Multi-source data fusion and intelligent scheduling method under virtual power plant platform

Abstract

The invention discloses a multi-source data fusion and intelligent scheduling method under a virtual power plant platform, relates to the technical field of virtual power plant optimal scheduling, and solves the technical problem that accurate adjustment of energy output is difficult to realize due to lack of a scientific energy scheduling priority evaluation system. According to the method, comprehensive data cleaning and deep fusion are carried out on the multi-source data, the data quality is improved, the potential correlation between the data is mined, the LSTM model and the fusion feature data are combined to establish the short-term load prediction model, the load change trend can be better captured, and the load prediction accuracy is improved. An energy scheduling priority evaluation system is constructed from three dimensions of policy scoring, economical efficiency scoring and adjustment capability scoring, comprehensive scoring and sorting are calculated through weighting, utilization of high-priority energy is guaranteed preferentially, and a refined scheduling strategy is formulated based on priority sorting and a load difference value when a load is short of supply or is over supply.

Description

Technical Field

[0001] The present invention relates to the technical field of virtual power plant optimization scheduling, and specifically to a multi-source data fusion and intelligent scheduling method under a virtual power plant platform. Background Art

[0002] With the introduction and development of the concept of virtual power plants, its ability to integrate distributed energy, energy storage, controllable loads and other resources has brought new possibilities for the efficient operation of power systems.

[0003] According to the patent application with publication number CN119721448A, an artificial intelligence-based virtual power plant optimization scheduling method is disclosed, which further reduces the power operation cost through integrated intelligent management and scheduling strategies. The system consists of key components such as the power generation end, energy storage end, power consumption end, system management platform, power grid and power market. The core is to use the deep deterministic policy gradient (DDPG) algorithm to build a simulation environment model and design a reward function to evaluate the economic benefits of the strategy. By initializing the Actor and Critic neural network, the system can generate and evaluate the scheduling strategy, and continuously iterate and update the network through the experience replay mechanism to optimize the economic benefits.

[0004] However, existing short-term load forecasting methods are unable to accurately capture the laws of load changes and are unable to effectively combine the characteristics of multi-source data, resulting in a large deviation between the forecast results and the actual load demand, and unable to provide a reliable basis for scheduling decisions. At the same time, there is a lack of a scientific energy scheduling priority evaluation system, and it is impossible to quantitatively rank different energy types according to factors such as policy, economy and regulation capacity; when there is an imbalance between load supply and demand, it is difficult to achieve precise regulation of energy output, resulting in energy waste or insufficient power supply, and unable to ensure stable operation of the power grid and maximize economic benefits. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a multi-source data fusion and intelligent scheduling method under the virtual power plant platform, which solves the problem of lack of a scientific energy scheduling priority evaluation system and difficulty in achieving precise regulation of energy output.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a multi-source data fusion and intelligent scheduling method under a virtual power plant platform, the method specifically comprising the following steps:

[0007] The collected multi-source data of virtual power plants are cleaned to obtain pre-processed data, and multi-source fusion is performed to obtain fused feature data;

[0008] A short-term load forecasting model is established based on the fused feature data and combined with the LSTM model. The short-term predicted load is obtained and compared with the preset value to generate real-time monitoring or real-time dispatch signals.

[0009] Analyze real-time dispatch signals, assign policy scores, economic scores, and regulatory capacity scores to different energy types, and then perform weighted summation to obtain a comprehensive score. At the same time, sort the scores from largest to smallest to generate dispatch priority ranking information.

[0010] Calculate the real-time load and determine the relationship between the value and zero to generate a signal of insufficient supply or oversupply;

[0011] Analyze the supply-demand imbalance signal to determine whether the real-time output of the energy type with the highest scheduling priority has reached the output upper limit, generate a round-robin scheduling signal, and then adjust the energy type allocation based on the predicted load difference to generate energy scheduling information;

[0012] Analyze oversupply signals, calculate real-time predicted load differences and determine warning levels, classify energy types, and perform energy scheduling to generate energy scheduling information.

[0013] As a further solution of the present invention, the specific method of generating the real-time monitoring or real-time scheduling signal is:

[0014] By using the fused feature data and combining it with the LSTM model, a short-term load forecasting model is constructed to predict the short-term load. The operator sets the preset value and compares the prediction result with the short-term predicted load.

[0015] If the short-term load is greater than the preset value, it means that the total power of the power plant meets the demand and a real-time monitoring signal is generated; if it is less than the preset value, it indicates that there is insufficient power and a real-time dispatching signal is generated.

