CN-121984121-A - Micro-grid peak regulation and frequency modulation cooperative control method based on AI and numerical calculation

CN121984121ACN 121984121 ACN121984121 ACN 121984121ACN-121984121-A

Abstract

The invention relates to a micro-grid peak regulation and frequency modulation cooperative control method based on AI and numerical calculation, and belongs to the technical field of automatic control of electric power systems. The method solves the technical problems that a traditional numerical calculation model is poor in self-adaption and an artificial intelligent model is low in initial reliability in micro-grid dispatching under the condition that wind, solar and fire comprehensive energy is accessed, the numerical calculation dispatching is preferentially adopted in the initial stage of micro-grid operation by combining an AI algorithm and a numerical algorithm, safety is guaranteed, the AI dispatching model continuously learns and optimizes after data are accumulated, the AI dispatching is switched to AI dispatching after the standard is judged based on the maturity of multi-index weighted scoring, and meanwhile, equipment such as a capacitor, electrochemical energy storage and the like is coordinated by adopting a multi-time-scale hierarchical control strategy. The technical effect is to improve the safety, accuracy, economy and reliability of micro-grid dispatching and realize the comprehensive performance optimization of the whole life cycle.

Inventors

WU YA
WANG FANGMING
WANG YI
ZHU MENG
Du Jiantai
ZHOU TAO
LI HOUXING
CHEN LI
LI BIN

Assignees

重庆赛迪热工环保工程技术有限公司

Dates

Publication Date: 20260505
Application Date: 20260116

Claims (10)

