).

Core module implementation

The core functions of this power marketing information system cover four modules: customer management, electricity billing, payment and data analysis. The customer management module uses Vue.js to create an interactive interface and Spring Data JPA to simplify database operations, and Apache Shiro or Spring Security to implement strict rights management to protect customer information security. The electricity billing module uses intelligent algorithms to consider peaks and valleys and ladder tariffs, efficiently handles billing through Java multi-threading technology, and utilizes message queues to achieve asynchronous communication to ensure system response. The payment module seamlessly integrates WeChat Pay, Alipay and other mainstream payment methods to ensure smooth and safe payment, and realizes real-time tracking of order status and exception management to maintain transaction integrity. The data analysis and reporting module utilizes Hadoop, Spark and other big data technologies to deeply explore the value of data, and combines Tableau or Power BI and other BI tools to present intuitive visualization reports, providing management with a strong basis for data decision-making. This series of well-designed modules together build an efficient, secure and intelligent power marketing service platform (Frei et al. 2018).

System testing and optimization

First unit and integration testing, we use JUnit and Mockito for unit testing of back-end services to ensure the stability and correctness of each microservice. Integrate Spring Cloud Contract for consumer-driven contract testing to ensure the consistency of the contract between services. We conduct stress testing with the help of JMeter or Gatling to simulate high concurrency scenarios, identify and solve performance bottlenecks, such as optimizing database queries and increasing caching strategies. We organize real users to participate in Beta testing to collect feedback and tune the ease of use of the front-end interface and the responsiveness of the back-end (Silva et al. 2019).

Algorithm optimization

LSTM networks are selected for power demand forecasting because of their ability to deal effectively with long-term dependence of time series data. Their performance exceeds ARIMA models, showing lower error and higher stability. Ensemble learning combines the efficient anomaly detection of isolation forest with the feature extraction ability of automatic encoder, which can identify anomalies more accurately than single model and improve the detection rate. Therefore, LSTM and integrated learning are the preferred solutions to solve complex power system challenges with unique advantages.

Electricity demand forecasting is an important part of the electricity marketing information system, which is optimized using the long short-term memory network (LSTM) in deep learning to improve the accuracy and stability of forecasting. LSTM deals with the long-term dependency problem in time-series data through a unique gating mechanism (Sirjani and Rahimiyan 2018).

The key steps in the LSTM cell include the computation of the forgetting gate, the input gate, the cell state update, and the output gate. The formula for the forgetting gate is ${f_t} = \sigma ({W_f} \cdot [{h_{t1}},{x_t}] + {b_f})$. Where ${f_t}$ denotes the output of the oblivion gate, $\sigma$ is the sigmoid activation function, ${W_t}$ is the weight matrix of the oblivion gate, $\left[ {{h_t}_1,{x_t}} \right]$ is the splicing of the hidden state of the previous time step with the current input, and ${b_f}$ is the bias term. Input gates: ${i_t} = \sigma \left( {{W_i} \cdot \left[ {{h_i},{x_t}} \right] + {b_i}} \right)$, ${{\tilde c}_t} = \tanh \left( {{W_c} \cdot \left[ {{h_i},{x_t}} \right] + {b_c}} \right)$ denote the activation and candidate cell states of the input gates, respectively. The cell state update formula is ${c_t} = {f_t} \odot {c_n} + {i_t} \odot {{\tilde c}_t}$. Where $\odot$ denotes the element multiplication and ${c_t}$ is the updated cell state. The output gates ${o_t} = \sigma \left( {{w_o}\left[ {{h_t}_1,{x_t}} \right] + {b_o}} \right)$, ${h_1} = {o_t}\, \odot \,\tan h\left( {{c_t}} \right)$ determine which cell state information is used for output.

We use the attention mechanism to increase the attention to key information through weighted summation. The formula is as follows: ${a_1} = soft\max \left( {{w_a} \cdot \tanh \left( {{W_a} \cdot {h_t} + {b_a}} \right)} \right)$, ${a_t} = soft\max \left( {{w_a} \cdot \tanh \left( {{W_a} \cdot {h_t} + {b_a}} \right)} \right)$ where ${a_t}$ is the attention weight and ${c_t}$ is the weighted context vector (Mozdawar et al. 2022).

Anomaly detection plays a crucial role in the operation and maintenance of power systems, which can quickly identify and warn of potential abnormal operating states, thus effectively preventing failures and safeguarding the stability and security of power grids. In order to further improve the accuracy and efficiency of anomaly detection, an innovative approach has emerged, which is the fusion of the Isolation Forest algorithm in integrated learning and the Autoencoder in deep learning technology.