[0016] As a further solution of the present invention, the specific method of generating the scheduling priority sorting information by sorting from largest to smallest is:

[0017] The energy type is labeled as i, and i = 1, 2, ..., j, where j represents the energy type. A comprehensive analysis is conducted on the scheduling priority of each energy type from the perspectives of policy, economy, and regulation capacity to obtain the corresponding policy score Zi, economy score Ji, and regulation capacity score Ti;

[0018] The three are weighted and added together to calculate the comprehensive score Qi according to the formula Qi=Zi×a1+Ji×a2+Ti×a3, where a1, a2 and a3 are the corresponding weight coefficients. The comprehensive score Qi is sorted from large to small to generate the scheduling priority sorting information.

[0019] As a further solution of the present invention, the specific method of generating the supply shortage signal or the oversupply signal is:

[0020] Then calculate the real-time load difference according to the formula ΔL 预测 =L 预测-(P 优先 +P 其他 )×(1-η) to calculate the predicted load difference ΔL 预测 , where L 预测 Expressed as short-term forecast load, P 优先 Represents the output load of the priority energy, P 其他 Represents the output load of other energy sources, η is the grid loss, and the predicted load difference ΔL 预测 The numerical value of is judged;

[0021] If it is greater than zero, a shortage signal is generated; if it is less than zero, an oversupply signal is generated.

[0022] As a further solution of the present invention, the specific method of analyzing the supply-demand imbalance signal is:

[0023] The energy type with the highest scheduling priority is obtained as the type to be analyzed, and its real-time output and output upper limit are obtained. If the real-time output has reached the upper limit, a round-robin scheduling signal is generated and analyzed at the same time;

[0024] If the upper limit is not reached, the output is adjusted to the upper limit and energy scheduling information is generated.

[0025] As a further solution of the present invention, the specific method of analyzing the round-robin scheduling signal is:

[0026] Obtain the secondary energy type with the highest scheduling priority and mark it as the type to be adjusted, denoted as n, where n = 1, 2, ..., m, where m represents the energy type corresponding to the type to be adjusted. Calculate the adjustable capacity of the secondary priority energy in turn and allocate output according to the formula: Calculate the distribution output corresponding to the type n to be adjusted, where min(……) is the minimum value function, P max It is the maximum adjustable power of the secondary priority energy source itself, and energy scheduling information is generated at the same time.

[0027] As a further solution of the present invention, the specific method of analyzing the oversupply signal is:

[0028] Obtain the real-time working energy type of the power plant and classify it into renewable energy and other energy. At the same time, according to the formula ΔL1=(P 优先 +P 其他 )×(1-η)-L 预测 Calculate the real-time predicted load difference ΔL1, and perform classification processing according to the calculated real-time predicted load difference ΔL1 to obtain corresponding classification information;

[0029] Then, based on the obtained classification information, the renewable energy and other energy sources are dispatched and analyzed. For renewable energy, real-time output is maintained and a priority consumption strategy is adopted. For other energy sources, real-time output is adjusted and not lower than the minimum technical output, and energy dispatch information is generated at the same time.

[0030] The present invention provides a multi-source data fusion and intelligent scheduling method under the virtual power plant platform. Compared with the existing technology, it has the following advantages:

[0031] The present invention improves data quality and mines potential correlations between data by performing comprehensive data cleaning and deep fusion on multi-source data. It combines the LSTM model and fused feature data to establish a short-term load forecasting model, which can better capture the load change trend and construct an energy scheduling priority evaluation system from three dimensions: policy score, economic score, and regulation capability score. By weighted calculation and comprehensive scoring and sorting, the utilization of high-priority energy is prioritized. At the same time, when the load is in short supply or in oversupply, a refined scheduling strategy is formulated based on priority sorting and load difference. When the supply is insufficient, the output of high-priority energy is prioritized, and the secondary priority energy is reasonably allocated to ensure power supply. When the supply is in oversupply, the excess power is processed in a hierarchical manner, renewable energy is preferentially absorbed, and the output of other energy sources is scientifically regulated, effectively avoiding energy waste, improving energy utilization efficiency, and ensuring the stability and economy of power grid operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a diagram of the steps of the present invention. DETAILED DESCRIPTION

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0034] See also Figure 1 , this application provides a multi-source data fusion and intelligent scheduling method under a virtual power plant platform, which specifically includes the following steps:

[0035] Step S1: Collect multi-source data of the virtual power plant, where the multi-source data includes distributed energy data, power grid operation data, meteorological and environmental data, user behavior data and market transaction data, and perform data cleaning on it to obtain pre-processed data. The data cleaning includes outlier detection and missing value filling. At the same time, the multi-source data is normalized and standardized, and the pre-processed data is multi-source fused to obtain fused feature data. The multi-source fusion of data here is an existing technology and will not be elaborated on here.

