The introduction of the electricity market will inevitably bring significant changes in the function and connotation of the energy management system (EMS). Load forecasting (especially short-term and ultra-short-term load forecasting) is the basic function of EMS, and is the premise and guarantee for the safe and economic operation of the power system. In the market environment, the basis of the scheduling plan and the power supply plan trading plan is different from the monopoly operation mode. The market economy operation mode with the main purpose of introducing competition puts forward new standards and requirements for power system load forecasting. These requirements include accuracy and real-time performance. Reliability, reliability, etc. They are interconnected and mutually constrained. 1 load forecasting system architecture Under the electricity market operation mode, a set of accurate, real-time, reliable and intelligent short-term load forecasting systems requires the following links: data acquisition, Data analysis, load forecasting, planning, planning reporting, information monitoring, and planning assessment. Among them, the forecasting link includes short-term, extended short-term, and ultra-short-term load forecasting. The three forecasting systems are based on domestic and foreign practices; the operating conditions of the planned day are also monitored, and the remaining time of the day is completed in time when the original plan and the actual load are seriously deviated. Re-predicting and planning adjustment of the load, so as to restore the economic loss as much as possible. The adjustment of the plan generally requires a 2h margin. That is, the power sector can only adjust the load plan at the current time after 2 hours, to ensure that the dispatching department and the power plant have sufficient Time is ready to execute the new plan.
The function of extending short-term load forecasting is that the load-related factors on the current day deviate significantly from the original forecasted amount, resulting in the adoption of the latest acquired information when the current load has a large deviation (> 3%) from the original plan or may be likely to occur. , re-run the load forecast for the next 2h to several hours on the day, and propose new forecast results for the forecasters and determine whether the plan needs to be adjusted. Due to the latest collected information, the extended short-term load forecasting is more accurate than the short-term load forecasting result completed 1d earlier.
Load forecasting system planning and reporting system information monitoring system, plan assessment system, etc. 21 Information collection and management system Effective information management is the basis of load forecasting. Adequate and perfect information is the premise of forecasting, and information reliability is prediction. The basis of accuracy and reliability In order to complete the reliable prediction of quasi-sulfur, the prediction system needs to collect as much forecast information as possible. It also needs to analyze the collected data, correct the information error, and determine the value and utilization of each information. The system uses the data acquisition link and the data analysis link to complete the above functions. 21.1 Data collection link data collection link collection load and related information data, including: historical daily load information, current day live load information; historical day weather information, current day weather information date The latest forecast meteorology, the latest forecast meteorological data; historical daily electricity price information, current day electricity price information, electricity price forecast information; large user plan load information, small thermal power small hydropower generation plan information; other measurable information (such as maintenance power outage plan, load shedding plan Large event plan, etc.) 21.2 Data analysis link Data analysis link completes information data correction and regular information stripping, improves information accuracy and reliability as much as possible, and then uses these information data to calculate daily factor similarity and daily load similarity. The degree reflects the correlation between the daily load related information and the load-related information of the standard day given by the system; the daily load similarity reflects the correlation between the daily load data and the load data of the standard day given by the system. The ratio of the similarity of a 2d power system is the relative similarity of the 2d. It embodies the information integration of the 2d. 21.21 Data correction The automation level in China is not high, and the sampling equipment or transmission line fault is easy to cause false data, called bad data. In the market mode, various information is randomly added, and the random data amount is superimposed on the amount of natural change data to form malformed data, which is called bad data. If these data cannot be effectively corrected, they will be provided to the load forecasting method in the form of false information and false change laws. It is necessary to mislead the establishment of the load forecasting model and affect the accuracy and reliability of the forecasting.
