Abstract
With the accelerating pace of China's electricity market reform, the construction of the electricity spot market has been put on the agenda. However, as the number and scope of market participants gradually expand, the market-oriented transaction power continues to rise, and the intraprovincial and interprovincial transaction varieties are increasingly abundant. How to design an intelligent and powerful system that can meet the performance requirements of high concurrency and high-frequency transactions in the future market is a major problem in power reform. Based on the research of theoretical research results, this paper builds the front-end interaction platform of the southern spot electricity system based on the regional center to provide a data declaration interface for market users, including market management, market declaration, market release, market evaluation, intelligent analysis, front-end data interface, security protection, and other functional modules, provide declaration information and some market evaluation results to the southern regional spot power system platform, and obtain clearance results and published information from this platform.
1. Introduction
Power market refers to a combination of electricity commodity, transaction period, and way of transaction and a market for electricity trade. The market trending is affected by energy policy, power net structure, demand and supply, market participants' management decisions, etc. [1]. In 2017, a notice on the implementation of the pilot construction of the electric power spot market [2] was published by the general office of the national development and reform commission and the National Energy Administration’s general office, which chooses the southern region (starting from Guangdong), western Inner Mongolia, Zhejiang, Shanxi, Shandong, Fujian, Sichuan, Gansu, etc. and counts up to eight regions as the first pilot of the spot market, making sure that the mechanism of power fare should be formed. Based on the goal of optimization of the spot market, we establish the clearing mechanism and blocking management mechanism of the spot market under security constraints. So far, three of the first eight electric power spot market pilot projects have been launched. 2019 is the first year of China's electricity spot market, based on the pilot operation of the spot market in south China (starting from Guangdong) and Gansu and Shanxi in 2018, and the remaining five of the first eight pilot areas will start trial operation in mid-2019 as required. The pilot operation of the electric power spot market will elevate the reform of the electric power system centering on the construction of the electric power market to a new level, further activate the electric power market, and put forward higher requirements on the power trading ability and market risk control ability of power generation enterprises and power selling companies. On the other hand, the establishment of a fully functional and smoothly connected technical support system is also the support and important technical guarantee of spot trading pilot implementation.
Domestic scholars have carried out a series of studies on domestic and foreign electricity markets in recent years. Savelli et al. [3–5] studied and analyzed the actual cases of the construction of foreign electric power spot market; Barazza and Strachan [6, 7] discussed the construction path of the power market from the perspectives of the construction mode of the spot market in the southern region and the mechanism design of the power market in Yunnan; Santos et al. [8, 9], respectively, designed the electric power spot market mechanism from the perspective of real-time price elasticity of user demand and marketization operation of incremental distribution network; López-Salamanca et al. [10], based on the similar network topology of blockchain and electric power spot market system, designed a brand new electric power spot real-time scheduling transaction and regulatory model for the intelligence and real-time and security requirements of electric power spot market transactions. The above literature provides valuable references for the construction of the spot market from multiple perspectives but lacks analysis on the construction of the complete technical support system of the spot market of electric power. As a matter of fact, the electricity market is a market with very strong technical characteristics, and economic problems are closely related to technical problems. With the growth of the number of users, the business volume of marketized electricity transactions increases at a geometric level. The existing technical support system is difficult to support the performance demand of high concurrency and high-frequency transactions in the future market. However, each upgrade is not a local software and hardware repair, but a change in the system architecture level. Thus, it is necessary to design an electric power spot system architecture with high connectivity, security, and scalability.
On the other hand, China’s power market has gradually entered a new stage oriented by market demand, and the classification of power users is one of the necessary means to realize the optimal allocation of power resources [11]. The development of the electric power spot market is based on the premise of meeting the diversified demand of the market's main body, possessing the remarkable feature of bidirectional selection. In such an environment, the new generation of power trading platforms needs to adopt price measures and incentive policies based on user clustering and user portrait, support business decisions with data, and ultimately promote the smooth and efficient operation of the power system.
Aiming at the fact that the characteristics of domestic electricity spot market business are complicated, we put forward the southern regional power spot system front-end interactive platform construction principle and partition of hierarchical application deployment business framework, through the “loose coupling” business architecture that can be extended according to the actual needs of users of the many new features, thus providing more modular, humanization, and intelligent service.
At the same time, the system is developed based on the high performance and high availability middleware and the technology platform oriented to the Internet distributed system, which increases the security isolation and reliability isolation between the systems and provides a reference for the construction of the domestic electric power spot market technical support system.