1. A micro-grid peak regulation and frequency modulation cooperative control method based on AI and numerical calculation is characterized by comprising the following steps: Step 1, initializing a data source, configuring an electric equipment and power generation equipment information database, a whole plant process production plan, a meteorological environment information database, a power frequency record database, a peak regulation and frequency modulation equipment information database and a market price information database, wherein a user can autonomously set, update or delete data in the database; Step 2, initializing constraint conditions, and setting multi-time layered control constraint, equipment service life constraint, power balance constraint, unit operation constraint and energy storage system constraint; step 3, based on the data source in the step1, predicting future electric power, generating power and frequency parameters in parallel through a numerical calculation scheduling model and an AI scheduling model; step4, parallelly calculating the future peak regulation demand and the frequency modulation demand through a numerical calculation scheduling model and an AI scheduling model; Step 5, comparing the adjustment demand of the numerical calculation scheduling model with the adjustment demand of the AI scheduling model, and judging the accuracy; step 6, respectively calculating peak regulation and frequency modulation capacity ratios of different devices based on the constraint conditions of the step 2 by using the numerical calculation scheduling model and the AI scheduling model; Step 7, carrying out comprehensive performance evaluation calculation on the basis of the capacity ratio of the step 6 by using a numerical calculation scheduling model and an AI scheduling model to optimize and generate a scheduling scheme; Step 8, judging the maturity of the comprehensive performance evaluation result of the AI scheduling model, wherein the maturity judgment comprises a speed index, a precision index, an economic index and a reliability index, wherein the instruction of the AI scheduling model is selected to be executed as a scheme 1 when the maturity is qualified, and the instruction of the numerical calculation scheduling model is selected to be executed as a scheme 2 when the maturity is unqualified; step 9, executing the load control instruction of the scheme selected in the step 8; And 10, tracking operation data and feeding back the operation data to the database in the step1, wherein the operation data comprise power, frequency, electricity consumption data, power generation data and equipment states.
2. The micro-grid peak regulation and frequency modulation cooperative control method based on AI and numerical calculation of claim 1 is characterized in that the micro-grid peak regulation and frequency modulation control adopts a multi-time scale layered control strategy, including primary frequency modulation, secondary frequency modulation, tertiary frequency modulation, low frequency load shedding and high frequency cutting; the primary frequency modulation uses at least one device of a capacitor, a wind power active reserve, a photovoltaic active reserve, a sagging characteristic of a thermal power generating unit, an electrochemical energy storage device and a fused salt electric heater; the secondary frequency modulation uses at least one device of electrochemical energy storage, sagging characteristic of a thermal power generating unit and a molten salt energy storage system; All the generator sets are used for the tertiary frequency modulation; the low-frequency load shedding is performed through a low-frequency relay when the frequency is lower than a first safety threshold value; the high-frequency cutting machine operates through the fused salt electric heater and reduces load of the generator set cutting machine when the frequency is higher than a second safety threshold; ‌ in normal operation of ‌, dividing the frequency deviation limit value of the isolated network system according to the capacity, wherein the frequency deviation of the system with the capacity of ‌ of 300 kilowatts and ‌ above is not more than ‌ +/-0.2 Hz ‌, the frequency deviation of the system with the capacity of ‌ of less than 300 kilowatts of ‌ is not more than ‌ +/-0.5 Hz ‌; ‌ under ‌ accidents or abnormal working conditions, and the frequency deviation of the isolated network system is not more than ‌ +/-1 Hz ‌; ‌ setting a first primary frequency modulation dead zone, wherein the frequency range of a power grid is 50.03 Hz-50.10 Hz; If the frequency deviation exists for more than 30 seconds after the adjustment and exceeds the allowable range, namely 50.40Hz, a secondary frequency modulation system is activated to process frequency components; if the secondary frequency modulation resource is more than 90% of rated capacity and the frequency deviation exists all the time, the frequency deviation cannot be recovered to 50.00Hz, or the power generation electricity utilization economy is temporarily required to be adjusted, and the tertiary frequency modulation is activated; if the frequency modulation resource is used up for three times or the frequency deviation exceeds a third safety threshold value, namely, exceeds 50.50Hz, triggering a high-frequency cutting machine which is divided into more than 2 rounds of step-by-step cutting machines until the frequency stops to rise and starts to fall back; setting a second primary frequency modulation dead zone, wherein the frequency range of a power grid is 49.97Hz to-49.90 Hz, and starting primary frequency modulation equipment when the frequency of the power grid is lower than 49.90Hz, and if the frequency deviation exists for more than 30 seconds after adjustment and exceeds the allowable range, namely 49.60Hz, activating a secondary frequency modulation system and processing frequency components; If the secondary frequency modulation resource is more than 90% of rated capacity and the frequency deviation exists all the time, the frequency cannot be recovered to 50Hz, or the power generation electricity economy needs to be adjusted temporarily, and the tertiary frequency modulation is activated; If the frequency deviation exceeds a fourth safety threshold value, namely 48.50Hz, the frequency deviation is used up for three times, the low-frequency load shedding is triggered, and the low-frequency load shedding is performed in more than 2 rounds of step-by-step load shedding until the frequency stops dropping and starts rising.