The Isolation Forest algorithm, as an efficient unsupervised learning method, is unique in that it adopts the framework of a random forest, but with a different goal than traditional classification or regression tasks. Each decision tree is random and shallow, and “isolates” data points by minimizing path lengths. In this process, points that are at the edge of the data distribution and can be easily separated from the rest of the data by a small number of cuts are given a high anomaly score. This means that Isolation Forest quantifies the degree of anomaly by calculating the average number of times a data point is randomly cut (the so-called path length), where a smaller average number of splits implies that the data point is deviating from the normal pattern, and is most likely anomalous behavior in the system (Xie et al. 2023).

The core capability of self-encoders, an important branch of deep learning, lies in learning an efficient low-dimensional representation of the data through a process of encoding and then decoding the data. In an anomaly detection scenario, the self-encoder is trained to reconstruct normal input data. When anomalous samples are encountered, the self-encoder struggles to reconstruct them accurately because they are significantly different from the normal patterns in the training set, thus recognizing these anomalies through reconstruction errors. The advantage of this approach is that it automatically captures complex patterns and structures in the data, improving the adaptability and accuracy of anomaly detection. Combining Isolation Forest with self-encoder, this hybrid strategy not only fully utilizes Isolation Forest’s efficiency in processing high-dimensional data and large-scale datasets, but also leverages the self-encoder’s ability to learn deep features and patterns of the data. Working in tandem, the two not only enhance the model’s sensitivity to recognizing anomalous behaviors, but also improve its ability to capture complex and hidden anomalous patterns. This comprehensive approach not only responds more quickly to abnormal behaviors in the power system, but also further ensures the comprehensiveness and accuracy of the detection, providing strong intelligent support for the safe operation of the power system. Self-encoder loss function: $L\left( {x,\hat x} \right) = {1 \over N}\sum\limits_{i = 1}^N {\parallel {x_i},{{\hat x}_i}{\parallel ^2}}$, where L is the mean square error loss function, ${x_i}$ is the input data, and ${{\hat x}_i}$ is the output of the self-encoder reconstruction (He et al. 2020).

In our implementation of LSTM for electricity demand forecasting, we fine-tuned several key parameters to optimize performance. These include the number of LSTM layers, the size of the hidden layer, and the dropout rate to mitigate overfitting. We employed grid search and cross-validation techniques to systematically explore the parameter space, with an emphasis on minimizing prediction errors while maintaining model robustness. To ensure the best fit, we also incorporated early stopping based on validation loss to prevent unnecessary training iterations. Additionally, batch normalization was applied after each LSTM layer to normalize the outputs and accelerate the training process. Through this rigorous tuning process, we achieved a forecasting accuracy improvement of 5% compared to a baseline LSTM model without such optimizations.

Comprehensive security architecture and management

When addressing the multifaceted nature of security requirements in power marketing information systems, a dedicated approach to integrating and managing these measures is critical. This section delves into the complexity of the system security architecture and clarifies the strategies used to protect data integrity, confidentiality, and availability at all operational layers.

The foundation of our security framework is a comprehensive threat modeling exercise. By identifying potential vulnerabilities and mapping attack vectors, we build an active defense strategy. This involves categorizing threats from external hacking attempts to internal threats and assessing the likelihood and impact of each scenario. Regular risk assessments ensure that our defences evolve in step with emerging threats to maintain continuous security readiness.

Encryption techniques such as Transport Layer Security (TLS) for data in transit and Advanced Encryption Standard (AES) for data at rest protect sensitive information from unauthorized access. Multi-factor authentication (MFA) is a mandatory implementation of authentication across all user access points to reduce the risk of unauthorized access. Intrusion detection systems (IDS) and intrusion prevention systems (IPS) are deployed to monitor and react to potential threats in real time.

We strictly comply with international and regional data protection regulations, including the General Data Protection Regulation (GDPR). This involves implementing privacy through design principles, ensuring data minimization, providing data subjects with rights to their personal information, and conducting regular audits to verify compliance. Perform privacy impact assessments (Pias) for new projects or system changes to assess their potential privacy risks and mitigation strategies.

A comprehensive incident response plan outlines procedures for detecting, controlling, eliminating, and recovering from security incidents. Regular training and simulations enable response teams to deal with incidents quickly and efficiently, minimizing losses and downtime. Clear communication channels and predefined escalation paths ensure fast decision-making during crises.

To maintain a reliable security posture, system logs, network traffic, and user activity are continuously monitored. Machine learning algorithms analyze patterns to detect anomalies and flag them for human review or automatic response. Regular security audits and penetration tests further enhance the resilience of the system and feed results back into the improvement cycle.

By adopting this holistic approach to security architecture and management, the power marketing information system not only meets but exceeds industry standards for cybersecurity. It ensures stakeholder trust.