The data correction link performs bad data and bad data identification and correction on the collected information. Among them, the bad data correction mainly refers to the bad data correction, and the bad data correction includes the natural data complement (such as load shedding, line repair and power failure, etc.) and impact data stripping (such as large users, large event shock load stripping, etc.) live load data information. It is the main information of load forecasting, which reflects the change of load-related information. Therefore, it is especially important to correct the data of live load data. The following describes the principle of load data correction for the characteristics of electric load. Other load-related information is based on the characteristics of electric load, and the load does not change suddenly before or after the load is also called the principle of load viscosity. Some obvious bad data points are repeated. These data points are repeatedly corrected by the three-point averaging method until the values ​​conform to the principle of load viscosity. From the statistical point of view, the load at the same time of each day is normally distributed, and the same continuous time period of each day. The load change rate is also normally distributed. The historical multi-day simultaneous load data is used as a sample for probability and statistical analysis. The expected value and variance estimate in the two normal distribution models above the time period are completed, and then the confidence is set to complete the load. The horizontal confidence interval estimate investigates whether the load value to be detected is within the confidence interval, and discriminates whether it is normal data or bad data or bad data. If it belongs to the latter, it increases the confidence and conducts a bilateral hypothesis test to determine that it is Bad data or bad data versus bad data, using the aforementioned method For the bad data, the maximum likelihood estimation is used to estimate the most likely load value of the point under the normal distribution. It is considered that the natural load value of the point is replaced by the natural load value, that is, the bad value is completed. The complement of the data will be recorded as the impact load before and after the complementation, that is, the peeling of the impact load is completed.
2.1.2.2 Similarity calculation The calculation of two similarities on the completion date of this link: daily factor similarity and daily load similarity. The factor similarity and load similarity at the same time on the same day are mutually verified. The verification information is used for data correction, that is, the similarity and load similarity of the same time should be consistent at the same time. If they are inconsistent, the information reflected by at least one similarity has bad data or bad data. This is the data correction link. The basis for the correction is provided.
On the basis of the similarity of the daily factors and the similarity of the load, the calculation of the relative similarity can be completed. It includes the calculation of the relative similarity of the daytime information and the relative similarity of the information measured during the day (measured at different times). The former reflects the difference in the power system. The similarity of the information contained in the date (including the load-related information and the load information itself) is the embodiment of the comprehensive “closeness†of the 2d information. It is mainly used as a forecast for the short-term load forecasting system; the latter reflects the collected at different times on the same day. The similarity of information is the embodiment of the comprehensive “closeness†of the two time information on that day. It is used as the monitoring basis for the information monitoring system. It is used to predict the short-term ultra-short-term load forecasting system as the forecast. 2.2 Load forecasting system The load forecasting system is the load forecasting. The core, including short-term ultra-short bottle expansion short-term load forecasting, they respectively perform different functions, and interact to form a complete system 2.21 short-term load forecasting (such as electricity price) repair 4m similar to ernieJournalElectronicPublish in the implementation method will be multiple loads The model is weighted. The merger will complete the future 1d~7d full-day load forecast 1d ahead of schedule, mainly used to arrange the unit start-stop plan, power balance, and safety check.
The weight coefficients of each model algorithm are determined by optimization method, taking full account of meteorological and electricity price information, and using the forecasted real-time load information to correct the prediction results, and introducing a new prediction strategy based on the concept of similarity, which effectively improves the prediction accuracy. In terms of algorithm intelligence, firstly, the least square method is used to perform the weight detection before weighting the above multiple load models, and some algorithms that are not applicable or excessively biased for the current prediction date are eliminated to avoid the algorithm weight optimization. Secondly, the system uses a comprehensive model such as pattern recognition and artificial neural network to match the changes of the system load model under different prediction dates; again, the intelligence of the system algorithm is also reflected in automatic information recognition and deviation identification. When the information deviation exceeds a certain limit, the new load forecasting algorithm is automatically selected to combine with the weight optimization, which effectively improves the adaptability of the algorithm. 22.2 Ultra-short-term load forecasting Under the guidance of the previous 1d plan, the real-time information of the day, Complete the load from 5min to 60min on the same day Forecasting, providing scheduling to complete automatic power generation control, real-time economic dispatch, online safety monitoring, and as a premise of real-time spot price forecasting in the electricity market. The results are also provided to the information monitoring system in terms of implementation methods, using linear extrapolation, time The weighted combination of sequence, Kalman filter, neural network and other methods effectively utilizes and reflects the influence of various known information on the load, and improves the accuracy of the prediction. 22.3 Expands the load forecast from 2 hours to several hours after the current time of the short-term load forecasting. The prediction results are used by the forecasters in order to decide whether to make adjustments. The principle information of the extended short-term load forecasting is similar to the short-term load forecasting, and the implementation is the same. The differences are shown in Table 1.