3. The micro-grid peak regulation and frequency modulation cooperative control method based on AI and numerical calculation of claim 1, wherein the AI scheduling model adopts a machine learning algorithm to conduct data prediction, and the machine learning algorithm comprises a neural network, knowledge transfer learning, gradient lifting tree model or imitation learning; the AI scheduling model enables data prediction by: Learning historical data of active power and reactive power of electric equipment in the micro-grid to predict future power; Learning the frequency deviation historical data of the micro-grid system to predict future frequency parameters; The method comprises the steps of counting and learning equipment information in a micro-grid, monitoring equipment states and providing overhaul and maintenance guidance; Learning equipment operation fault history data to conduct fault prediction; Learning meteorological parameter historical data to predict wind power and photovoltaic power generation power, wherein the meteorological parameters comprise weather, temperature, humidity and altitude; learning peak shaving service, paid auxiliary service and capacity lease market price historical data to predict price; learning market energy prices and pollutant treatment price historical data to predict prices; real-time accounting peak and frequency regulation capacity of power generation equipment, electric equipment, energy storage equipment and peak and frequency regulation equipment in the micro-grid; Distributing equipment peak regulation and frequency modulation proportion by adopting a scheduling algorithm, wherein the scheduling algorithm comprises deep reinforcement learning, a neural network, knowledge transfer learning or imitation learning; The scheduling model is updated and optimized periodically.
4. The micro-grid peak regulation and frequency modulation cooperative control method based on AI and numerical calculation of claim 1, wherein the numerical calculation scheduling model adopts a numerical algorithm to conduct data prediction, and the numerical algorithm comprises a hierarchical clustering algorithm, a K-means algorithm, a Gaussian mixture model clustering algorithm, a DBSCAN algorithm or a statistical analysis and probability analysis method; The numerical computation scheduling model generates scheduling instructions by: Real-time accounting of peak and frequency regulation available capacity of power generation equipment, electric equipment, energy storage equipment and peak and frequency regulation equipment in the micro-grid; Adopting a scheduling algorithm to accurately solve equipment peak regulation and frequency modulation proportion distribution; Based on the multi-time scale accounting speed index, the accuracy index, the economic index and the reliability index, an equipment load control instruction is generated as a reference execution scheme 2.
5. The micro-grid peak regulation and frequency modulation cooperative control method based on AI and numerical calculation of claim 1, wherein the regulation rate deviation in the maturity judgment is calculated by evaluating response time and regulation rate, the response time is a time interval from the dispatch instruction to the beginning of the change of the power of the controlled unit, and the regulation rate is the power change in unit time; The regulation precision deviation is calculated by evaluating frequency deviation, voltage deviation and power angle stability, wherein the frequency deviation is according to GB/T15945 standard, the voltage deviation is according to GB/T12325 standard, the power angle stability requires that the system is stable under single fault, and the damping ratio of a key oscillation mode is more than 5%; the economic benefit index is based on scheduling benefit and cost accounting of thermal power units, electrochemical energy storage, wind power units, photovoltaic power units and molten salt energy storage; The model reliability index is calculated by the availability and instruction rejection rate.
6. The micro-grid peak regulation and frequency modulation cooperative control method based on AI and numerical calculation of claim 5, wherein the maturity judgment adopts a hierarchical analysis method to determine weights of economic benefit indexes, regulation rate deviation, regulation precision deviation and model reliability indexes, and obtains a comprehensive score through weighted scoring, wherein the standard value of the weighted scoring can be manually adjusted.
7. The micro-grid peak regulation and frequency modulation cooperative control method based on AI and numerical calculation of claim 1, wherein the method further comprises the step of judging the accuracy of power frequency prediction by an AI scheduling model; The accuracy judgment is carried out by comparing a numerical calculation scheduling model result with an AI scheduling model prediction result, estimating the prediction accuracy by using a Root Mean Square Error (RMSE) or a Mean Absolute Error (MAE), and carrying out capacity ratio accounting when the accuracy reaches the standard, otherwise, returning to reinforcement learning optimization.
8. The micro-grid peak regulation and frequency modulation cooperative control method based on AI and numerical computation of claim 1 is characterized in that the method is used for executing a numerical computation scheduling model initially by default, the AI scheduling model is continuously learned and optimized online, and is automatically switched to the AI scheduling model when the maturity is judged to be up to standard and supports manual switching of the scheduling model by a user.
9. The micro-grid peak regulation and frequency modulation cooperative control method based on AI and numerical calculation as set forth in claim 1, wherein in said comprehensive performance evaluation, economic benefit index S is calculated by the formula Calculating, wherein C is thermal power unit scheduling benefit, C is thermal power unit scheduling cost, D is electrochemical energy storage scheduling benefit, D is electrochemical energy storage scheduling cost, F is wind power unit scheduling benefit, F is wind power unit scheduling cost, G is photovoltaic power unit scheduling benefit, G is photovoltaic power unit scheduling cost, E is fused salt energy storage scheduling benefit, E is fused salt energy storage scheduling cost, and equipment peak regulation and frequency modulation capacity proportion is dynamically adjusted according to the emergency degree and economic degree of the system to optimize.
10. A micro-grid peak regulation and frequency modulation cooperative control system based on artificial intelligence AI and numerical calculation is characterized by comprising a microprocessor and a computer readable storage medium, wherein the microprocessor is programmed to execute the method of any one of claims 1-9, and the computer readable storage medium stores a computer program which when executed realizes the method of any one of claims 1-9.

