Industry is one of the main sectors that generate carbon emissions. Relevant data show that in 2022, China's carbon emissions will total 11 billion tons, of which industrial emissions will be 4.2 billion tons, accounting for 38.18% of the country's emissions. The "Action Plan for Carbon Peaking Before 2030" issued by the State Council listed the "Industrial Carbon Peaking Action" as one of the "Ten Actions for Carbon Peaking", requiring the industrial sector to accelerate green and low-carbon transformation and high-quality development, and strive to Take the lead in achieving carbon peaking.
How can industrial parks achieve low-carbon development? Refined management of energy consumption is the key. However, traditional industrial parks often lack a complete energy consumption monitoring and management system, as well as a professional "double carbon" knowledge reserve - "I don't know how much energy consumption is generated by which link", I don't know which energy consumption is not rigidly needed", "I don't know how to save energy How to formulate carbon reduction strategies", "I don't know how to implement carbon reduction measures", "I don't know whether specific carbon reduction measures are effective" and other issues have greatly restricted the achievement of the "double carbon" goal in the industrial field.
Shanghai Nanyang Wanbang, affiliated to Shanghai Yidian Group, is a solution company based on cloud computing, big data, and intelligent IoT, serving government and enterprises to realize the integration of data construction and transportation. Based on the insight into the rigid demand for energy conservation and carbon reduction in the industrial field, Nanyang IMCC created a one-stop energy conservation and carbon reduction solution based on the flying paddle and Wenxin large model, providing industrial parks with full-link services from energy consumption management to dual-carbon consultation, using artificial intelligence Technology empowers traditional industrial parks to achieve the goal of "double carbon".
Abnormal energy consumption is difficult to find, and peak energy consumption is difficult to predict
Two problems restrict the optimization of energy consumption in industrial parks
"Effective ways to achieve carbon emission reduction include energy consumption optimization, use of clean energy, comprehensive transformation, and forest carbon sequestration." Zeng Youxuan, manager of Nanyang Wanbang Intelligent IoT Solution Department, introduced, "Combining Nanyang Wanbang's related technical advantages, we choose to use 'energy' Energy consumption optimization' is the core entry point. Among them, the discovery and prediction of abnormal energy consumption and peak energy consumption are very important."
Abnormal energy consumption and energy consumption peaks often lead to an increase in carbon emissions, and are also important clues to reflect energy management problems in enterprises. Managers must clearly know in which scenarios and periods of time the enterprise has abnormal energy consumption or peak energy consumption, so as to implement targeted energy-saving and carbon-reduction measures. However, this is precisely the core pain point of many enterprises and parks.
Energy consumption data must be available before analysis and prediction can be carried out. The first difficulty comes from the collection of energy consumption data. The traditional method of energy consumption data collection is mainly based on manual copying of energy consumption meters (such as instrument panels, water meters, etc.), which is time-consuming and labor-intensive, and data accuracy and immediacy are difficult to guarantee, which greatly restricts the timely detection of abnormal energy consumption . The energy consumption prediction is also manually estimated based on the comparison of the data of the same period in the past. For example, if the electricity consumption was a certain value last Wednesday, it is estimated that this Wednesday will also be a certain value. The lack of a more accurate prediction system brings a negative impact on the energy consumption optimization of enterprises. Come challenge.
traditional instrument panel
In the eyes of Nanyang Wanbang, these two problems are just very suitable to be solved by artificial intelligence technology.
Use AI to draw "energy consumption map" and "energy consumption electrocardiogram" to capture abnormal energy consumption and predict peak energy consumption
Carbon reduction, safety and cost reduction
Using artificial intelligence technology to improve the efficiency and accuracy of energy consumption data collection, discover abnormal energy consumption in a timely manner, and realize scientific peak energy consumption prediction are the goals of the first stage of Nanyang Wanbang's technical solutions. After research and comparison, and comprehensive consideration of technical capabilities, eco-friendliness, and the localization needs of Xinchuang's customers, the team finally chose to conduct research and development based on the Paddle text recognition development kit PaddleOCR and the timing modeling algorithm library PaddleTS.
PaddleOCR has a wealth of components and a series of basic pre-training models such as open text detection and text recognition, which greatly reduces the difficulty of use. For developers, they only need to have a certain python development ability to test and tune on the basis of the open source model, and finally complete the deployment and launch in a short time.
PaddleTS provides a series of advanced timing modeling algorithms and related components based on deep learning technology, covering various scenarios such as timing prediction, timing anomaly detection, timing classification, etc., which can realize more accurate identification and detection of devices. The full name of time series is time series, which is a data form related to time changes (such as electrocardiogram, stock index, etc.), and is used to present information change trends. At present, PaddleTS performs well not only in the richness of timing modeling functions, but also in the number of integrated timing algorithms and end-to-end execution efficiency.
Nanyang Wanbang's overall solution steps are as follows:
- Collect energy consumption data in industrial parks through IoT devices. Devices with data transmission interfaces collect data through access gateways, and devices without interfaces, such as traditional instrument panels, take photos by installing cameras;
- Use the tool library provided by Baidu Flying Paddle text recognition development kit PaddleOCR to train the recognition model for traditional energy consumption meters in industrial parks, and deploy it on the edge computing gateway self-developed by Yidian Research Institute to realize rapid recognition of text in photos and convert them into data. Model accuracy reaches 99.5%;
- Data cleaning: data deduplication, blank value filling and splicing, outlier screening;
- Based on the complete and real-time energy consumption data, using PaddleTS, the timing modeling algorithm library of flying paddles, based on the advanced anomaly detection algorithm and timing prediction algorithm, build an abnormal energy consumption detection and timing prediction model in industrial parks, quickly identify abnormal energy consumption, and predict energy consumption peak. The model prediction accuracy reached 95%.