It is not easy to open the spot electricity market. There are many problems that need to be solved. How do we find the price of electricity? What data are used as the basis for the price discovery process? What to do with all this data? How do we design corresponding rules? As an electricity selling company, after determining the spot price, how do we sign the medium- and long-term contract? In the wholesale market, how do we sell electricity and customers to sign a contract? How do we sign a contract with a power company? In the spot process, how do we distribute the quantity, how do we decompose, and so on?
Naturally, all models in the spot power market can be abstracted into a huge algorithmic model, and many methods in artificial intelligence have natural advantages for processing big data. By integrating computing and data into the electronic market, the power spot system based on artificial intelligence will be more competitive.
The data involved in the power market are diverse, including multidimensional data such as climate, environment, and fuel price; large sample data of grid side and whole network; cluster data of users on the selling side; user side depth behavior data; and a large amount of data generated by high-frequency trading in the spot scenario. At the same time, China's electricity market data is slightly different from other data in the world. China's data mainly involves some security issues, some of which can be made public (climate, environment, and other data), and some of which can be made semitransparent (grid-side data). Therefore, more attention needs to be paid to data security when using cloud services in the power market. For the electricity market, we want to use certain technical means, in the case of semitransparent calculation results.
In the wholesale market and big data environment, artificial intelligence has a profound future in the power market. In the spot power market, price forecasting is the key. In order to achieve this goal, it is necessary to start with simulation modeling, fundamental analysis, load forecasting, game behavior, strategy optimization, and other aspects. For different types of benchmarks, different treatment conditions are adopted to maximize the control cost. At the same time, for the market and declaration management subsystem, market evaluation subsystem, portal website, intelligent customer service four major business modules, the data analysis, and prediction can better clear the communication channel between the modules.
To solve the above problems, the paper proposes a new architecture based on an Intelligent Technical Support System for the Electricity Spot Market: (1) analysis of the construction of a complete power spot market technical support system; (2) user behavior analysis, intelligent customer service, and other key technologies; and (3) providing a reference for the construction of technical support system in the domestic power spot market.
2. Construction Principles
2.1. Design Principles
The system construction strictly follows the following design principles:(1)Openness. The platform can be widely interfaced with a variety of external systems, which can be interface specifications for the platform or use existing interface standards on the market. External systems include power grid dispatch control systems, power grid dispatch management systems, and mid-long-term trading systems.(2)Expandable. The system adopts a modular design. When the business development needs to expand new business and new functional modules are needed, the original system architecture is not affected, and the risk of major modification of existing systems is reduced.(3)Configurable. Different business models adopt the parametric design, which can be flexibly configured according to business needs.(4)High Availability. The system uses clustering, fusing, backup, and other technical means to enhance the system's availability. When the system is abnormal, it has a self-recovery ability to prevent downtime.(5)High Security. The system has a complete security mechanism, including a rights management system, a resource management system, and an operation management system to ensure that authorized business users and system users can use related functions, access related data, and implement operational function settings to effectively avoid operational risks. At the same time, the system has the function of data encryption protection, adopting the digital signature as the core digital signature, electronic seal, timestamp, and other means to encrypt the core data of the system business in the process of storage and transmission to prevent the leakage of transaction data. In addition, the system has a network attack prevention function on the Internet exportation and protects against network attacks such as Distributed Denial of Business (DDOS) attacks through various means such as flow cleaning to ensure that normal business is not affected by network attacks.(6)Manageable. The system adopts the ACM platform to provide monitoring, management, control, and other functions for the business in operation. When abnormal transaction processing is found, it can provide multiple alarm paths and has a humanized monitoring and management interface.
According to the complex and changeable business characteristics of the domestic power spot market, this paper proposes a power architecture based on high connectivity, security, and scalability.
2.2. Technical Indexes
The front-end interactive platform needs to meet the following performance index, including the following:(1)System availability is ≥ 99.99%. The system front-end interaction module can use the time/total time ≥99.99% for the core business in the statistical period (monthly).(2)It should support concurrent online users of no less than 5,000 core business (e.g., price declaration) concurrent requests for users not less than 500.(3)The message processing capability is not less than 6000 strips/second. Specifically, the number of messages processed per second by the system is not less than 6000 strips/second.(4)It should support the permanent preservation of all historical data during the system life cycle.