[0036] Step S2: Establish a short-term load forecasting model based on the obtained fusion feature data and in combination with the LSTM model, and perform short-term load forecasting. At the same time, compare the model with a preset value, and the specific value of the preset value is set by the operator;

[0037] If the short-term load is greater than the preset value, it means that the current total power of the power plant meets the short-term load demand, and a real-time monitoring signal is generated. Conversely, if the short-term load is less than the preset value, it means that the current total power of the power plant does not meet the short-term load demand, and a real-time power dispatching signal is generated.

[0038] Consider a scenario where, at 9:00 AM on a summer weekday, a city power grid dispatch center needs to forecast power load for the next four hours (9:00 AM to 1:00 PM) and make dispatch decisions. The dispatch center collects load data from the past week, meteorological data (temperature, humidity, wind speed), and the current date and time, and performs data cleaning and feature engineering. For example, derived features such as the average load over the past three hours and the difference between the current temperature and the historical temperature over the same period are calculated. All data is then normalized and converted into a format acceptable to the LSTM model.

[0039] The optimized LSTM model is trained on processed historical data. After training, the real-time feature data at 9 o'clock is fed into the model to predict the load for the next four hours. Assuming the predicted results are: 1200MW at 9 o'clock, 1300MW at 10 o'clock, 1350MW at 11 o'clock, 1400MW at 12 o'clock, and 1380MW at 13 o'clock, the operator sets the preset value to 1300MW based on historical data and the weather conditions of the day. The operator compares the predicted average load ((1200+1300+1350+1400+1380) / 5=1326MW) with the preset value. Since 1326MW is greater than 1300MW, it indicates that the current total power generation of the power plant meets the short-term load demand. The system generates a real-time monitoring signal, and the dispatcher continues to monitor the grid operation status.

[0040] Step S3: Analyze the generated real-time power dispatch signal, obtain the energy type of the power plant and label it as i, where i=1, 2, ..., j, where j represents the energy type. At the same time, analyze and determine the dispatch priority corresponding to energy type i. The specific dispatch priority is comprehensively analyzed from three aspects: policy score, economic score, and regulation capability score. As shown in the following energy type single indicator score table, the policy score Zi, economic score Ji, and regulation capability score Ti corresponding to energy type i are obtained;

[0041]

[0042] At the same time, the weight coefficients corresponding to the policy score Zi, economic score Ji, and regulation capacity score Ti are obtained and recorded as a1, a2, and a3, respectively. The specific values of the weight coefficients are set by the operator. Generally, a1 = 0.58, a2 = 0.27, and a3 = 0.15. The three are weighted and added together to obtain the comprehensive score Qi corresponding to energy type i. The calculation formula is Qi = Zi × a1 + Ji × a2 + Ti × a3. The comprehensive score Qi values are sorted from large to small to generate the scheduling priority sorting information. Combined with the above score table, the comprehensive scores of different types of energy are calculated as follows:

[0043] Wind power / photovoltaic: Q1 = 5 × 0.58 + 5 × 0.27 + 2 × 0.15 = 4.55;

[0044] Gas-fired power plant: Q2 = 3 × 0.58 + 3 × 0.27 + 5 × 0.15 = 3.36;

[0045] Coal-fired power plant: Q3 = 2 × 0.58 + 2 × 0.27 + 1 × 0.15 = 1.85;

[0046] Battery energy storage: Q4 = 4 × 0.58 + 4 × 0.27 + 5 × 0.15 = 4.17;

[0047] After sorting according to the calculated comprehensive score, wind power / photovoltaic power > energy storage > gas-fired power plant > coal-fired power plant.

[0048] Then calculate the real-time load difference according to the formula ΔL 预测 =L 预测 -(P 优先 +P 其他 )×(1-η) to calculate the predicted load difference ΔL 预测 , where L 预测 Expressed as short-term forecast load, P 优先 Represents the output load of the priority energy, P 其他 Indicates the output load of other energy sources, η is the grid loss, generally taken as (3%-5%), and the predicted load difference ΔL预测 The numerical value of is judged;

[0049] If the predicted load difference ΔL 预测 If the short-term forecast load exceeds the total output load of the priority energy, a supply-demand shortfall signal is generated. Conversely, if the forecast load difference ΔL 预测 If it is less than zero, it means that the short-term forecast load does not exceed the total output load of the priority energy resources and generates an oversupply signal.

