Power grid operation optimization system and method based on artificial intelligence

Abstract

The invention relates to the technical field of power grid operation, in particular to a power grid operation optimization system and method based on artificial intelligence, and the system comprises a dynamic adaptive unit which collects real-time data, and predicts the follow-up power grid operation dynamic state according to the real-time data, thereby obtaining prediction data; when the system is used, the uncertainty in power grid dispatching is reduced, the dispatching strategy of the generator set is optimized, the energy consumption and the operation cost are reduced, the load fluctuation or fault possibly occurring in the power grid can be recognized in advance through the dynamic monitoring module, and the power grid dispatching efficiency is improved. The method improves the reliability of the power grid, discovers potential risks in advance, avoids power grid collapse or large-scale faults, formulates a more accurate unit scheduling strategy according to the real-time data and historical data of the power grid, improves the operation efficiency and economy of the power grid, provides effective support for power grid fault prediction, can quickly respond to the power grid change, and improves the reliability of the power grid. And the risk of fault occurrence is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid operation, and in particular to a power grid operation optimization system and method based on artificial intelligence. Background Art

[0002] The AI-based power grid operation optimization system is a solution that uses advanced AI technology to improve the efficiency, reliability and economy of power system operation, and realize intelligent management of the power grid.

[0003] The patent application number is CN202011462629.X, which states in the specification that "the present invention discloses a method for optimizing power grid operation, including: establishing a power grid energy-saving and economic optimization model; obtaining the calculation parameters required for the optimization model and the active output adjustment measures of the candidate units; correcting the adjustable space of the active output adjustment measures of the candidate units; adjusting the increase and decrease space of the generator according to multiple proportions to generate adjustable space plans of each proportion; optimizing the economic operation of each adjustable space plan; performing safety verification on each optimized plan under expected faults; selecting the plan with the greatest benefit from the plans that pass the safety verification as the final plan. The present invention can provide decision-making support for the optimization of power grid energy-saving and economic operation". With the increase in electricity demand and the continuous increase in the proportion of renewable energy, the above-mentioned technology faces challenges in unit scheduling, spare capacity configuration and fault warning. In addition, the above-mentioned technology relies on fixed parameters and it is difficult to provide efficient and flexible decision-making support in a complex power grid environment, especially when the power grid changes dynamically, the load fluctuates and the market price fluctuates. The performance of the above-mentioned technology will be affected, and it is difficult to adapt to various emergencies in power grid operation in real time.

[0004] To sum up, developing an artificial intelligence-based power grid operation optimization system and method is still a key issue that needs to be urgently solved in the field of power grid operation technology. Summary of the invention

[0005] The purpose of the present invention is to solve the problems existing in the prior art that with the increase in electricity demand and the continuous rise in the proportion of renewable energy, the above-mentioned technology faces challenges in unit scheduling, spare capacity configuration and fault warning. In addition, the above-mentioned technology relies on fixed parameters and is difficult to provide efficient and flexible decision support in a complex power grid environment. In particular, when the power grid changes dynamically, the load fluctuates and the market price fluctuates, the performance of the above-mentioned technology will be affected, and it will be difficult to adapt to various emergencies in the operation of the power grid in real time. The present invention provides a power grid operation optimization system and method based on artificial intelligence.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The present invention provides an artificial intelligence-based power grid operation optimization system, comprising: a dynamic adaptive unit, collecting real-time data, predicting subsequent power grid operation dynamics based on the real-time data, thereby obtaining prediction data;

[0008] An intelligent verification unit, which combines the predicted data with artificial intelligence to conduct dynamic security verification and formulate strategies;

[0009] A flexible optimization unit dynamically schedules access points according to the prediction data and continuously optimizes the formulated strategy;

[0010] The multi-level simulation unit simulates the synergistic effect between different units according to the dynamic scheduling access point.

[0011] Furthermore, the dynamic adaptive unit includes:

[0012] An adaptive module, which establishes a power grid load prediction model and a renewable energy prediction model according to the real-time data collected from the power grid, and automatically adjusts input parameters to reflect the dynamic changes of the power grid in real time;

[0013] A dynamic monitoring module monitors the real-time data of the power grid, analyzes and processes the real-time data using big data and machine learning, and then predicts the future power grid load and renewable energy power generation to obtain prediction data;

[0014] The intelligent calibration unit includes:

[0015] A scenario simulation module uses artificial intelligence to monitor the real-time data of the operating power grid in real time, automatically identifies potential fault data in the real-time data, simulates multiple fault scenarios based on the fault data, and performs dynamic safety verification;

[0016] The strategy formulation module formulates a unit output dispatch strategy based on the real-time data obtained by real-time monitoring and the fault data identified by the scenario simulation module using integrated learning and neural network algorithms.

[0017] Furthermore, the flexible optimization unit includes:

[0018] A dynamic prediction module, which uses artificial intelligence to dynamically schedule access points of renewable energy sources based on the prediction data and the fault data;

[0019] An automatic optimization module, based on deep reinforcement learning and optimization algorithms, intelligently dispatches the spare capacity of the renewable energy access point, and optimizes the output dispatch strategy of the unit in combination with big data analysis;

[0020] The multi-level simulation unit includes:

[0021] The co-evolution module solves the coordination problem of the renewable energy access points between multiple generator sets through the co-evolution algorithm, and uses artificial intelligence to simulate the synergy between the various generator sets;

[0022] The iterative learning module simulates and learns the synergistic effect between the units through artificial intelligence to obtain simulation data, identifies the static and dynamic safety of each unit based on the simulation data, and automatically learns and optimizes after obtaining the identification data to accelerate the iteration progress.

