Algorithm Research Of Non-invasive Household Appliances Recognition

Load monitoring is very important for power system. In conventional load monitoring systems, the hardware circuits are installed on each load to be monitored. The load of invasive monitoring method in terms of installation and maintenance requires a lot of time and money, and hardware maintenance cost is higher. Therefore, researchers propose noninvasive load monitoring (NILM), need to installed at the entrance of the power monitoring equipment, monitoring the population place such as voltage, current signal can be decomposed system within a single category and load operation. For energy provider, NILM help power providers understand the user’s load, electricity usage and energy usage, strengthen the load power monitoring and management, the use of reasonable arrangement of the load time, such as poor regulating peak valley and reduce transmission losses; considering from the technology itself, help to improve the prediction accuracy of power load, simulation analysis for load monitoring system planning and provide more accurate data; for power users, through the NILM can analysis of the load data of energy consumption effectively, reduce unnecessary energy consumption, achieve the goal of saving energy and reducing consumption.The on-line monitoring household appliances power is of great importance for users of electricity management. Add noninvasive household appliances in smart meters power monitoring module, in order to meet online electricity management to provide effective and comprehensive data support. This article conducted the research of noninvasive simple load identification from three aspects. First of all, according to the air conditioning load in summer is the main energy-consuming household electricity load components, improvement based on k-means algorithm is applied to air conditioning load decomposition, Using edge detection and k-means clustering method to classify data, then the data will be used to determine the key parameters of conditioned behavior, this parameter is used to confirm the start and stop air conditioner event; Second extract load current parameters, selects the current maximum, average and mean square error as a characteristic parameter to identify load for simple identification, Load startup transient current waveforms can be acquired, several numerical transient excitation characteristics extracted from the acquired transient current waveform parameters associated with the three properties, Extracting a transient parameters, which were perfect training, identified as load recognition feature parameters, and then simulated to verify recognition results; Finally, according to the extract to the current and voltage waveform, more computational load characteristic parameters of weighted assignment method to complete the load type matching, Select the electricity load to simulate, the experimental data will be brought into the recognition algorithm, authenticate the algorithm accuracy and applicability. Specific work is as follows:(1)First detected load start-stop, according to the difference of current waveform, cutting load waveform figure, After the strength of each of the current cycle for difference operation, get the general transient time, And then extract the period the maximum load current, average and mean square error as the parameter setting load identification. Extract more load these three transient identification parameters, and simulation algorithm accuracy.(2)According to the steady state and transient characteristics of household electrical appliances, extract more feature parameters of the household electrical appliances. Experiment selected 16 kinds of household electrical appliances as a reference device, Steady state voltage and current waveform data sampling, calculating the characteristic parameters, characteristic parameters of the model library as electrical type recognition database.(3)Household appliances type identification algorithm is put forward. Select some electrical appliances simulation identification, analyze and calculate the voltage and current waveform data, the results prove the correctness of the identification algorithm, using the above method, Choose two household appliances as the mixed type recognition experiment. Results prove that the identification algorithm can identify multiple devices at the same time online operation success.