Experimental evaluation

Experimental design

This evaluation aims to comprehensively analyze the integrated performance of the power marketing information system supported by the cloud computing platform, focusing on its performance, security, scalability and cost-effectiveness. The experiment relies on mainstream public cloud platforms (e.g., Aliyun, AWS) and builds an infrastructure that includes ECS instances, RDS database, Kafka message queues and Redis caching services. The dataset is designed to cover 100,000 customer information, 1,000 daily electricity transactions in a year, and high-frequency electricity consumption data of 10,000 users in a consecutive month, which is used as the basis for in-depth testing.

To ensure realistic evaluation, detailing the dataset’s selection and composition is vital. Mirroring the power marketing sector, our dataset contains 100,000 diverse customer profiles, reflecting varied consumption habits and demographics, ideal for CRM testing. Incorporating 1,000 daily electricity transactions over a year captures seasonal fluctuations and billing complexities, enabling rigorous billing module testing under dynamic loads. High-frequency consumption data from 10,000 users for a month adds precision to demand forecasting, assessing grid management responsiveness.

Acknowledging limitations, the synthesized dataset, despite its representativeness, may not cover all real-world edge cases. Its scale, though extensive, is a subset of the broader, evolving power market. Thus, while experiments offer insightful performance benchmarks, continuous validation with live data and system adjustments are necessary for practical implementation.

In essence, the dataset design focuses on realism and relevancy, bridging the simulation-to-application gap. This meticulous construction establishes a solid foundation for a rigorous experimental framework, credibly evaluating the cloud-powered marketing system’s capabilities.

By using JMeter to implement concurrent access stress testing, we simulate peak user behavior and quantify system response time and processing throughput; at the same time, we perform penetration testing to verify the system’s ability to resist security threats such as SQL injection and XSS attacks. To assess scalability, virtual machine instances are gradually expanded and the trend of system processing power increase is monitored. Cost-benefit analysis involves documenting the consumption of cloud resources under different load conditions and comparing it with the cost of a traditional self-built data center to assess the economic rationality of the cloud computing solution. The overall evaluation framework aims to provide empirical evidence and strategic guidance for the design and optimization of power marketing information systems in a cloud computing environment.

Experimental results

Table 1 demonstrates the response time of the main functional modules (login page, query electricity bill, online payment) of the power marketing information system under different load levels. The data is obtained by JMeter recording during simulated user operations, reflecting the performance of the system under low, medium and high load pressure. The response time rises as the load increases, reflecting the change in the system’s processing power.

Table 1 System response time (in milliseconds)

Table 1 shows the response time of the power marketing information system for three main operations (login page, query electricity bill, and online payment) under different load levels. The response time is the elapsed time from the user initiating the request to the system completing the response. As the load level increases (low, medium, and high), the response time generally increases, reflecting the performance of the system under the pressure of higher user access. For example, under low load, the response time for the login page is 150 milliseconds, while under high load, this time extends to 450 milliseconds.

Table 2 records the number of requests per second that the system is able to handle in response to different loads, for the same functional modules (login page, query electricity bill, online payment). The throughput increases with the number of instances, showing the ability of the system to handle concurrent requests. The data is derived by gradually increasing the number of concurrent users and measured in the test tool. Throughput is the number of requests that the system can handle per unit of time and is an important indicator of the system’s processing power. Under low load, the system was able to handle 20 login requests/second, while under high load, this capability was enhanced to 60 requests/second.

Table 2 Throughput (in requests/second)

Table 3 summarizes the test results of system security against three common network security threats: SQL injection, XSS attack and DDoS protection. All the test results are “Pass”, which means that the system has taken effective security measures and can successfully defend against these types of attacks to protect user data and system security.

Table 3 Safety test results (pass/fail)

Table 4 demonstrates the scalability of the system’s processing power by increasing the number of system instances. As the number of instances increases from 1 to 4, the throughput of the system increases from 30 to 90 transactions/second, indicating that the system is designed to have good horizontal scalability and to be able to improve the overall performance by adding resources when needed.

Table 4 Scalability tests (in number of transactions processed per second)

Table 5 compares the cost components and their total costs for the two models of traditional and cloud deployment. It can be seen that the cloud deployment demonstrates higher cost effectiveness in general, with the total cost reduced by $9,000/month compared to the traditional deployment.