2.3. Security Indexes
The front-end interactive platform needs to fully consider the requirements of security protection, including the following:(1)The system architecture should consider the isolation of the network. In safety zone III and the demilitarized zone (DMZ), the internal and external network interaction platform should be used for isolation. The forward and reverse isolation devices should be used between safety zone III and safety zone II.(2)Network security access must meet the security protection strategy of system access. The system should support security encryption of network communication, encrypted communication at the application layer, identity authentication, access control, and other measures.(3)The system should have the identity authentication function, and the identity of the user should be verified by a digital certificate to establish a unified authentication system to prevent illegal access.(4)The system should provide monitoring and security auditing interfaces, monitor the status of servers, workstations, and network devices in conjunction with the security situational perception system to achieve audits during and after the event, and trace responsibility for malicious actions and violation operations.(5)The system should have the function of data encryption protection. It adopts a digital signature, electronic seal, and time stamp with a digital certificate as the core to encrypt the core data of the system business during storage and transmission to prevent the leakage of transaction data.(6)The system should have the network attack prevention function in the Internet exportation and protect against network attacks such as DDOS attacks through various means such as flow cleaning to ensure that normal business is not affected by network attacks.
3. System Architecture Design
Based on the design principles and construction principles proposed in the previous section, we develop a technology platform based on high performance, high availability middleware for Internet distributed systems.
3.1. Business Architecture
In order to prevent the system from self-recovery and downtime in case of abnormality, the whole architecture is divided into four subsystems due to the principle of high availability middleware. The system consists of four business subsystems, namely, the market and declaration management subsystem, the market evaluation subsystem, the portal website, and the intelligent customer business, as shown in Figure 1.(1)Market and Declaration Management Subsystem. It provides various businesses related to power trading required by the electricity market operation rules and provides technical support for constraint disclosure, price declaration, market release, and information release in the power market. As the core module of the whole system, after the module is implemented, it should be able to realize all kinds of transactions in the day-ahead, within-day, and real-time market according to the division of the trading cycle, to meet the development requirements of the power market; improve technical support for data reporting, market release, market analysis, and information release in the power market; provide market participants with convenient data declaring means, verify the validity of the declared data, and ensure the certainty and integrity of the data declaration; analyze statistics of market information, analysis, evaluation, and prediction of market operation; publish market information based on market planning; and ensure the validity, correctness, completeness, and security of the information.(2)Market Evaluation Subsystem. In the electricity market environment, different types of market participants want to maximize economic returns in the market, and they also face different risk assessment and management issues. In order to control risks and ensure smooth operation of the regional power market, it is urgent to quantitatively study the market risks and price trends. At the same time, from the data point of view, combined with statistical analysis methods, we collect, process, and analyze a large number of market member bidding behavior information, tracking, analyzing, real-time forecasting, and early warning the characteristics of power supply and demand of various industries, regions, and people, and providing user classification and personalized services, which will help improve the level of meticulous operation management and demand-side management. The submodule provides quantitative analysis tools for power market operations and supervisors to analyze financial market financial risks, forecast electricity prices, and monitor electricity market bidding behaviors, including data modeling, user behavior analysis, user portraits, data statistics, risk monitoring, and other functions.(3)Market Member Management Subsystem. For all kinds of market members, such as power grid enterprises, power generation enterprises, power sales enterprises, power users, and independent ancillary service providers, the functions of registration, alteration, exit, maintenance, and export of market management information are provided. At the same time, it provides unit and power generation matching and constraint condition parameter management functions, manages various configuration parameters of the spot market, and sets the basic running environment of the spot market. The parameter categories include but are not limited to basic parameters, security correction parameters, optimization calculation parameters, unit constraints, system balance constraints, network constraints, and other constraints set parameters.(4)Intelligent Customer Service. It provides fast and efficient customer service for end-users of the spot system, effectively saves the human resource requirements of customer service, and provides unified customer service for different types of platforms through rich interfaces, specifically including Automatic Speech Recognition (ASR) service for user voice-to-text and Text to Speech (TTS) service for answer text-to-speech, through large-scale knowledge processing, natural semantic understanding, and other technologies. According to the application scenarios of different businesses at the same time, layering and grouping the structure of the knowledge base to realize a multipurpose “people” to improve work efficiency and completely automatic learning of unknown problems or unknown questions based on precise algorithms make the knowledge base constantly updated and optimized and reduces maintenance costs, use the crawler system to collect unstructured data, dock data centers or other data source information, and generate the basic market, information, and analytical data of knowledge base through the processing of the data.