[0023] Furthermore, the workflow of the dynamic adaptive unit is:

[0024] The adaptive module collects the real-time data from the power grid, uses the Kalman filter algorithm to remove the influence of sensor errors on the real-time data, uses the mean filling method to interpolate the missing data in the real-time data to obtain pre-processed data, and then analyzes the pre-processed data based on the online learning algorithm to obtain analysis data, and automatically adjusts the power grid input parameters according to the analysis data to make it more suitable for the current state of the power grid. The state transfer equation is: X k =QX k-1 +WU k +E k , where X k is the system state vector, Q is the state transfer matrix representing the dynamic changes of the power grid, and U k is the control input, E k is the process noise, and the observation update formula is: R k =TX k +Y k , where R kk is the power grid measurement value, T is the observation matrix, Y k is the observation noise, and the state update formula is: Among them, k-1 is the covariance matrix of the previous step, P is the observation noise covariance, I k is the Kalman gain, O k-1 is the estimated error covariance matrix at the previous time step k-1, is the output result, is the state estimate of the previous moment, R k is the observation value at time step k, T is the observation matrix, is the difference between the observed and predicted values, T S is the transpose of the observation matrix, P is the observation noise covariance matrix, TO k-1 T S +P is the inconsistency between system prediction and observation, mean filling method formula: Where D is the number of valid data, i is the missing value A at time point i i The estimated value after filling, A i is the valid i-th data value in the data set, It is the sum of the effective data.

[0025] The dynamic monitoring module, combined with the real-time data of the monitored power grid, uses time series to predict the power grid load, calculates the changes in power grid load demand in the future, and uses machine learning to predict the power generation of renewable energy, and combines the predicted value with the load forecast to predict the supply and demand balance of the power grid. The time series prediction formula is:

[0026] where f s , L s , v s are the activation functions of the input gate, forget gate, and output gate, respectively. s and b s are the cell state and output, respectively, x s is the historical data of power grid load, j s-1 is the output of the previous time step, G i is the weight matrix of the input gate, H i is the influence of the hidden state of the previous moment on the current input gate, k i is the bias term of the input gate, The activation function output value is between 0 and 1. is a candidate memory cell, G Z , H Z , k Z are the weights and biases of the candidate memory cells, Z s is the cell state at the current moment, L s ×Z s-1 is the information retained in the previous cell state, is the update effect of the current input information on the current state, G v , H v , k v is the weight and bias term of the output gate, b s is the hidden state at the current moment, tanh(Z s ) is the value of the cell state after the tanh activation function is transformed, and the prediction output formula is: Machine learning model formula: Where N re is the final output of the model, α q is the weight of the qth base learner, β q (x) is the output of the qth base learner, x is the feature vector of the input data, and ∪ is the total number of base learners.

[0027] Furthermore, the workflow of the intelligent verification unit is as follows:

[0028] The scenario simulation module detects abnormalities in the real-time data of the power grid, obtains abnormal data, and uses the AC power flow formula of the power grid to simulate various fault conditions, and calculates the power flow under the simulated fault scenario. At the same time, the decision tree formula is used to detect the fault condition. The active power formula of the AC power flow equation of the power grid is: Where W ij is the active power from node i to node j, E ia is the admittance from node i to node a, |R i |,|R a | are the voltage amplitudes of nodes i and a, respectively, and χ i , χ a are the phase angles of node i and node a respectively, sin(χ i -x a ) is the sinusoidal function of the phase angle difference between node i and node a, B is the total number of nodes in the power grid, and the reactive power formula of the power grid AC power flow equation is: Among them, M ij is the reactive power from node i to node a, E ia is the admittance from node i to node a, |R i |,|R a | are the voltage amplitudes of nodes i and a, respectively, and χ i , χ a are the phase angles of node i and node a respectively, cos(χ i -x a ) is the cosine function of the phase angle difference between node i and node a, B is the number of nodes in the power grid, and the decision tree formula is: in is the predicted output, δ i is the weight corresponding to the i-th feature, ε i (h) is the output value of the ith base learner, υ is the total number of features in the model, and Sign[·] is the sign function;

[0029] The strategy formulation module optimizes the unit output by using an integrated learning algorithm according to the real-time data and the fault data identified by the scenario simulation module, dynamically adjusts the unit scheduling strategy according to the real-time load and power generation conditions, and finally outputs the unit scheduling strategy.

[0030] Furthermore, the workflow of the flexible optimization unit is:

[0031] The dynamic prediction module collects the prediction data and the fault data, uses regression algorithms and deep neural networks to predict the future renewable energy generation, and dynamically adjusts the access point of renewable energy generation according to the prediction results, optimizes the load distribution of the power grid, and prioritizes the dispatch of renewable energy generation in combination with the load demand of the power grid. The renewable energy generation prediction formula is: in is the slack variable, A is the penalty factor, γ is the normal vector of the classification hyperplane, |γ 2 is the square norm of the γ vector, and N is the total number of training samples;

[0032] The automatic optimization module calculates the reserve capacity demand of the power grid based on the intelligent scheduling of the reserve capacity of the renewable energy access point, and uses deep reinforcement learning and particle swarm optimization to intelligently schedule the reserve capacity and unit output. The deep reinforcement learning update formula is:

[0033] Z(B v , Y v ) is the Z value of v at the current moment, B v is the state of v at the current moment, Y v is the action taken at time v, d v is the immediate reward obtained at time v, ι is the discount factor used to control the importance of future rewards, η is the speed at which the learning rate controls the value update, and the spare capacity is calculated as: Among them J re is the reserve capacity, i.e. the additional power capacity required by the grid, W i is the active power of the i-th unit, represents the power generation of each generator set in the power grid at the current moment, ZS is the total number of all running units in the power grid, W ae is the power capacity of the power grid that can currently meet the load demand, |·| is the absolute value of the total power, J re =max(·, 0) is the maximum value function, and the particle swarm optimization update formula is: in is the position of particle i at time s, representing the current solution of the particle, is the velocity of the particle at time s, is the new position of the particle at time s+1, i is the updated solution, m is the inertia weight that controls the influence of the particle's current velocity during the update, g1 and g2 are learning factors that control the particle to its optimal position LZ i and the degree of attraction of the global optimal position j, h1, h2 are random numbers between [0, 1], LZ i is the historical optimal position of the particle, and j is the historical optimal position of the group.

[0034] Furthermore, the workflow of the multi-level simulation unit is:

[0035] The collaborative evolution module uses a genetic algorithm to optimize the joint scheduling scheme of multiple units based on the joint adjustment of the renewable energy access points between multiple generator sets and the synergy between the units, and considers the interactive effects of load adjustment, fuel efficiency and startup costs between the units. The genetic algorithm optimization formula is: Where W i is the output power of the unit, m i , n i ,u i is a coefficient related to the unit characteristics;

[0036] The iterative learning module identifies the static and dynamic safety of each unit based on the simulation data, simulates the operation status of the power grid under different loads, faults and environmental conditions, uses the simulated annealing algorithm to perform static and dynamic safety analysis on the power grid, identifies the potential safety risks of the power grid, and automatically optimizes the system model based on the simulation results. The simulated annealing update rule formula is: where ΔQ is the energy difference between the current solution and the new solution, S j is the current temperature, S j+1 is the temperature of the next step, κ is the temperature attenuation coefficient, is the mathematical form of the probability of acceptance, GL(ΔQ, S j ) is the probability of accepting the new solution.

[0037] On the other hand, the present invention also provides a method for optimizing power grid operation based on artificial intelligence, which comprises the following steps:

[0038] S1. Collecting real-time data, and predicting subsequent power grid operation dynamics based on the real-time data, thereby obtaining prediction data;

[0039] S2. Perform dynamic security verification and formulate strategies by combining the predicted data with artificial intelligence;

[0040] S3. Dynamically schedule access points according to the predicted data, and continuously optimize the formulated strategy;

[0041] S4. Simulate the synergy effect between different units according to the dynamic scheduling access point.

[0042] Furthermore, in step S1, real-time data is collected, and the subsequent power grid operation dynamics are predicted based on the real-time data, so as to obtain the predicted data by:

[0043] According to the real-time data collected from the power grid, a power grid load prediction model and a renewable energy prediction model are established, and input parameters are automatically adjusted to reflect the dynamic changes of the power grid in real time;

[0044] By monitoring the real-time data of the power grid, analyzing and processing the real-time data using big data and machine learning methods, and then predicting the future power grid load and renewable energy power generation to obtain prediction data;

[0045] In step S2, the method of performing dynamic security verification and formulating strategies by combining artificial intelligence with the predicted data is as follows:

[0046] Using artificial intelligence to monitor the real-time data of the running power grid in real time, automatically identifying potential fault data in the real-time data, simulating multiple fault scenarios based on the fault data, and performing dynamic safety verification;

[0047] Based on the real-time data obtained by real-time monitoring and the fault data identified by the scenario simulation module, an integrated learning and neural network algorithm is used to formulate a unit output scheduling strategy.

[0048] Furthermore, in step S3, the method of dynamically scheduling access points and continuously optimizing the strategy formulation is as follows:

[0049] Dynamically dispatching access points of renewable energy using artificial intelligence according to the prediction data and the fault data;

[0050] Based on deep reinforcement learning and optimization algorithms, the spare capacity of the renewable energy access point is intelligently dispatched, and the output dispatch strategy of the unit is optimized in combination with big data analysis;

[0051] In step S4, according to the dynamic scheduling access point, the method for simulating the synergy effect between different units is:

[0052] Through the collaborative evolution algorithm, the coordination problem of the renewable energy access points between multiple generator sets is solved, and artificial intelligence is used to simulate the synergy between the various units;

[0053] The synergistic effect between the units is simulated and learned through artificial intelligence to obtain simulation data. The static and dynamic safety of each unit is identified based on the simulation data. After the identification data is obtained, automatic learning and optimization are carried out to accelerate the iteration progress.

[0054] Beneficial Effects

[0055] Compared with the known public technology, the technical solution provided by the present invention has the following advantages:

[0056] Beneficial effects:

[0057] When in use, the present invention is conducive to reducing the uncertainty in power grid dispatching, optimizing the dispatching strategy of generator sets, reducing energy consumption and operating costs, and can identify possible load fluctuations or failures in the power grid in advance through the dynamic monitoring module, which is conducive to improving the reliability of the power grid, discovering potential risks in advance, avoiding power grid collapse or large-scale failures, and formulating more accurate unit dispatching strategies based on the real-time and historical data of the power grid, improving the operating efficiency and economy of the power grid, providing effective support for power grid fault prediction, and being able to quickly respond to power grid changes and reduce the risk of failures.