Expanding upon the cost-benefit analysis presented in Table 5, a deeper dive into the financial benefits underscores the advantages of transitioning to a cloud-based infrastructure. Initial setup expenses for the cloud model, while not insignificant, are offset by the elimination of substantial capital expenditures typically associated with on-premise hardware installations. This transition translates to an upfront saving that redirects funds towards other strategic initiatives. In terms of ongoing maintenance, the cloud deployment model reduces labor costs by approximately $4,000 per month, highlighting the efficiency gains from leveraging managed cloud services. These services automate routine tasks, update software seamlessly, and reduce the need for dedicated IT personnel, thereby streamlining operations and lowering overheads. Other direct costs, such as energy consumption and physical space requirements, are also significantly curtailed under the cloud strategy. A reduction of $1,000 monthly in these costs underscores the economies of scale that cloud providers can offer, where infrastructure efficiencies are shared among multiple clients. To provide a comprehensive view of financial viability, it’s pertinent to assess the Return on Investment (ROI) and the payback period for the cloud migration. Considering the monthly savings of $9,000, if the one-time cost of migration is estimated at $200,000, the investment would break even in approximately 22 months. Beyond this point, the savings directly contribute to ROI, with potential for a substantial return over the system’s lifecycle. Moreover, factoring in the potential savings from avoided downtime and enhanced productivity, the ROI calculation becomes even more favorable. Cloud systems’ inherent reliability and scalability facilitate business continuity and agility, translating into indirect benefits that could further accelerate the payback period and amplify long-term profitability.

Table 5 Cost-benefit analysis (in $/month)

Figure 6 assesses customer satisfaction and internal efficiency gains for several core functional modules within the system. Customer satisfaction reflects the degree of user satisfaction with the system in terms of ease of use and functionality implementation, while internal efficiency enhancement measures the effectiveness of the system in optimizing internal workflows. As can be seen from the table, the Data Analytics module scored the highest in both assessments, showing its excellent performance in providing high-quality services and improving internal operational efficiency.

Fig. 6

Case evaluation (Satisfaction Score out of 5)

Our system underwent a comprehensive penetration testing phase, performed by certified ethical hackers, to verify its ability to withstand SQL injection and XSS attacks. The test scenarios were designed to simulate real-world attack vectors, including attempts to manipulate queries and inject malicious scripts. The results show zero successful violations, validating the effectiveness of implemented measures such as SQL query preparation statements and input cleanup combined with output encoding to eliminate XSS threats. In addition, we have integrated a Web Application Firewall (WAF) to provide an additional layer of protection, logging and blocking suspicious requests in real time. Regular security audits and updates to our defense mechanisms ensure ongoing protection from emerging threats.

To rigorously assess the scalability of the system, we simulated rapid growth of the user base by gradually deploying additional VM instances on the cloud platform. Load testing was performed using a virtual model tool of real user behavior, gradually increasing the load until the expected peak usage was reached or higher. Testing has shown that even with a 1000% increase in user traffic, our system maintains sub-second response times and CPU and memory utilization remain within acceptable thresholds. However, we identified database read operations as potential bottlenecks, which prompted us to implement database sharding and caching policies. These enhancements ensure linear scalability, allowing future growth to be handled seamlessly without compromising performance.

We conducted structured surveys using the System Availability Scale (SUS) before and after implementing system optimization. After optimization, the SUS score increased significantly from 72 to 86, indicating significant improvements in user satisfaction and usability. Qualitative feedback collected through open-ended questions highlighted faster response times and intuitive interface design as key factors in enhancing the user experience. In addition, we integrate real-time user analytics to monitor engagement metrics such as session duration, click-through rates, and conversion events to provide continuous feedback for iterative improvement. This data-driven approach to UX enhancement ensures that our systems are aligned with user needs and preferences.

Conclusion

After comprehensive discussion and practice, this study successfully constructs a future-oriented power marketing information system, which not only meets the complex business needs of power enterprises, but also achieves remarkable results in enhancing user experience, optimizing operational efficiency, strengthening security compliance and adapting to future development. The system design strictly follows the business and customer needs, realizes the leap from traditional processes to digitalization and automation, and especially makes substantial progress in business process integration, data-driven decision-making, construction of security system and integration of multi-channel services. The technical innovation points focus on the in-depth optimization at the algorithm level, using LSTM and integrated learning technology, which not only improves the accuracy of the prediction model, but also brings new perspectives for anomaly detection and strengthens the intelligence level of the system. The microservicing design and cloud-native practices of the system architecture ensure that the system is highly flexible, scalable, and cost-effective, adapting to the rapidly changing needs of the power market. Experimental evaluation results strongly demonstrate the new system’s excellent performance, security, and scalability, especially in high concurrency processing, defense against security threats, and dynamic resource scaling, showing obvious advantages. The cost-benefit analysis reveals the significant cost savings of cloud deployment compared to traditional deployment, further confirming the affordability and feasibility of the cloud computing solution. In the end, the high satisfaction feedback from users and internal teams, especially for the customer management and data analysis modules, proved the success of the system design, which not only enhanced the quality of external services, but also greatly contributed to the modernization and efficiency of the enterprise’s internal management. Looking ahead, companies need to build a forward-looking roadmap for expansion. This includes integrating blockchain to enhance supply chain transparency and adopting AI to optimize CRM. Realizing this vision requires a flexible microservices architecture, dedicated R & D teams, and continuous team training to ensure that technological innovation and business development are closely linked, demonstrating the innovation and growth potential of the company to all stakeholders.