Table 1 Comparison of short-term load forecasting and extended short-term load forecast Table1Comparisonbetweentheshort-term comparison Project 1 forecasting model short-term load forecasting extended short-term load forecasting function description completed in the future 1d7 d full-day completion date 2h to load forecasting several h load forecasting is mainly historical fact information Historical fact information, current day information, original plan information sampling point fixed, 96 points are not fixed, less than 96 points optimization target full-day load 96 points mean square error minimum prediction period multi-point load mean square error minimum expansion short-term load forecasting mode The implementation method can further study whether the expansion of short-term load forecasting is controlled by the information monitoring system. When the information monitoring system monitors that the actual operating load will be significantly different from the original plan (>%), the 22.4 prediction system will be integrated. It is feasible, necessary and beneficial. Based on the same information database, they share some mathematical models and algorithms (such as linear extrapolation, exponential smoothing, neural network, etc.), and use similar prediction strategies to extract the information they need to complete the prediction. Their prediction results are mutually correct and corrected. The results of the same set of artificial interfaces are displayed on the same set of interfaces as the manual intervention. The forecasters can simultaneously observe the rich information changes. The short-term load forecast results are extended short-term load forecasting and super Short-term load forecasting information; ultra-short-term load forecasting results are one of the monitoring information of the information monitoring system that controls whether extended short-term load forecasting needs to be initiated; extended short-term load forecasting is a backup of short-term load forecasting, and corrects its forecasting result. 2.3 Information The monitoring system information monitoring system conducts on-line monitoring of the latest forecast information (weather forecast, electricity price forecast, large event forecast, etc.) of the current day's live information (load information, weather information, electricity price information, etc.), and adopts the data analysis link within the day (different time The information relative similarity data is used to calculate the degree of deviation of information in different time periods, which is called information deviation degree.
In particular, it is necessary to complete the relative similarity and deviation calculation of the information currently collected on the current day and the 1d pre-forecast information, so as to theoretically determine whether the current load and the original planned load will deviate greatly.
The system adopts the latest forecast results in the last 1h given by the ultra-short-term load forecasting, compares with the original plan, and calculates the predicted load deviation degree; collects real-time assessment data from the planned assessment system to calculate the actual load deviation degree.
When the above deviation exceeds the respective tolerance limits (such as weather changes, electrical circuit information changes, large circuit events, etc.), the system gives a plan deviation from the operational information warning through the information window, and starts the extended short-term load forecast to re-complete the future. 2h to several hours load forecast, for forecasters to decide whether to correct the plan 2.4 plan assessment system plan assessment system for the historical assessment of the historical and planned load data, including online assessment and offline assessment online assessment calculation of the day's live load data and the deviation of the day plan Provides the basis for planning deviation for ultra-short-term load forecasting and information monitoring systems; the latter calculates historical date load data and its planned deviations, provides information for short-term load forecasting, and is also the main source of information for electricity market settlement. Offline assessment can also be managed information system. (MIS) Completion 2.5 Planning and Reporting System From the perspective of automation, the load forecasting system can include the planning and reporting system. It completes the planning based on the load forecasting results, and then completes the planning report (mainly the trading planning department). ) and the planned release (mainly power generation plan, power supply plan part), these automatic functions are the embodiment of the software automation level. The above three sets of prediction systems interact to form a whole tblishingI. The short-term load forecasting system of the above framework has been in many domestic provinces. The ground adjustment is practical. The research and application results show that under the electricity market operation mode, the load forecasting operation mode has undergone major changes. The power system puts forward new requirements and standards for the connotation and evaluation of the load forecasting function, and proposes a new model demanding load for the load forecasting framework. As a complete system, the forecasting system consists of information collection and management system load forecasting system information monitoring system, planning and reporting system planning and evaluation system. The system load is social and random, and the load chaos and complexity are increased. The management system filters, corrects and distributes the short-term load forecasting of the load forecasting information. It must not only complete the full-day load forecast for the next 1d~7d, but also provide the re-forecasting tool for the future time of the day. Introduce the concept of extended short-term load forecasting, and complete the 2h within the day. Load forecast to a few hours, in order to timely correct the operational plan e short-term load forecast, extended short-term load forecast, ultra-short-term load forecast as a whole inseparable i
Rayon Woven Printed And Dyed Fabric
Rayon Woven Print Fabric,Rayon Twill Fabric,Rayon Satin Fabric,Rayon Crepe Fabric
Shaoxing RongMai Trading Co., Ltd , https://www.rm-fabric.com