[0058] When used, the present invention is conducive to improving the utilization rate of renewable energy, ensuring that the power grid can flexibly respond to load fluctuations and uncertainties of renewable energy, and is conducive to giving full play to the synergistic effect of multiple generator sets, improving overall power generation efficiency, enhancing simulation accuracy and accelerating the power grid optimization process, and reducing human intervention. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 A system diagram of a power grid operation optimization system based on artificial intelligence according to the present invention;

[0060] Figure 2 The present invention is a flow chart of a method for optimizing power grid operation based on artificial intelligence. DETAILED DESCRIPTION

[0061] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0062] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0063] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0064] Embodiment 1:

[0065] like Figure 1 As shown, the present invention provides an artificial intelligence-based power grid operation optimization system, comprising: a dynamic adaptive unit, collecting real-time data, predicting subsequent power grid operation dynamics according to the real-time data, thereby obtaining prediction data;

[0066] An adaptive module, which establishes a power grid load prediction model and a renewable energy prediction model according to the real-time data collected from the power grid, and automatically adjusts input parameters to reflect the dynamic changes of the power grid in real time;

[0067] A dynamic monitoring module monitors the real-time data of the power grid, analyzes and processes the real-time data using big data and machine learning, and then predicts the future power grid load and renewable energy power generation to obtain prediction data;

[0068] The workflow of the dynamic adaptive unit is:

[0069] The adaptive module collects the real-time data from the power grid, uses the Kalman filter algorithm to remove the influence of sensor errors on the real-time data, uses the mean filling method to interpolate the missing data in the real-time data to obtain pre-processed data, and then analyzes the pre-processed data based on the online learning algorithm to obtain analysis data, and automatically adjusts the power grid input parameters according to the analysis data to make it more suitable for the current state of the power grid. The state transfer equation is: X k =QX k-1 +WU k +E k , where X k is the system state vector, Q is the state transfer matrix representing the dynamic changes of the power grid, and U k is the control input, E k is the process noise, and the observation update formula is: R k =TX k +Y k , where R kk is the power grid measurement value, T is the observation matrix, Y k is the observation noise, and the state update formula is: Among them, k-1 is the covariance matrix of the previous step, P is the observation noise covariance, I k is the Kalman gain, O k-1 is the estimated error covariance matrix at the previous time step k-1, is the output result, is the state estimate of the previous moment, R k is the observation value at time step k, T is the observation matrix, is the difference between the observed and predicted values, T S is the transpose of the observation matrix, P is the observation noise covariance matrix, TO k-1 T S +P is the inconsistency between system prediction and observation, mean filling method formula: Where D is the number of valid data, i is the missing value A at time point i i The estimated value after filling, A i is the valid i-th data value in the data set, It is the sum of the effective data.

[0070] The dynamic monitoring module, combined with the real-time data of the monitored power grid, uses time series to predict the power grid load, calculates the changes in power grid load demand in the future, and uses machine learning to predict the power generation of renewable energy, and combines the predicted value with the load forecast to predict the supply and demand balance of the power grid. The time series prediction formula is:

[0071] where f s , L s , v s are the activation functions of the input gate, forget gate, and output gate, respectively. s and b s are the cell state and output, respectively, x s is the historical data of power grid load, j s-1 is the output of the previous time step, G i is the weight matrix of the input gate, H i is the influence of the hidden state of the previous moment on the current input gate, k i is the bias term of the input gate, The activation function output value is between 0 and 1. is a candidate memory cell, G Z , H Z , k Z are the weights and biases of the candidate memory cells, Z s is the cell state at the current moment, L s ×Z s-1 is the information retained in the previous cell state, is the update effect of the current input information on the current state, G v , H v , k v is the weight and bias term of the output gate, b s is the hidden state at the current moment, tanh(Z s ) is the value of the cell state after the tanh activation function is transformed, and the prediction output formula is: Machine learning model formula: Where N re is the final output of the model, α q is the weight of the qth base learner, β q (x) is the output of the qth base learner, x is the feature vector of the input data, and ∪ is the total number of base learners;

[0072] Specifically, in a certain area, the load demand for the next week is predicted through the power grid load forecasting model, and the wind power generation in the area is predicted based on the renewable energy forecasting model. After combining these predicted values, a supply and demand balance optimization strategy is generated for scheduling different types of generator sets. Through high-precision load and power generation forecasts, it is helpful to reduce the uncertainty in power grid scheduling, optimize the scheduling strategy of generator sets, reduce energy consumption and operating costs, and the dynamic monitoring module can identify possible load fluctuations or failures in the power grid in advance, which is beneficial to improve the reliability of the power grid.

[0073] An intelligent verification unit, which combines the predicted data with artificial intelligence to conduct dynamic security verification and formulate strategies;

[0074] The intelligent calibration unit includes:

[0075] A scenario simulation module uses artificial intelligence to monitor the real-time data of the operating power grid in real time, automatically identifies potential fault data in the real-time data, simulates multiple fault scenarios based on the fault data, and performs dynamic safety verification;

[0076] A strategy formulation module, which formulates a unit output dispatch strategy using integrated learning and neural network algorithms based on the real-time data obtained by real-time monitoring and the fault data identified by the scenario simulation module;

[0077] The workflow of the intelligent verification unit is as follows:

