Research On Frequency Coordination Control Of Microgrid Based On Deep Reinforcement Learning

With the continuous increase of the proportion of new energy in the distribution network,while reducing carbon emissions and promoting carbon neutrality,the intermittent and volatility of new energy operation also bring a lot of security and stable operation of the distribution grid.The risk,microgrid technology provides an effective way for new energy such as wind and light.And how to maintain the internal voltage,frequency stability,and how to use the stable operation of the microgrid to support the distribution grid have become the focus of attention.Microgrids have two operating modes: islands and grid-connected.When the microgrid is in a state of lonely islands,drooping control is usually used to maintain the stability of its voltage and frequency through distributed power output power ratio.However,the distributed power supply is affected by the environmental region in the process of incorporating the microgrid.A large number of line impedance will be introduced,which causes the microgrid to be weak and has a non-power ring flow,which will affect the stability of the operation of the islanded microgrid.When the microgrid is in the grid-connected state,it usually needs to respond to the business needs of the distribution grid on time,such as participating in the second frequency adjustment task,etc.,and the microgrid is mainly based on new energy sources not only needs to cope with the load change during the grid process but also needs to be caused by load changes.Frequency fluctuations also need to solve the problem of frequency disturbance caused by the intermittent volatility of the new energy itself,thereby improving the ability to regulate the frequency adjustment of the microgrid support system.For the above-mentioned problems,based on the interaction of artificial intelligence algorithms with the power grid system,the process of controlling and optimizing the process is converted into a smart body training process.Dual-feed wind turbines participated in the second frequency adjustment control scheme of multiple regional interconnection grid systems under the running state of the microgrid.The main contributions of this article are as follows:1)Optimization control of the islanded microgrid based on deep strengthening learning:This article first controls a drooping control to control the distributed power supply in the microgrid network,so that the power is distributed as the assignment,and the system frequency is stabilized within the safe range.Secondly,the low-voltage micropower grid was analyzed by the problem of drooping control in the running state of the islands,and the virtual impedance method was introduced to optimize the drooping control of the drooping control.Finally,the virtual impedance method needs to be detected in real-time to detect the impedance resistance value of the environmental climate change,and it may increase the problem of the output voltage value of the inverter and the system frequency offset rate.Improve.By defining the status,rewards,and action functions in the algorithm,the intelligent body can automatically iterate the optimal virtual impedance value scheme according to the size of the non-meritocratic ring flow,and the adaptive update will be adaptable as it does not need to measure the impedance of the line;Add the constraints of the system frequency and voltage amplitude in the reward function of the algorithm,so that the smart body is optimized secondary and voltage amplitude in the iteration process.2)Based on dual-feed wind turbine secondary frequency control:When the microgrid is switched to the grid-connected operation,it is not enough to restrain frequency fluctuations by using the drooping control and using adaptive virtual impedance.It is necessary to use the dual feeders in the network to adjust the output of the output to restore the stability of the system frequency.Essence This article first introduces the mathematical model of the dual-feed and analyzes the vector control principle of its output power.Secondly,based on the frequency response model of traditional generators,a multi-regional interconnection system frequency response model containing non-linear factors is established,and the simplified fan control model is added to the frequency response model of multi-regional interconnection systems.Multiregional interconnection system frequency response modeling.Finally,for the system frequency fluctuations caused by load changes and fans torque disturbances,the dual-feeds and traditional generators based on dual-feeds and traditional generators based on deep strong learning algorithms are designed.At the same time that the constraint fan rotor speed changes are within the scope of the stable operation,the fan,and the traditional generator independently iterate the optimal force solution according to the frequency deviation of each region,that is,the output variable volume of the traditional generator and the rotor voltage of the dual feeder fans regulating amount adjustment amount The control goals have successfully participated in the second frequency adjustment control of multiple regional interconnection systems.