Model recognition dashboard data Finally, the project team established an intuitive "energy consumption map" and "energy consumption electrocardiogram", which not only allows managers to control energy consumption in real time, but also detects abnormal energy consumption and early warning capabilities based on current data. Consumption peak, so as to help enterprise managers plan ahead and adjust energy-saving strategies and measures in a timely manner.
"Energy consumption map" shows the overall energy consumption of the park
For example, when there is a hot weather and the temperature of the factory rises with the outdoor temperature, the air conditioner will automatically increase energy consumption to cool down, which not only increases carbon emissions, but sometimes cannot guarantee the economical operation of the air conditioning system. Enterprise managers can use the "energy consumption map" and "energy consumption electrocardiogram" to detect abnormal energy consumption of air conditioners in a certain area in time, and notice the current operation strategy of the air conditioning system in the area, so as to make timely adjustments to achieve energy saving effects. When the system predicts that there will be a peak energy consumption, it will also issue an early warning in time to help managers formulate strategies to reduce peak loads in advance and help the power system meet the peak summer.
"Energy consumption electrocardiogram" displays energy consumption trends, warns of abnormalities and predicts peaks
"Abnormal energy consumption and peak energy consumption prediction can not only help industrial parks achieve energy conservation and carbon reduction, but also enable enterprises to better achieve safe production and reduce energy consumption costs." Zeng Youxuan added, "Abnormal energy consumption can sometimes reflect equipment failures and Potential safety hazards, such as abnormal power data are sometimes caused by production line failures, or a precursor to fires, which are directly related to the safety of energy use by enterprises. Peak energy consumption will cause enterprises to generate more electricity bills under the current electricity price system, affecting Reduce costs and increase efficiency. "Shaving peaks and filling valleys" is not only related to carbon emission reduction, but also related to the financial cost of enterprises." At present, the service has covered 10 large industrial parks in Shanghai, Hangzhou, Qingdao, and Nanjing, and can help The park saves an average of 10% of carbon emissions and corresponding energy consumption costs, while ensuring production safety. According to the statistics of a park, it can save about 100 million kilowatt-hours of electricity throughout the year and reduce the annual electricity bill by more than 100 million yuan. The original 1-2 fires per year no longer occur.
Based on the large-scale construction of "Double Carbon Consulting Intelligence Brain",
professional knowledge is at your fingertips
It is not enough to have "energy consumption map" and "energy consumption electrocardiogram" to help enterprises carry out refined energy consumption management. The path to achieve the goal of "double carbon" is also the rigid need of many business managers to guide their own production improvement in a targeted manner.
Based on the Wenxin large model, Nanyang Wanbang has created the "Double Carbon Consulting Brain", which supplements the "Double Carbon" related policy guidelines, document reports and other content to the large language model. With its super learning ability and natural language understanding ability And language generation ability, so that enterprises can enjoy professional and convenient "double carbon" consulting services anytime and anywhere. Enterprise personnel only need to ask questions through natural language, and they can get a reply from the "intelligent brain", so as to efficiently obtain the "double carbon" related knowledge that could only be obtained by consulting a large amount of information or consulting professionals in the past, saving time and money. Labor costs. At present, the service has covered all the "double carbon" service industrial parks of Nanyang Wanbang.
Questions and Answers on "Double Carbon"
At the same time, Nanyang Wanbang also uses Wenxin's large-scale model capabilities to create "AI knowledge base" and "AI personal assistant" for enterprises, helping enterprises to upgrade their office intelligence and improve business efficiency. The enterprise knowledge base transformed by the large language model can provide various services such as document summary, internal process query, report interpretation and analysis, report summary, business insight, etc. To gain more convenience in terms of other aspects, and better invest in daily work and the realization of energy saving and carbon reduction goals.
Realized with flying paddle + Wenxin large model
"One-stop energy saving and carbon reduction solution" full-link service
Continue to promote the achievement of industrial "double carbon" goals
Flying Paddle + Wenxin Large Model enables Nanyang Wanbang's one-stop energy-saving and carbon-reduction solution to realize the full-link service from energy consumption management to double-carbon consultation. Based on this, combined with the joint assistance of Nanyang Wanbang's third-party partners, the overall one-stop solution covers five major links: carbon perception, carbon calculation, carbon emission reduction, carbon management and carbon certification, helping enterprises achieve energy saving, carbon reduction and cost reduction The ultimate goal of increasing efficiency and obtaining carbon certification.
Nanyang Wanbang's one-stop energy-saving and carbon-reduction plan has five major links
Talking about the future of artificial intelligence technology in the energy-saving and carbon-reduction scenarios of the manufacturing industry, Zeng Youxuan said: "With the proposal of the 'dual carbon' goal, the intelligent transformation of enterprises is also more urgent. Energy-saving and carbon-reduction are difficult but correct things, but at present In industrial scenarios, many small models are still being used to optimize specific scenarios. With the advent of the era of large models, how to use the intelligence of large models to better solve more problems of energy saving and carbon reduction in industrial manufacturing is a key issue. Nanyang Wanbang continues to explore the direction."
At present, Nanyang Wanbang has formed an overall business layout integrating "cloud-digital-intelligence-security". In addition to providing energy-saving and carbon-reduction services based on intelligent IoT for industrial parks, it is also continuing to cultivate in the fields of smart government affairs and information security. . The Flying Paddle and Wenxin large-scale models will also continue to improve, and join hands with partners from thousands of industries to jointly explore the future of artificial intelligence empowering thousands of industries. While helping enterprises achieve intelligent upgrading, it also creates value for society.