[0078] The scenario simulation module detects abnormalities in the real-time data of the power grid, obtains abnormal data, and uses the AC power flow formula of the power grid to simulate various fault conditions, and calculates the power flow under the simulated fault scenario. At the same time, the decision tree formula is used to detect the fault condition. The active power formula of the AC power flow equation of the power grid is: Where W ij is the active power from node i to node j, E ia is the admittance from node i to node a, |R i |,|R a | are the voltage amplitudes of nodes i and a, respectively, and χ i , χ a are the phase angles of node i and node a respectively, sin(χ i -x a) is the sinusoidal function of the phase angle difference between node i and node a, B is the total number of nodes in the power grid, and the reactive power formula of the power grid AC power flow equation is: Among them, M ij is the reactive power from node i to node a, E ia is the admittance from node i to node a,

[0079] |R i |,|R a | are the voltage amplitudes of nodes i and a, respectively, and χ i , χ a are the phase angles of node i and node a respectively, cos(χ i -x a ) is the cosine function of the phase angle difference between node i and node a, B is the number of nodes in the power grid, and the decision tree formula is: in is the predicted output, δ i is the weight corresponding to the i-th feature, ε i (h) is the output value of the ith base learner, υ is the total number of features in the model, and Sign[·] is the sign function;

[0080] A strategy formulation module, which optimizes the unit output by using an integrated learning algorithm according to the real-time data and the fault data identified by the scenario simulation module, and dynamically adjusts the unit dispatch strategy according to the real-time load and power generation conditions, and finally outputs the unit dispatch strategy;

[0081] Specifically, in a certain power grid, through the scenario simulation module, the working status of the power grid is monitored in real time, data from each node is collected, and anomaly detection is performed, including but not limited to discovering that the output power of a wind turbine has dropped significantly due to changes in wind speed. The system calculates the load distribution in other areas of the power grid and determines the type of fault occurring in the area. At the same time, through the formulation of a strategy module, the unit scheduling is optimized based on historical data, and the output of the standby unit is adjusted in time, which is conducive to discovering potential risks in advance and avoiding power grid collapse or large-scale failures. A more accurate unit scheduling strategy is formulated based on the real-time and historical data of the power grid, which improves the operating efficiency and economy of the power grid, provides effective support for power grid fault prediction, and can quickly respond to power grid changes and reduce the risk of failures.

[0082] A flexible optimization unit dynamically schedules access points according to the prediction data and continuously optimizes the formulated strategy;

[0083] The flexible optimization unit includes:

[0084] A dynamic prediction module, which uses artificial intelligence to dynamically schedule access points of renewable energy sources based on the prediction data and the fault data;

[0085] An automatic optimization module, based on deep reinforcement learning and optimization algorithms, intelligently dispatches the spare capacity of the renewable energy access point, and optimizes the output dispatch strategy of the unit in combination with big data analysis;

[0086] The workflow of the flexible optimization unit is:

[0087] The dynamic prediction module collects the prediction data and the fault data, uses regression algorithms and deep neural networks to predict the future renewable energy generation, and dynamically adjusts the access point of renewable energy generation according to the prediction results, optimizes the load distribution of the power grid, and prioritizes the dispatch of renewable energy generation in combination with the load demand of the power grid. The renewable energy generation prediction formula is: in is the slack variable, A is the penalty factor, γ is the normal vector of the classification hyperplane, |γ| 2 is the square norm of the γ vector, and N is the total number of training samples;

[0088] The automatic optimization module calculates the reserve capacity demand of the power grid based on the intelligent scheduling of the reserve capacity of the renewable energy access point, and uses deep reinforcement learning and particle swarm optimization to intelligently schedule the reserve capacity and unit output. The deep reinforcement learning update formula is:

[0089] Z(B v , Y v ) is the Z value of v at the current moment, B v is the state of v at the current moment, Y v is the action taken at time v, d v is the immediate reward obtained at time v, ι is the discount factor used to control the importance of future rewards, η is the speed at which the learning rate controls the value update, and the spare capacity is calculated as: Among them J re is the reserve capacity, i.e. the additional power capacity required by the grid, W i is the active power of the i-th unit, represents the power generation of each generator set in the power grid at the current moment, ZS is the total number of all running units in the power grid, W ae is the power capacity of the power grid that can currently meet the load demand, |·| is the absolute value of the total power, J re =max(·, 0) is the maximum value function, and the particle swarm optimization update formula is: in is the position of particle i at time s, representing the current solution of the particle, is the velocity of the particle at time s, is the new position of the particle at time s+1, i is the updated solution, m is the inertia weight that controls the influence of the particle's current velocity during the update, g1 and g2 are learning factors that control the particle to its optimal position LZ i and the degree of attraction of the global optimal position j, h1, h2 are random numbers between [0, 1], LZ i is the historical optimal position of the particle, j is the historical optimal position of the group;

[0090] A multi-level simulation unit simulates the synergy between different units according to the dynamic scheduling access point;

[0091] The multi-level simulation unit includes:

[0092] The co-evolution module solves the coordination problem of the renewable energy access points between multiple generator sets through the co-evolution algorithm, and uses artificial intelligence to simulate the synergy between the various generator sets;

[0093] An iterative learning module simulates and learns the synergistic effects between the units through artificial intelligence to obtain simulation data, identifies the static and dynamic safety of each unit based on the simulation data, and automatically learns and optimizes after obtaining the identification data, thereby accelerating the iterative progress;

[0094] The workflow of the multi-level simulation unit is:

[0095] The collaborative evolution module uses a genetic algorithm to optimize the joint scheduling scheme of multiple units based on the joint adjustment of the renewable energy access points between multiple generator sets and the synergy between the units, and considers the interactive effects of load adjustment, fuel efficiency and startup costs between the units. The genetic algorithm optimization formula is: Where W i is the output power of the unit, m i , n i ,u i is a coefficient related to the unit characteristics;