[0050] For example, if the short-term forecast load L is known 预测 =2000MW, output load of priority energy (wind power and photovoltaic) P 优先 =1200MW, output load of other energy sources P 其他 =500MW, grid loss rate η = 4%, calculate the predicted load difference: ΔL 预测 =2000-(1200+500)x(1-0.04)=332MW>0. Since the predicted load difference is greater than zero, a supply-demand imbalance signal is generated.

[0051] Step S4: Analyze the generated supply-demand imbalance signal, obtain the energy type with the highest scheduling priority, mark it as the type to be analyzed, and obtain the real-time output and output upper limit of the type to be analyzed. Then determine whether the real-time output reaches the output upper limit. If so, generate a round-robin scheduling signal. If not, adjust the real-time output of the type to be analyzed based on the output upper limit to generate energy scheduling information.

[0052] Upon receiving a signal of insufficient supply, the dispatch system first retrieves the dispatch priority ranking from the energy dispatch priority database, which is calculated based on a combination of factors including policy scores, economic scores, and regulatory capacity scores. Using real-time monitoring from a distributed sensor network, the system identifies the energy type with the highest priority and marks it as a candidate for analysis.

[0053] At the same time, with the help of high-precision metering equipment and equipment operation parameter monitoring modules, the current real-time output power value of the energy type to be analyzed can be accurately obtained, as well as the output upper limit value determined by the energy equipment based on its own technical parameters (such as unit rated power, energy storage charging and discharging limits, etc.) and grid operation constraints (such as voltage and frequency limits, etc.).

[0054] The system accurately compares the real-time output power and output upper limit of the energy type to be analyzed. If the real-time output power reaches or exceeds the output upper limit (considering a certain error range, such as ±1%), it indicates that the energy can no longer meet the load demand by increasing its own output. At this time, the system will generate a round-robin scheduling signal according to the pre-set round-robin scheduling rules (such as switching to the next priority energy in sequence according to the energy type priority list). The signal will be transmitted to the corresponding energy scheduling execution unit to trigger the start-up or output adjustment operation of the next level of priority energy.

[0055] If the real-time output power does not reach the output upper limit, the system will calculate a reasonable adjustment amount based on the current load gap size and the regulation characteristics of the energy type to be analyzed (such as ramp rate, response time, etc.), with the output upper limit as the target, through advanced automatic control algorithms (such as PID control algorithm, model predictive control algorithm, etc.). Then, the adjustment instruction is sent to the control terminal of the energy type to be analyzed to accurately adjust its real-time output and generate detailed energy scheduling information. The energy scheduling information includes the real-time output power before adjustment, the adjustment target value, the changes in key parameters during the adjustment process, and the expected output power after adjustment, so that the dispatcher can conduct real-time monitoring and subsequent analysis.

[0056] Suppose a city's power grid receives a signal of insufficient supply during peak summer electricity demand. The dispatch system analyzes wind power and identifies it as the energy source with the highest dispatch priority, making it the target for analysis. The current real-time output of wind power is 80 MW, while the upper limit is 100 MW based on wind turbine equipment parameters and grid operating conditions.

[0057] Because 80MW falls short of the 100MW output ceiling, the system uses a model predictive control algorithm, taking into account the current grid load shortfall of 50MW and the wind power ramp rate, to calculate the need to increase wind power output to 100MW. The system then sends a regulation command to the wind power control terminal and generates energy dispatch information, recording the pre-regulation output of 80MW, the regulation target of 100MW, the estimated regulation time of 5 minutes, and the expected post-regulation output of 100MW, for dispatchers to monitor and analyze.

[0058] If the real-time wind power output reaches the upper limit of 100MW at this time, the system will generate a round-robin dispatch signal according to the round-robin dispatch rules, notify the energy storage system to prepare to start discharging to supplement the power supply gap, and feedback the relevant operation records to the dispatcher.

