Hello everyone, I participate in a project in which we intend to forecast the demand of customers With installed and photovoltaic panels. These are the so-called microinstallations, i.e. their power does not exceed 40 kw. Due to the fact that the distribution system operator installs two-way meters that measure consumption and delivery, we are not able to find out what the customer's auto consumption level is without additional metering. When it comes to forecasting the demand of these customers, usually after the installation of photovoltaics, customers most often decide to install additional receivers so that they can consume the energy produced. Increasingly, these receivers are heat pumps. Thus, the demand of such customers changes dynamically with the change in temperature. To solve this problem, we believe that it will be appropriate to use, for example, linear regression with explanatory variables such as temperature, sunlight, humidity. As for forecasting what customers give to the grid, in this case an efficiency matrix will be built Determining how much energy is given away at a given temperature and sunlight Along with taking into account the Parameters of the sun, i.e. azimuth and Height of the sun. I would like to know what methods do you use to solve this type of task?
regards
Paweł
Dear Pawel
You can find correlation among variables to find exact temperature dependency on demand. However, this correlation will be location dependent. And, if you are having large database, you can use hybrid deep learning models, otherwise decision tree based machine learning models for demand forecasting.
In forecasting day-ahead power generation and consumption, you may use a deep learning model (LSTM, Bi-LSTM, RNN/CNN, etc), which gives you higher prediction accuracy.
Dear Paweł Machał :
I agree with Sir. Md Morshed Alam . You could use a deep learning classifier (MLP, LSTM, DNN, etc.) to forecast the generation and consumption of power for the next day ahead. This would give you a higher level of accuracy in your predictions.
Regards;
Ehtisham
Thanks for yours adwise. I have first data of my calculation - RMSE seems to bee high about 600 Watts. My pv instlation has 17 kW peak power installed. MAPE is also high about 450%. But I have simulated whole year and there was snow on panels also. Do you have better accurancy?
- Badges
- Science topic
- Similar topics
- Biological Science
- Bioecology
- Systems Ecology
- Landscape Ecology
- Connectivity