[0096] The iterative learning module identifies the static and dynamic safety of each unit based on the simulation data, simulates the operation status of the power grid under different loads, faults and environmental conditions, uses the simulated annealing algorithm to perform static and dynamic safety analysis on the power grid, identifies the potential safety risks of the power grid, and automatically optimizes the system model based on the simulation results. The simulated annealing update rule formula is: where ΔQ is the energy difference between the current solution and the new solution, S j is the current temperature, S j+1 is the temperature of the next step, κ is the temperature attenuation coefficient, is the mathematical form of the probability of acceptance, GL(ΔQ, S j) is the probability of accepting the new solution;

[0097] Specifically, in a multi-unit power system, the collaborative evolution module can establish a multi-unit synergy effect model based on the output power, fuel efficiency and startup cost of the multiple units, and perform optimized scheduling. Through the iterative learning module, it can conduct safety analysis of the power grid under different load and fault conditions, automatically identify and optimize the operation strategy of the power grid, which is conducive to giving full play to the synergy of multiple generating units, improving the overall power generation efficiency, enhancing the simulation accuracy, accelerating the power grid optimization process, and reducing human intervention.

[0098] Embodiment 2:

[0099] like Figure 2 As shown, embodiment 2 provides a method for optimizing power grid operation based on artificial intelligence, which includes the following steps:

[0100] S1. Collecting real-time data, and predicting subsequent power grid operation dynamics based on the real-time data, thereby obtaining prediction data;

[0101] S2. Perform dynamic security verification and formulate strategies by combining the predicted data with artificial intelligence;

[0102] S3. Dynamically schedule access points according to the predicted data, and continuously optimize the formulated strategy;

[0103] S4. Simulate the synergy effect between different units according to the dynamic scheduling access point.

[0104] Furthermore, in step S1, real-time data is collected, and the subsequent power grid operation dynamics are predicted based on the real-time data, so as to obtain the predicted data by:

[0105] According to the real-time data collected from the power grid, a power grid load prediction model and a renewable energy prediction model are established, and input parameters are automatically adjusted to reflect the dynamic changes of the power grid in real time;

[0106] By monitoring the real-time data of the power grid, analyzing and processing the real-time data using big data and machine learning methods, and then predicting the future power grid load and renewable energy power generation to obtain prediction data;

[0107] In step S2, the method of performing dynamic security verification and formulating strategies by combining artificial intelligence with the predicted data is as follows:

[0108] Using artificial intelligence to monitor the real-time data of the running power grid in real time, automatically identifying potential fault data in the real-time data, simulating multiple fault scenarios based on the fault data, and performing dynamic safety verification;

[0109] Based on the real-time data obtained by real-time monitoring and the fault data identified by the scenario simulation module, an integrated learning and neural network algorithm is used to formulate a unit output scheduling strategy.

[0110] Furthermore, in step S3, the method of dynamically scheduling access points and continuously optimizing the strategy formulation is as follows:

[0111] Dynamically dispatching access points of renewable energy using artificial intelligence according to the prediction data and the fault data;

[0112] Based on deep reinforcement learning and optimization algorithms, the spare capacity of the renewable energy access point is intelligently dispatched, and the output dispatch strategy of the unit is optimized in combination with big data analysis;

[0113] In step S4, according to the dynamic scheduling access point, the method for simulating the synergy effect between different units is:

[0114] Through the collaborative evolution algorithm, the coordination problem of the renewable energy access points between multiple generator sets is solved, and artificial intelligence is used to simulate the synergy between the various units;

[0115] The synergistic effect between the units is simulated and learned through artificial intelligence to obtain simulation data. The static and dynamic safety of each unit is identified based on the simulation data. After the identification data is obtained, automatic learning and optimization are carried out to accelerate the iteration progress.

[0116] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A power grid operation optimization system based on artificial intelligence, characterized in that: include: A dynamic adaptive unit collects real-time data and predicts subsequent power grid operation dynamics based on the real-time data, thereby obtaining prediction data; An intelligent verification unit, which combines the predicted data with artificial intelligence to conduct dynamic security verification and formulate strategies; A flexible optimization unit dynamically schedules access points according to the prediction data and continuously optimizes the formulated strategy; The multi-level simulation unit simulates the synergistic effect between different units according to the dynamic scheduling access point.

2. The power grid operation optimization system based on artificial intelligence according to claim 1, characterized in that: The dynamic adaptive unit includes: An adaptive module, which establishes a power grid load prediction model and a renewable energy prediction model according to the real-time data collected from the power grid, and automatically adjusts input parameters to reflect the dynamic changes of the power grid in real time; A dynamic monitoring module monitors the real-time data of the power grid, analyzes and processes the real-time data using big data and machine learning, and then predicts the future power grid load and renewable energy power generation to obtain prediction data; The intelligent calibration unit includes: A scenario simulation module uses artificial intelligence to monitor the real-time data of the operating power grid in real time, automatically identifies potential fault data in the real-time data, simulates multiple fault scenarios based on the fault data, and performs dynamic safety verification; The strategy formulation module formulates a unit output dispatch strategy based on the real-time data obtained by real-time monitoring and the fault data identified by the scenario simulation module using integrated learning and neural network algorithms.