[0059] Then, the generated round-robin scheduling signal is analyzed to obtain the next energy type with the highest scheduling priority, and marked as the type to be adjusted, denoted as n, where n = 1, 2, ..., m, where m represents the energy type corresponding to the type to be adjusted. The real-time output of the type to be adjusted is adjusted and analyzed based on the predicted load difference. The specific adjustment analysis method is as follows:

[0060] Calculate the adjustable capacity of secondary priority energy sources (such as energy storage and gas-fired power generation) in turn, and allocate output according to the formula: Calculate the distribution output corresponding to the type n to be adjusted, where min(……) is the minimum value function, P max The maximum adjustable power of the secondary energy source itself, and energy scheduling information is generated at the same time;

[0061] Assume that the current forecast load difference ΔL 预测 = 100MW, there are two secondary priority energy sources: energy storage (energy and gas power generation, the maximum adjustable power of energy storage P max The maximum adjustable power of gas-fired power generation is 30MW, with a comprehensive score of 8. max The total energy consumption is 50MW and the comprehensive score is 7. First, the total comprehensive score of all energy resources to be used is calculated to be 15.

[0062] For energy storage: Theoretically, the proportional output is ΔL 预测 =100×8 / 15=53.3MW, but due to the maximum adjustable power P of energy storage max If it is 30MW, the min(……) function takes the smaller value, so the final output of the energy storage is 30MW;

[0063] For gas-fired power generation: First calculate the remaining load difference as 100-30=70MW. The theoretically proportional output of gas-fired power generation is 100×7 / 7=70MW, but the maximum adjustable power P of gas-fired power generation is max It is 50MW, so the final output of energy storage is 70MW.

[0064] Step S5: Analyze the generated oversupply signal to obtain the real-time working energy type of the power plant, and classify it into renewable energy and other energy sources. At the same time, according to the formula ΔL1=(P 优先 +P 其他 )×(1-η)-L 预测 The real-time forecast load difference ΔL1 is calculated. For example, the short-term forecast load is 300MW, the wind power output is 200MW, and the coal-fired unit output is 150MW. The difference is (200+150)-300=50MW. The calculated real-time forecast load difference ΔL1 is then used for classification processing to obtain corresponding classification information. The specific classification processing method is as follows:

[0065] Level 1 (0MW<ΔL1≤50MW): Slight overcapacity, flexible regulation is activated first;

[0066] Level 2 (50MW<ΔL1≤150MW): moderate overcapacity, requiring the linkage of multiple regulation measures;

[0067] Level 3 (ΔL1>150MW): Severe overcapacity, emergency absorption plan activated;

[0068] Then, the renewable energy and other energy sources are dispatched and analyzed based on the obtained classification information;

[0069] For renewable energy, maintain real-time output and adopt a priority consumption strategy. For other energy sources, adjust real-time output and make sure it is not lower than the minimum technical output, and generate energy scheduling information at the same time.

[0070] For example, a regional power grid experiences oversupply at night in summer. The specific parameters are as follows:

[0071] The short-term forecast load is 300MW, with renewable energy sources including 200MW of wind power and 50MW of photovoltaic power (zero output at night). Other energy sources include 150MW of coal-fired units and 80MW of gas-fired units. The grid loss coefficient η is 4%. The corresponding real-time forecast load difference calculated using the formula is 122.88MW. This results in a Level 2 warning. The dispatch analysis for renewable energy and other energy sources is then conducted separately.

[0072] Renewable energy dispatch: Wind power maintains 200MW output; inter-provincial interconnection lines are activated to transmit 50MW; energy storage systems are charged at 40MW power (SOC = 75%); and data centers are encouraged to increase electricity consumption by 30MW.

[0073] Other energy scheduling: gas-fired units reduce output to 30MW (minimum technical output); coal-fired units reduce output to 100MW (regulation cost is lower than gas-fired units).

[0074] Energy scheduling information is generated based on the above scheduling analysis.

[0075] Some of the data in the above formulas are calculated based on their numerical values and are not substituted into parameter units for calculation. At the same time, the contents not described in detail in this specification belong to the existing technology known to those skilled in the art.

[0076] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A multi-source data fusion and intelligent scheduling method under a virtual power plant platform, characterized in that: The method specifically comprises the following steps: The collected multi-source data of virtual power plants are cleaned to obtain pre-processed data, and multi-source fusion is performed to obtain fused feature data; A short-term load forecasting model is established based on the fused feature data and combined with the LSTM model. The short-term predicted load is obtained and compared with the preset value to generate real-time monitoring or real-time dispatch signals. Analyze real-time dispatch signals, assign policy scores, economic scores, and regulatory capacity scores to different energy types, and then perform weighted summation to obtain a comprehensive score. At the same time, sort the scores from largest to smallest to generate dispatch priority ranking information. Calculate the real-time load and determine the relationship between the value and zero to generate a signal of insufficient supply or oversupply; Analyze the supply-demand imbalance signal to determine whether the real-time output of the energy type with the highest scheduling priority has reached the output upper limit, generate a round-robin scheduling signal, and then adjust the energy type allocation based on the predicted load difference to generate energy scheduling information; Analyze oversupply signals, calculate real-time predicted load differences and determine warning levels, classify energy types, and perform energy scheduling to generate energy scheduling information.