Description

Micro-grid peak regulation and frequency modulation cooperative control method based on AI and numerical calculation Technical Field The invention belongs to the technical field of automatic control of power systems, and relates to a micro-grid peak regulation and frequency modulation cooperative control method based on AI and numerical calculation. Background As the duty cycle of renewable energy sources in electrical power systems increases gradually, the stable operation of micro-grids faces new challenges. Renewable energy sources such as wind power, photovoltaic and the like have intermittence and fluctuation in output, and have obvious influence on the frequency and power balance of the micro-grid. The traditional peak regulation and frequency modulation mode mainly using thermal power generating units is difficult to adapt to the rapid regulation requirement under the condition of high-proportion renewable energy access. The micro-grid system comprises various peak regulation and frequency modulation devices such as a capacitor, electrochemical energy storage, molten salt energy storage, photovoltaic, wind power and thermal power units and the like, and the control complexity is obviously increased. At present, micro-grid peak regulation and frequency modulation mainly depend on a traditional numerical calculation method, such as a numerical optimization algorithm. The method solves based on an accurate physical model, and has a certain effect in a small-scale system. However, as the source network and the load storage are coexisted, the system scale is enlarged, the traditional numerical calculation method faces the problem of geometric multiple increase of the calculated amount, and efficient solution of large-scale combination optimization is difficult to realize. In addition, the numerical calculation method relies on manual parameter adjustment, so that the adjustment difficulty is high when the external conditions change, and the self-adaptive capacity is lacked. In recent years, artificial intelligence (ARTIFICIALINTELLIGENCE, AI) technology has been increasingly applied to the field of power system scheduling. Machine learning algorithms offer advantages in dealing with non-linear, high dimensional problems by analyzing historical data to implement predictions and decisions. However, the artificial intelligence method has the problem of a black box model, and the decision process is difficult to explain. Under the condition of insufficient data accumulation or insufficient training, the reliability of the model prediction result is low, and the model prediction result is directly used for scheduling control and has safety risk. Relying solely on artificial intelligence scheduling may result in system operational instability due to inaccurate initial data and immature models. In the prior art, the numerical calculation method and the artificial intelligence method are independently applied and cannot be effectively combined. Although the numerical calculation method is high in reliability and suitable for small-scale solution, the optimal solution is not easy to find for large-scale combination optimization. Artificial intelligence methods are good at handling uncertainty but have insufficient initial reliability. The scheduling system with the numerical calculation and the AI intelligent algorithm is provided, the micro-grid operates in an initial stage, numerical calculation scheduling is prioritized, and initial data is ensured to be accurately and accurately checked. And in the micro-operation process stage, a micro-network operation database is continuously accumulated and perfected, an AI scheduling model is gradually perfected through AI continuous reinforcement learning until the maturity and accuracy of the estimated scheduling model reach the standards, and then the scheduling model is switched by feedback, so that the economical controllability of the scheduling process is fully ensured. Disclosure of Invention In view of the above, the present invention aims to provide a micro-grid peak regulation and frequency modulation cooperative control method based on AI and numerical calculation. In order to achieve the above purpose, the present invention provides the following technical solutions: A micro-grid peak regulation and frequency modulation cooperative control method based on AI and numerical calculation comprises the following steps: Step 1, initializing a data source, configuring an electric equipment and power generation equipment information database, a whole plant process production plan, a meteorological environment information database, a power frequency record database, a peak regulation and frequency modulation equipment information database and a market price information database, wherein a user can autonomously set, update or delete data in the database; Step 2, initializing constraint conditions, and setting multi-time layered control constraint, equipment service lif