3. The power grid operation optimization system based on artificial intelligence according to claim 2 is characterized in that: The flexible optimization unit includes: A dynamic prediction module, which uses artificial intelligence to dynamically schedule access points of renewable energy sources based on the prediction data and the fault data; An automatic optimization module, based on deep reinforcement learning and optimization algorithms, intelligently dispatches the spare capacity of the renewable energy access point, and optimizes the output dispatch strategy of the unit in combination with big data analysis; The multi-level simulation unit includes: The co-evolution module solves the coordination problem of the renewable energy access points between multiple generator sets through the co-evolution algorithm, and uses artificial intelligence to simulate the synergy between the various generator sets; The iterative learning module simulates and learns the synergistic effect between the units through artificial intelligence to obtain simulation data, identifies the static and dynamic safety of each unit based on the simulation data, and automatically learns and optimizes after obtaining the identification data to accelerate the iteration progress.

4. The power grid operation optimization system based on artificial intelligence according to claim 3 is characterized in that: The workflow of the dynamic adaptive unit is: The adaptive module collects the real-time data from the power grid, uses the Kalman filter algorithm to remove the influence of sensor errors on the real-time data, uses the mean filling method to interpolate the missing data in the real-time data to obtain pre-processed data, and then analyzes the pre-processed data based on the online learning algorithm to obtain analysis data, and automatically adjusts the power grid input parameters according to the analysis data to make it more suitable for the current state of the power grid. The state transfer equation is: X k =QX k-1 +WU k +E k , where X k is the system state vector, Q is the state transfer matrix representing the dynamic changes of the power grid, and U k is the control input, E k is the process noise, and the observation update formula is: R k =TX k +Y k , where R kk is the power grid measurement value, T is the observation matrix, Y k is the observation noise, and the state update formula is: Among them, k-1 is the covariance matrix of the previous step, P is the observation noise covariance, I k is the Kalman gain, O k-1 is the estimated error covariance matrix at the previous time step k-1, is the output result, is the state estimate of the previous moment, R k is the observation value at time step k, T is the observation matrix, is the difference between the observed and predicted values, T S is the transpose of the observation matrix, P is the observation noise covariance matrix, TO k-1 T S +P is the inconsistency between system prediction and observation, mean filling method formula: Where D is the number of valid data, i is the missing value A at time point i i The estimated value after filling, A i is the valid i-th data value in the data set, It is the sum of the effective data. The dynamic monitoring module, combined with the real-time data of the monitored power grid, uses time series to predict the power grid load, calculates the changes in power grid load demand in the future, and uses machine learning to predict the power generation of renewable energy, and combines the predicted value with the load forecast to predict the supply and demand balance of the power grid. The time series prediction formula is: where f s , L s , v s are the activation functions of the input gate, forget gate, and output gate, respectively. s and b s are the cell state and output, respectively, x s is the historical data of power grid load, j s-1 is the output of the previous time step, G i is the weight matrix of the input gate, H i is the influence of the hidden state of the previous moment on the current input gate, k i is the bias term of the input gate, The output value of the activation function is between 0 and 1, Z~ s is a candidate memory cell, G Z , H Z , k Z are the weights and biases of the candidate memory cells, Z s is the cell state at the current moment, L s ×Z s-1 is the information retained in the previous cell state, is the update effect of the current input information on the current state, G v , H v , k v is the weight and bias term of the output gate, b s is the hidden state at the current moment, tanh(Z s ) is the value of the cell state after the tanh activation function is transformed, and the prediction output formula is: Machine learning model formula: Where N re is the final output of the model, α q is the weight of the qth base learner, β q (x) is the output of the qth base learner, x is the feature vector of the input data, and ∪ is the total number of base learners.

5. The power grid operation optimization system based on artificial intelligence according to claim 4 is characterized in that: The workflow of the intelligent verification unit is as follows: The scenario simulation module detects abnormalities in the real-time data of the power grid, obtains abnormal data, and uses the power grid AC power flow formula to simulate various fault conditions, and calculates the power flow under the simulated fault scenario. At the same time, the decision tree formula is used to detect the fault condition. The active power formula of the power grid AC power flow equation is: Where W ij is the active power from node i to node j, E ia is the admittance from node i to node a, |R i |, R a | are the voltage amplitudes of nodes i and a, respectively, and χ i , χ a are the phase angles of node i and node a respectively, sin(χ i -x a ) is the sinusoidal function of the phase angle difference between node i and node a, B is the total number of nodes in the power grid, and the reactive power formula of the power grid AC power flow equation is: Among them, M ij is the reactive power from node i to node a, E ia is the admittance from node i to node a, R i |, R a | are the voltage amplitudes of nodes i and a, respectively, and χ i , χ a are the phase angles of node i and node a respectively, cos(χ i -x a ) is the cosine function of the phase angle difference between node i and node a, B is the number of nodes in the power grid, and the decision tree formula is: in is the predicted output, δ i is the weight corresponding to the i-th feature, ε i (h) is the output value of the ith base learner, υ is the total number of features in the model, and Sign[·] is the sign function; The strategy formulation module optimizes the unit output by using an integrated learning algorithm according to the real-time data and the fault data identified by the scenario simulation module, dynamically adjusts the unit scheduling strategy according to the real-time load and power generation conditions, and finally outputs the unit scheduling strategy.