2. The multi-source data fusion and intelligent scheduling method under the virtual power plant platform according to claim 1 is characterized in that: The specific method of generating the real-time monitoring or real-time scheduling signal is: By using the fused feature data and combining it with the LSTM model, a short-term load forecasting model is constructed to predict the short-term load. The operator sets the preset value and compares the prediction result with the short-term predicted load. If the short-term load is greater than the preset value, it means that the total power of the power plant meets the demand and a real-time monitoring signal is generated; If it is less than the preset value, it indicates insufficient power and a real-time dispatch signal is generated.

3. The multi-source data fusion and intelligent scheduling method under the virtual power plant platform according to claim 1 is characterized in that: The specific method of generating the scheduling priority sorting information by sorting from largest to smallest is: The energy type is labeled as i, and i = 1, 2, ..., j, where j represents the energy type. A comprehensive analysis is conducted on the scheduling priority of each energy type from the perspectives of policy, economy, and regulation capacity to obtain the corresponding policy score Zi, economy score Ji, and regulation capacity score Ti; The three are weighted and added together to calculate the comprehensive score Qi according to the formula Qi=Zi×a1+Ji×a2+Ti×a3, where a1, a2 and a3 are the corresponding weight coefficients. The comprehensive score Qi is sorted from large to small to generate the scheduling priority sorting information.

4. The multi-source data fusion and intelligent scheduling method under the virtual power plant platform according to claim 1 is characterized in that: The specific method of generating the supply shortage signal or the oversupply signal is: Then calculate the real-time load difference according to the formula ΔL 预测 =L 预测 -(P 优先 +P 其他 )×(1-η) to calculate the predicted load difference ΔL 预测 , where L 预测 Expressed as short-term forecast load, P 优先 Represents the output load of the priority energy, P 其他 Represents the output load of other energy sources, η is the grid loss, and the predicted load difference ΔL 预测 The numerical value of is judged; If it is greater than zero, a shortage signal is generated; if it is less than zero, an oversupply signal is generated.

5. The multi-source data fusion and intelligent scheduling method under the virtual power plant platform according to claim 1 is characterized in that: The specific method of analyzing the signal of supply exceeding demand is as follows: The energy type with the highest scheduling priority is obtained as the type to be analyzed, and its real-time output and output upper limit are obtained. If the real-time output has reached the upper limit, a round-robin scheduling signal is generated and analyzed at the same time; If the upper limit is not reached, the output is adjusted to the upper limit and energy scheduling information is generated.

6. The multi-source data fusion and intelligent scheduling method under the virtual power plant platform according to claim 5 is characterized in that: The specific method of analyzing the round-robin scheduling signal is as follows: Obtain the secondary energy type with the highest scheduling priority and mark it as the type to be adjusted, denoted as n, where n = 1, 2, ..., m, where m represents the energy type corresponding to the type to be adjusted. Calculate the adjustable capacity of the secondary priority energy in turn and allocate output according to the formula: Calculate the distribution output corresponding to the type n to be adjusted, where min(……) is the minimum value function, P max It is the maximum adjustable power of the secondary priority energy source itself, and energy scheduling information is generated at the same time.

7. The multi-source data fusion and intelligent scheduling method under the virtual power plant platform according to claim 1 is characterized in that: The specific method of analyzing the oversupply signal is as follows: Obtain the real-time working energy type of the power plant and classify it into renewable energy and other energy. At the same time, according to the formula ΔL1=(P 优先 +P 其他 )×(1-η)-L 预测 Calculate the real-time predicted load difference ΔL1, and perform classification processing according to the calculated real-time predicted load difference ΔL1 to obtain corresponding classification information; Then, based on the obtained classification information, the renewable energy and other energy sources are dispatched and analyzed. For renewable energy, real-time output is maintained and a priority consumption strategy is adopted. For other energy sources, real-time output is adjusted and not lower than the minimum technical output, and energy dispatch information is generated at the same time.

Images