3.2. Technical Architecture
The basic idea behind microservices is to think about creating applications around business domain components that can be independently developed, managed, and accelerated. An advocate will apply this model to a series of small services; each service focuses on a single business function, running in separate processes, which make the boundary between the services is clear, and can use lightweight communication mechanism (such as HTTP/REST) to communicate with each other, cooperate to achieve a complete application, and meet the needs of business and user [12]. Therefore, this system uses the microservice environment provided by cloud computing to build the development, deployment, and operation environment of the software. Meanwhile, according to different responsibilities, the whole system is divided into a terminal display, access network, background service, operation and maintenance monitoring, platform, and middleware layer, so as to meet the requirements of high availability of the business system. The overall technical architecture is shown in Figure 2.
3.2.1. Terminal Presentation
The terminal is mainly divided into transaction clients, management clients, and third-party clients. The front gateway of the front-end interaction module of the spot trading system is accessed through HTTP/HTTPS protocol. The request/reply mode is the main message communication mode to realize the interaction with the server system and complete the business operation and system maintenance functions.
3.2.2. Access Gateway
The main task is to convert external HTTP/HTTPS protocol requests into internal microservice calls and implement access control. It mainly provides session management, channel management, and other functions, through service discovery and service routing to locate the request service address and complete the service call. In addition, the access gateway also has load balancing, flow control, and other functions to protect the back-end service and prevent it from being overwhelmed by the large flow.
3.2.3. Back-Office Services
It mainly provides various services and components on which the business functions of the front-end interactive platform of the spot trading system depend, including application services and basic services, to simplify the business development process and improve the reuse of applications.
3.2.4. Operational Monitoring
It mainly realizes the unified monitoring and management of all kinds of resources in the application system and provides up and down service, service degradation, and real-time warning. It can be convenient for troubleshooting and solving, for each application system provides the most convenient monitoring and management mode.
3.2.5. Platform and Middleware Layer
It mainly provides the basic services needed by the application in the development, testing, and running process.
3.3. System Deployment Plan
Figure 3 shows the network topology of the front-end interactive platform of the power spot system. DMZ is a buffer between nonsecure systems and a secure system to solve the problem that access users of the external network cannot access the internal network server after installing a firewall [13]. The system deploys application services in the DMZ area and the secure III area, respectively. Meanwhile, an internal and external network interactive platform is built between the secure III area and the DMZ area to realize data transmission between the internal and external networks, as shown in Figure 4. For security III zone deployment firewall to strengthen the restriction on the security of the internal network, and through the security route to force the security settings of the network equipment inspection, we shield the network connection to nonsecurity equipment, to ensure the security of the internal network. Among them, the application deployed in the DMZ is targeted at market members and accessed through Internet exit, mainly providing various member interaction functions such as market information release, price declaration, and member information management. Safety III area is the production of administrative zones, mainly for market operation and maintenance personnel, and provides all kinds of process management, market management, application of price management, market information release management operations, etc.
3.4. Interface Docking
As the subsystem needs to be connected with the power spot system of the southern region, we interact through E file format and coordinate the interaction process by using the data exchange gateway to realize information sharing and integration between each business subsystem and the external business system, as shown in Figure 4. The southern regional power platform transmits data between each subsystem through the Aliyun cloud object storage service (OSS of Aliyun) and exchanges data with each gateway through the Aliyun cloud message service (MNS of Aliyun). In order to prevent the leakage of sensitive information during the transmission of E files, the system encrypts the contents of the files and adopts an asymmetric encryption algorithm. Among them, the encryption key is provided by the file receiver. The file sender encrypts the content of the file by using the encryption key, and the receiver decrypts the content by using the decryption key and then carries out subsequent business processing. During this period, the protocol transformation gateway will always monitor the call record and call status of the interface and alert the channel exception, data exception, and illegal access through the ACM platform.
4. The Core Modules
4.1. Market Declaration Module
The market declaration module is mainly used for various market members to declare data and verify and process the declaration data received. It includes the following functions.
It supports price declaration and provides grid constraint information to users for reference on the declaration page. Constraint information includes declaration period, declaration upper and lower limit price, equipment maintenance plan information, tie line plan information, grid safe operation index, and grid safety constraint information.