6. The power grid operation optimization system based on artificial intelligence according to claim 5, characterized in that: The workflow of the flexible optimization unit is: The dynamic prediction module collects the prediction data and the fault data, uses regression algorithms and deep neural networks to predict the future renewable energy generation, and dynamically adjusts the access point of renewable energy generation according to the prediction results, optimizes the load distribution of the power grid, and prioritizes the dispatch of renewable energy generation in combination with the load demand of the power grid. The renewable energy generation prediction formula is: in is the slack variable, A is the penalty factor, γ is the normal vector of the classification hyperplane, |γ 2 is the square norm of the γ vector, and N is the total number of training samples; The automatic optimization module calculates the reserve capacity demand of the power grid based on the intelligent scheduling of the reserve capacity of the renewable energy access point, and uses deep reinforcement learning and particle swarm optimization to intelligently schedule the reserve capacity and unit output. The deep reinforcement learning update formula is: Z(B v , Y v ) is the Z value of v at the current moment, B v is the state of v at the current moment, Y v is the action taken at time v, d v is the immediate reward obtained at time v, ι is the discount factor used to control the importance of future rewards, η is the speed at which the learning rate controls the value update, and the spare capacity is calculated as: Among them J re is the reserve capacity, i.e. the additional power capacity required by the grid, W i is the active power of the i-th unit, represents the power generation of each generator set in the power grid at the current moment, ZS is the total number of all running units in the power grid, W ae is the power capacity of the power grid that can currently meet the load demand, |·| is the absolute value of the total power, J re =max(·, 0) is the maximum value function, and the particle swarm optimization update formula is: in is the position of particle i at time s, representing the current solution of the particle, is the velocity of the particle at time s, is the new position of the particle at time s+1, i is the updated solution, m is the inertia weight that controls the influence of the particle's current velocity during the update, g1 and g2 are learning factors that control the particle to its optimal position LZ i and the degree of attraction of the global optimal position j, h1, h2 are random numbers between [0, 1], LZ i is the historical optimal position of the particle, and j is the historical optimal position of the group.

7. The power grid operation optimization system based on artificial intelligence according to claim 6 is characterized in that: The workflow of the multi-level simulation unit is: The collaborative evolution module uses a genetic algorithm to optimize the joint scheduling scheme of multiple units based on the joint adjustment of the renewable energy access points between multiple generator sets and the synergy between the units, and considers the interactive effects of load adjustment, fuel efficiency and startup costs between the units. The genetic algorithm optimization formula is: Where W i is the output power of the unit, m i , n i ,u i is a coefficient related to the unit characteristics; The iterative learning module identifies the static and dynamic safety of each unit based on the simulation data, simulates the operation status of the power grid under different loads, faults and environmental conditions, uses the simulated annealing algorithm to perform static and dynamic safety analysis on the power grid, identifies the potential safety risks of the power grid, and automatically optimizes the system model based on the simulation results. The simulated annealing update rule formula is: where ΔQ is the energy difference between the current solution and the new solution, S j is the current temperature, S j+1 is the temperature of the next step, κ is the temperature attenuation coefficient, is the mathematical form of the probability of acceptance, GL(ΔQ, S j ) is the probability of accepting the new solution.

8. A method for optimizing power grid operation based on artificial intelligence, based on a system for optimizing power grid operation based on artificial intelligence according to any one of claims 1 to 7, characterized in that: The following steps are involved: S1. Collecting real-time data, and predicting subsequent power grid operation dynamics based on the real-time data, thereby obtaining prediction data; S2. Perform dynamic security verification and formulate strategies by combining the predicted data with artificial intelligence; S3. Dynamically schedule access points according to the predicted data, and continuously optimize the formulated strategy; S4. Simulate the synergy effect between different units according to the dynamic scheduling access point.

9. The method for optimizing power grid operation based on artificial intelligence according to claim 8, characterized in that: In step S1, real-time data is collected, and the subsequent power grid operation dynamics are predicted based on the real-time data, so that the method for obtaining the predicted data is: According to the real-time data collected from the power grid, a power grid load prediction model and a renewable energy prediction model are established, and input parameters are automatically adjusted to reflect the dynamic changes of the power grid in real time; By monitoring the real-time data of the power grid, analyzing and processing the real-time data using big data and machine learning methods, and then predicting the future power grid load and renewable energy power generation to obtain prediction data; In step S2, the method of performing dynamic security verification and formulating strategies by combining artificial intelligence with the predicted data is as follows: Using artificial intelligence to monitor the real-time data of the running power grid in real time, automatically identifying potential fault data in the real-time data, simulating multiple fault scenarios based on the fault data, and performing dynamic safety verification; Based on the real-time data obtained by real-time monitoring and the fault data identified by the scenario simulation module, an integrated learning and neural network algorithm is used to formulate a unit output scheduling strategy.

10. The method for optimizing power grid operation based on artificial intelligence according to claim 9, characterized in that: In step S3, the method of dynamically scheduling access points and continuously optimizing the strategy formulation is as follows: Dynamically dispatching access points of renewable energy using artificial intelligence according to the prediction data and the fault data; Based on deep reinforcement learning and optimization algorithms, the spare capacity of the renewable energy access point is intelligently dispatched, and the output dispatch strategy of the unit is optimized in combination with big data analysis; In step S4, according to the dynamic scheduling access point, the method for simulating the synergy effect between different units is: Through the collaborative evolution algorithm, the coordination problem of the renewable energy access points between multiple generator sets is solved, and artificial intelligence is used to simulate the synergy between the various units; The synergistic effect between the units is simulated and learned through artificial intelligence to obtain simulation data. The static and dynamic safety of each unit is identified based on the simulation data. After the identification data is obtained, automatic learning and optimization are carried out to accelerate the iteration progress.

Images