It supports the daily market declaration, including the declaration of electricity consumption, unit electric energy, unit expected grid-connection time, unit maximum generating capacity, unit heat supply flow, and unit emergency minimum output. At the same time, it supports the declaration of unit electric energy in the real-time market and the declaration of unit frequency modulation and peak regulation in the frequency modulation market.
It supports the monitoring and approval of market declaration information and informs the applicant via message reminder after approval.
It supports input and verification of declaration data and import and export of declaration data according to declaration data category.
Payment duplicate declaration, power users, and power selling companies can copy the data of historical declaration records through this function.
4.2. Market Release Module
According to the information disclosure principle in market rules, the market release module is mainly responsible for releasing current and future grid operation, market operation, market supervision, and other pieces of information to various market members, government authorities, and power regulatory authorities.
It releases maintenance plan information, such as maintenance shutdown and return of key equipment.
It releases plan information of the liaison line.
It releases network operation information, including the actual load of the system, the actual load of the bus, the actual output of the unit, the exchange power of the tie line, the load of important sections, and the safe operation index information of the network.
It releases grid security constraint information, including stable section quota and equipment operation quota.
It releases market operation information, including min long-term contract volume price data and transaction records, day-to-day market node electricity price, real-time market node electricity price, daily market safety check result, real-time market safety check result, day-to-market market node partition information, real-time market node partition information, crew frequency mileage information, real-time clearing results of the FM market, and real-time clearing results of the peaking market.
5. Key Technologies and Implementation
User behavior analysis systems and intelligent customer service are the focus and key of this paper. According to the regularity of user behavior and intelligent customer service, the artificial intelligence (AI) method is used to study and analyze it. AI is the technological science that studies and develops theorem, algorithms, technology, and application to imitate, inherit, succeed, expand, and even surpass the intelligence of human beings.
AI intends to be aware of the primary essence of intelligence of humans and create a new category of human beings to some extent. Research areas in artificial intelligence include computer vision, natural language processing, machine learning, expert systems, recommendation systems, and fuzzy logic. In addition, its research field is still expanding.
As the most significant method to implement and accomplish AI, machine learning is meanwhile the mainstream orientation for research of AI. Machine learning uses experience to improve the performance of the system itself. Machine learning can be classified into two parts: traditional methods and advanced methods. The former consists of supervised learning and unsupervised learning. And the latter is comprised of deep learning, transfer learning, and reinforcement learning.
One application of machine learning, as an example, is to explain and predict the phenomenon of the rising and falling of the spot market prices, which attracted enough interest from many researchers and other relevant institutions. A lot of previous research works have been done, where the researchers used various techniques of regression to solve the question of the changing market prices.
One of the sections of this paper proposes the case of changing prices as a classifying question and utilizes machine learning technology to predict the rising or falling market price [14]. This work uses varieties of feature engineering technologies such as PCA (principal component analysis). Another critical technology applied is the data transformation techniques, which process missing data, outlier data, and change the distribution of data such as box-cox transformation technology. The performance of one machine learning method is usually measured by the following 4 metrics: accuracy, precision, recall (sensitivity), and specificity. In this work, the two values 0 and 1 represent the two classes, respectively, where the class value 0 represents that the price of the market decreases. And the class value 1 represents that the price of the market increases to the contrary. Generally, a lost function can be defined aswhere the linear relationship between h and x is given as
Then, gradient descent should be used to update all the parameters as formulas (3) and (4):
Most of the time, classification task needs to be done in the spot market. Generally, we use logistic regression to handle it. It clearly knows that object function should be firstly considered, which is defined as
Similarly, in order to minimize this cost function, gradient descent should be applied to train this model as formulas (6) and (7):
It is easy to know that the value of is 0 or 1, which is totally different from the linear regression.
Different from traditional low-depth learning, deep learning increases the number of layers and usually changes the structure of the model. In addition, it emphasizes the learning of features. Unlike traditional hand-designed feature extractors based on professional domain knowledge, deep learning extracts the feature from the input layer to the output layer and builds the mapping from the latent information to the semantics at a high level to obtain the characteristic expression of the data from the general learning process. Typical deep learning models include deep belief nets (DBNs), convolution neural networks (CNNs) [15], generative adversarial networks (GANs), recurrent neural networks (RNNs) [16], and stacked autoencoder (SAE) [17]. For the full connected network, the object function (without regularization) should be shown as
Clearly, the gradient descent is also needed to minimize. The gradient for the sigmoid function can be computed as
Then, in order to optimize all the parameters, the backpropagation algorithm is needed to be implemented.
Reinforcement learning [18], also known as evaluating learning and reinvigorating learning, is an effective technique for machine learning. Its foundation is to solve the problem of decision-making. In other words, it is a method to learn how to make correct decisions by itself. Reinforcement learning mainly consists of the four factors: individuality, situational state, behavior, and award. In the learning procedure, the individual learns on the basis of the situation, and the search tactics make the best choice, leading to the change of the state, so as to obtain the reward or penalty of feedback from the system. According to these reward and punishment values, the current strategy is adjusted and a new epoch of learning is processing, and the loop is repeated until the system evaluates the individual under certain conditions. There are several typical methods for reinforcement learning such as the Sarsa algorithm and Deep Q Network (DQN) [19] and Q-Learning method.
For DQN, in addition to using a DCNN (deep convolutional neural network) to approach the present value function, another independent network is implemented to yield the goal Q value [20]. Explicitly, indicates the value of the output layer in the network, . The output of the network represents the target, and generally, approximates the goal Q value. It updates the parameters of the network by minimizing a loss function, that is, the difference between the actual value of Q and the goal Q. The loss function of mean square error is indicated as
We obtain the gradient by partially derivative of the variable, as shown in
The purpose of migration learning [21] is to use the relevance of learning objectives and existing knowledge to apply existing knowledge to relevant but different areas to deal with the correlated questions. In many cases, a few tag samples in some application scenarios are even difficult to acquire samples and can not help construct the robust models. Migration learning can transfer the model parameters in related scenarios to another scenario to build the structure of models, by which the new model’s adaptive ability is improved. Quintessential migration learning methods include Tr Ada Boost (transfer adaptive boosting), Co CC (coclustering based classification), and self-learning.
Migration learning [22] refers to a method of migrating knowledge from a guidance model T to a learning model S. Therefore, the mean square error loss function can be used to train the parameters of the learning model S as shown in
Another way to migrate the Q-value function is to migrate only the action corresponding to the maximum Q value from T to S. Then, we use Negative Log-Likelihood (NLL) to predict the same optimal action value to train the relevant parameters of the learning model S as shown in
And in order to make the result more precise, Hinton [23] used Kullback-Leibler Divergence (KLD)to define the loss function as shown in
6. Conclusions
This paper studies the front-end interactive platform scheme of the power spot system in the southern region, puts forward the construction principle and architecture design of the system, and introduces the two core modules of market declaration and market release in detail. The proposed electricity spot market front-end interactive platform has the following characteristics: (1) based on high performance, high availability middleware, Internet technology platform for distributed system-oriented development, increasing the security isolation between the system and reliability, and with “loose coupling” business architecture, effectively improves the stability and reliability of the system; (2) it timely and accurately grasps the characteristics of customers' electricity consumption behavior through the analysis of user behavior, supports the decision-making of power marketing and dispatching of enterprises, and improves the service level of government, industry, and commerce and other departments; (3) it uses intelligent customer service to provide customers with all-round professional consulting services in the field of the power market. According to the construction principles and architecture design of high availability, we analyze and discuss the user behavior and intelligent customer service and design the technical support system scheme for the southern regional power spot market because the design can meet the demand of high concurrency and high-frequency trading in the future market. Meanwhile, it also provides an important reference for the construction of a trading platform in the domestic electric power spot market.
Nowadays, some power companies aim to improve quality and efficiency and fully apply technologies such as cloud computing, big data, the Internet of Things, mobile networks, and artificial intelligence. As an important driving force of the new round of scientific and technological revolution and industrial transformation, artificial intelligence is profoundly changing people's production, life, and learning methods and promoting the intelligent era of human society, human-machine synergy, cross-border integration, and sharing. Therefore, the application of artificial intelligence-related technology can be better incorporated into the grid spot system research, which can improve the performance of the system. For example, how to do real-time load forecasting of the sales company's proxy users and then improve the performance of the corresponding system modules will involve a large amount of data accumulation and calculation, and these can be solved by artificial intelligence, big data, and other technologies. Therefore, I believe that artificial intelligence will bring more surprises in future research.
Data Availability
The data used to support the findings of this study are available from the corresponding author upon request.
Conflicts of Interest
The authors declare that they have no conflicts of interest.